Sei sulla pagina 1di 186

Advances in Physics, 2002, Vol. 51, No.

1, 1 186

Intercalation compounds of graphite*


M. S. Dresselhaus{ and G. Dresselhaus{ Massachusetts Institute of Technology, Cambridge, MA 02139, USA
[Received 29 July 1980]

Abstract
A broad review of recent research work on the preparation and the remarkable properties of intercalation compounds of graphite, covering a wide range of topics from the basic chemistry, physics and materials science to engineering applications.

Contents
1. 2. Introduction Materials preparation and characterization 2.1. Intercalation methods 2.2. Materials characterization 2.3. Kinetics, thermodynamics and the intercalation process 2.4. The staging phenomenon Structural properties 3.1. Classication of types of crystalline order 3.2. X-ray, electron and neutron di raction for structural studies 3.3. Relation between the structure of intercalation compounds and their parent materials 3.4. Phase transitions Electronic properties 4.1. Electronic structure of graphite and graphite intercalation compounds 4.2. Transport properties 4.3. Optical properties 4.4. Electron spectroscopy studies 4.5. Fermi surface studies 4.6. Magneto-optical phenomena 4.7. Magnetic susceptibility, magnetic resonance and specic heat studies 4.8. Superconductivity Lattice mode spectra 5.1. Methods for studying lattice modes in intercalated graphite 5.2. Lattice modes in graphite 5.3. Raman scattering results on graphite intercalation compounds

page
2 5 5 16 24 31 36 36 42 46 55 65 65 85 103 114 118 127 133 148 151 151 153 155

3.

4.

5.

* This article was originally published in Advances in Physics, volume 30, 1981. It had attracted 876 citations by October 2001, and is ranked 11 in the index of articles attracting more than 100 citations. { Department of Electrical Engineering and Computer Science, and Center for Materials Science and Engineering. { Francis Bitter National Magnet Laboratory.
Advances in Physics ISSN 00018732 print/ISSN 14606976 online # 2002 Taylor & Francis Ltd http://www.tandf.co.uk/journals DOI: 10.1080 /0001873011011364 4

M. S. Dresselhaus and G. Dresselhaus


5.4. Infrared studies of the lattice modes 5.5. Inelastic neutron scattering studies 6. Technological applications of intercalated graphite 6.1. Introduction 6.2. Battery and electrode materials 6.3. Chemical catalytic applications 6.4. Conductivity applications 6.5. Carbon bres 6.6. Other applications 165 169 171 171 171 172 172 173 175 176 176

Acknowledgements References

1. Introduction Research on the preparation and properties of graphite intercalation compounds has recently undergone a resurgence of interest and our fundamental knowledge of the physics of these remarkable materials has increased substantially. It seems appropriate to review this subject now but, with research activity and interest in the eld remaining high, this review may well require updating in the near future. This is a broad review (an outgrowth of lecture notes developed for graduate students in our research group at MIT) of a wide range of topics from the basic chemistry and physics of intercalated graphite to engineering applications. It has not been possible therefore to include reference to all of the important papers that have been written in this eld (and we apologize to authors whose work would otherwise have merited inclusion). Graphite intercalation compounds are formed by the insertion of atomic or molecular layers of a di erent chemical species called the intercalant between layers in a graphite host material, as shown in gure 1. The intercalation compounds occur in highly anisotropic layered structures where the intraplanar binding forces are large in comparison with the interplanar binding forces. The most common examples of host materials for intercalation compounds are graphite and the transition metal dichalcogenides. Of the various types of intercalation compounds, the graphite compounds are of particular physical interest because of their relatively high degree of structural ordering. The most important and characteristic ordering property of graphite intercalation compounds is the staging phenomenon, which is characterized by intercalate layers that are periodically arranged in a matrix of graphite layers. Graphite intercalation compounds are thus classied by a stage index n denoting the number of graphite layers between adjacent intercalate layers, as is illustrated in gure 2. This staging phenomenon is a general phenomenon in graphite intercalation compounds, even in those samples with very dilute intercalate concentrations (n 10). Intercalation provides to the host material a means for controlled variation of many physical properties over wide ranges. Because the free carrier concentration of the graphite host is very low (104 free carriers/atom at room temperature), intercalation with di erent chemical species and concentrations permits wide variation of the free carrier concentration and thus of the electrical, thermal and magnetic properties of the host material. Of these properties, the e ect of intercalation on the electrical conductivity has probably attracted the greatest amount of attention because of the fabrication of an intercalation compound (CX AsF 5 ) with a reported room temperature conductivity exceeding that of copper

Intercalation compounds of graphite

Figure 1. Model for C8 K according to Rudor and Schulze (1954) showing the stacking of graphite layers (networks of small solid balls) and of potassium layers (networks of large hollow balls). The graphite and intercalate layers are arranged in an AAAA stacking sequence, where A refers to the graphite layers and the Greek letters to the intercalate layers.

(Foley et al. 1977). Perhaps even more striking is the range of electrical conductivity behaviour, ranging from almost insulating behaviour for the c-axis conductivity in certain acceptor compounds to superconducting in-plane behaviour below 1.0 K for the rst stage alkali metal donor compounds C8 K where neither of the parent chemical species individually exhibit superconductivity (Hanney et al. 1965, Koike et al. 1978). The large increase in conductivity in intercalated graphite results from a charge transfer from the intercalate layer where the carriers have a low mobility to the graphite layers where the mobility is high. Since the most signicant modications to the graphite involve graphite layers adjacent to the intercalate layer, it is convenient to distinguish between the graphite bounding layers adjacent to the intercalant, and the graphite interior layers that have only graphite nearest-neighbou r layers. The synthesis of a graphite intercalation compound was rst reported by Scha autl (1841). However, the rst systematic studies of these compounds began in the early 1930s with the introduction of X-ray di raction techniques for stage index determinations (Ho man and Frenzel 1931, Schleede and Wellman 1932). Though the systematic study of their physical properties began in the late 1940s, it is only in recent years that research on graphite intercalation compounds has become a eld of intense activity internationally. A large number (100) of reagents can be intercalated into graphite. These intercalants are commonly classied according to whether they form donor or acceptor compounds. The most common and most widely studied of the donor compounds are the alkali metal compounds with K, Rb, Cs and Li, though other

M. S. Dresselhaus and G. Dresselhaus

Figure 2. Schematic diagram illustrating the staging phenomenon in graphitepotassium compounds for stages 1 n 4. The potassium layers are indicated by dashed lines and the graphite layers by solid lines connecting open circles, and indicating schematically a projection of the carbon atom positions. The . . . ABAB . . . graphite layer stacking for stages n 2 is maintained between intercalate layers, although a rhombohedral stacking arrangement appears across intercalate layers. The stacking ordering is well-conrmed by X-ray di raction (00l) patterns. For each stage, the distance Ic between adjacent intercalate layers is indicated. For rst stage C8 K, the unit cell includes intercalate layers with stacking indices , , , (see gures 1 and 20).

donor intercalants are known, such as alkaline earth metals, lanthanides and metal alloys of these with each other or with alkali metals. Ternary donor intercalation compounds have also been prepared using alkali metals with hydrogen or polar molecules, such as ammonia and tetrahydrofuran , and aromatic molecules, such as benzene. A very large variety of acceptor compounds have also been prepared, and are often based on Lewis acid intercalants such as the halogen Br2 or halogen mixtures, metal chlorides, bromides, uorides and oxyhalides, acidic oxides such as N2 O5 and SO3 and strong Bronsted acids such as H2 SO4 and HNO3 . In the intercalation process, the molecular intercalants generally remain molecular in form. From gures 1 and 2 it is seen that intercalation causes crystal dilatation along the caxis: the larger the molecular intercalants, the larger the dilatation for compounds of comparable stage. In general, both chemical a nities and geometric constraints associated with intercalant size and intercalant bonding distances determine whether or not a given chemical species will intercalate. Many of these compounds are unstable in air, with donor compounds being easily oxidized and acceptors being easily desorbed. For this reason, most intercalation compounds require encapsulation to ensure chemical stability, though some compounds, such as graphiteFeCl3 and graphiteSbCl5 , are relatively stable in air. In addition to the large number of chemical species that can be intercalated, a number of di erent types of graphite host materials are used for each of the various applications. From a structural point of view, the simplest host material is a single crystal graphite ake, such as those separated from the limestone rocks found in the

Intercalation compounds of graphite

Ticonderoga mines of New York State. Because ake dimensions are 1 mm in diameter and only several hundredths of a millimetre in thickness, these materials often cannot be conveniently used for carrying out physical properties measurements. In such cases, samples of large physical dimensions based on highly oriented pyrolytic graphite (HOPG) are used (Moore 1973). HOPG is a synthetic graphite formed by cracking a hydrocarbon at high temperature and subsequent heat treatment, often combined with the application of pressure. The resulting material is highly oriented along the c-axis (orientational deviations less than 18) but in the layer planes consists of a randomly ordered collection of crystallites of 1 mm average diameter. For many physical measurements the greater exibility in sample size provided by the HOPG host material is of greater importance than the more perfect ordering of the single crystal akes. In fact, HOPG has been the most common host material for graphite intercalation compounds during the recent period of active research. Another type of graphite host material is kish graphite, obtained by the crystrallization of carbon from molten steel during the steel manufacturing process. Kish graphite samples typically contain several large single crystallites, exhibiting much higher structural ordering than HOPG, but not quite as ordered or as chemically pure as natural single crystal akes. On the other hand, kish graphite samples are normally an order of magnitude greater in area and in thickness when compared with single crystal akes. Though little use has so far been made of this host material, intercalation compounds based on kish graphite can also be prepared. Carbon bres represent yet another class of synthetic graphite with great mechanical strength, because the bre axis is along the graphite a-axis where the iteratomic bonding is extremely strong. The intercalation of carbon bres is under consideration as a method for variation of the electrical, mechanical and adhesive properties of this commercially important class of bre materials. In recent years several excellent review articles have appeared on the synthesis and structure of graphite intercalation compounds (Ebert 1986, Herold 1979), as well as on the electronic properties (Fischer 1979), the lattice properties (Dresselhaus and Dresselhaus 1979), and on applications areas (Whittingham and Ebert 1979). In addition, a large collection of research and review articles on these subjects have appeared in the Proceedings of the Franco-American Conference of La Napoule (1977, Mater. Sci. Engng, 31, 1). This review article on graphite intercalation compounds covers the eld broadly in an e ort to interrelate many of the factors which a ect the structural, electronic and lattice properties of these materials. Section 2 is devoted to the preparation and characterization of materials, as well as related topics such as intercalation kinetics and staging. Sections 3, 4 and 5 deal respectively with the structural electronic and lattice properties of graphite intercalation compounds, while some applications areas for these materials are briey discussed in section 6.

2.

Materials preparation and characterization 2.1. Intercalation methods

2.1.1. Introduction A number of general methods have been developed for the preparation of graphite intercalation compounds (Ebert 1976, Herold 1977, 1979) including the

M. S. Dresselhaus and G. Dresselhaus

two-zone vapour transport technique, the liquid intercalation method, the electrochemical method and cointercalation techniques. For the various techniques employed, the parameters of signicance are temperature, vapour pressure, the chemical and physical properties of the intercalant and the characteristics of the graphite host material. Small, thin samples intercalate more quickly and often yield better-staged, more homogeneous material than large, thick samples. Also single crystal graphite akes intercalate more readily than a highly oriented pyrolytic graphite (HOPG) specimen (Moore 1973), or than a carbon bre (Hooley 1977a). Although a given intercalation compound can often be prepared by alternative growth techniques, the physical and chemical properties of the intercalant play an important role in favouring one intercalation method over another. Excellent reviews of methods used for the preparation of specic intercalation compounds are given in the articles by Ebert (1976) and Herold (1977, 1979). Intercalation can be achieved starting from solid, liquid or gaseous reagents (Croft 1960, Holey 1977a, Herold 1977, 1979) though preparation using vapour transport with the two-zone method is the most common for the preparation of wellstaged specimens (Fredenhagen and Cadenbach 1926, Herold 1955). Intercalation occurs for many types of reagents (more than 100), ranging from simple ionic species such as alkali metals, diatomic molecules such as the halogens, metal chlorides, bromides, oxides and sulphides to large organic molecules such as benzene (Croft 1956, Ebert 1976, Stumpp 1977, He rold 1977, 1979). The simpler binary and ternary compounds are usually prepared by direct synthesis, and the more complicated materials by a variety of stepwise intercalation procedures. Because of the high reactivity of most graphite intercalation compounds, they are commonly prepared and stored in ampoules, containing either intercalate vapour, an overpressure of an inert gas or vacuum, depending on the intercalate species. Cooling samples (for example, to liquid nitrogen temperatures) greatly increases their stability. Certain intercalation compounds (for example, graphiteBr2 ), when removed from their encapsulating ampoules and exposed to the normal room temperature environment, will desorb for a period of time, after which there is negligible desorption. The resulting material, called a residue compound, is chemically stable and contains an intercalate concentration (usually small), dependent on the desorption temperature and on the intercalate concentration of the compound prior to desorption (Hennig 1952b). Lamellar compounds, denoting intercalation compounds that have not been desorbed, may be single-staged, multi-staged or have a random arrangement of the intercalate layers. 2.1.2. The two-zone vapour transport method In the two-zone vapour transport method, the intercalant is typically heated to some temperature Ti and the graphite, which is some distance away, is heated to a higher temperature Tg as illustrated in gure 3. The stage of the compound, or the

Figure 3. Schematic diagram of the two-zone vapour transport method where Tg and Ti indicate the temperature of the graphite and intercalant respectively.

Intercalation compounds of graphite

Figure 4. Isobar for growth of graphiteK compounds showing intercalant uptake versus temperature di erence Tg Ti between graphite and intercalant (from the work of Nixon and Parry 1968). Note the temperature ranges for staging stability, as indicated for stages n 1, 2, 3.

weight uptake of intercalant, is controlled by the temperature di erence Tg Ti , the smaller values of Tg Ti corresponding to compounds with lower stage. The region of stability for the formation of a given stage compound decreases with increasing temperature di erence Tg Ti (or equivalently with increasing stage index), as shown in gure 4. As a useful guide for the preparation of a stage n compound, listings of temperature di erences Tg Ti versus stage are available for a number of intercalants (for example, for K (Nixon and Parry 1968), Rb (Nixon 1966), Cs (Nixon 1966), Br2 (Sasa et al. 1971), FeCl3 (Hooley and Bartlett 1967)). In the actual sample preparation some adjustment of the operating temperatures is necessary to account for the dimensions and the defect density of the graphite specimen to be intercalated, the amount of intercalant in the reaction chamber and the geometric arrangement of the graphite relative to the intercalant (Hooley 1977a, b). Although each intercalant requires a special set of growth conditions, these conditions are closely related for intercalants with similar chemical properties. For example, the alkali metal donor compounds with the intercalants K, Rb and Cs are all prepared using a similar procedure, where Tg is varied to control the stage, and Ti is held constant, thereby giving rise to xed vapour pressure conditions for the growth process (Nixon and Parry 1968). Stage 1 samples are grown using very small di erences in Tg Ti , though Tg is maintained above Ti to prevent intercalant condensation on the sample. To achieve more dilute compounds, Tg is raised, keeping Ti constant, and the stage index is most sensitively determined by the temperature di erence Tg Ti, as shown in the intercalation isobar for the graphite K system of gure 4. Typical values for a variety of stages with the K, Rb and Cs intercalants are given in table 1 (Nixon 1966). From this table and from gure 4 it is seen that a given stage can be prepared for a range of Tg values. At least for the stage 1 compounds, more rapid intercalation proceeds when Tg is near the lower limit of the range, and slower intercalation at higher Tg . For xed Ti, the lower the stage, the larger the range of Tg over which that stage can be prepared as a single-staged compound. Donor intercalation requires su cient intercalant volatility and electropositivity. Moreover, very careful temperature control is necessary to prepare singlestaged materials with high stage index. The preparation of dilute, well-staged samples can also be carried out using the two-zone vapour transport technique, though in this case the stages that are achieved

M. S. Dresselhaus and G. Dresselhaus


Table 1. T i and Tg for preparation of alkali metal compounds with stages 1 n 3.a K Ti 2508C Stage 1 2 3
a

Rb Ti 2088C Tg 8C 215330 375430 450480

Cs Ti 1948C Tg 8C 200425 475530 550

Tg 8C 225320 350400 450480

From Nixon (1966).

depend on many parameters, including geometrical factors (sample size, such as thickness and cross-sectional area and ampoule size and shape) and the accuracy of temperature control (Hooley et al. 1965, Hooley 1977a). For the preparation of dilute alkali metal compounds, shaping the ampoule to minimize the free volume for intercalate vapour around the graphite, and necking down the connection between the graphite and intercalate zones increases the stage index of the sample (Underhill et al. 1980). The intercalation of small graphite samples promotes the formation of a single stage over the sample volume. For a given geometrical arrangement, the preparation of a dilute compound of specied stage is greatly aided by use of a growth diagram such as gure 4. An upper limit for Tg is imposed in practice by the softening of the encapsulating glass and by its increased reactivity with the alkali metal intercalants at high temperatures. A lower limit on Ti is imposed by the requirement of some minimum acceptable reaction rate and of the condition p > pt , where pt is the threshold pressure below which intercalation does not occur. It should be mentioned that although alkali metal compounds are typically grown under low vapour pressure (less than 1 torr) conditions, the intercalation rates are relatively rapid; for example, single-staged graphiteRb samples for n 8 can be prepared in less than 24 hours (Underhill et al. 1980). Preparation of a variety of single-staged acceptor compounds is also carried out using the two-zone technique, though the detailed growth conditions are quite di erent from the donor compounds and vary from one class of acceptors to another. In the case of acceptors, the intercalant is in molecular form, and normally a much larger c-axis expansion of the graphite host is required to accommodate the intercalant. For growth of acceptor compounds by the two-zone method, the graphite temperature Tg is typically held constant, and the intercalant, at a lower temperature Ti is varied to produce the desired stage (Hooley 1973). The acceptor compounds typically have a high threshold vapour pressure and therefore are prepared under high vapour pressure conditions. The metal chloride acceptor compounds with AlCl3 and FeCl3 are both prepared in a similar way (Hooley 1973). The metal chloride intercalant is rst produced in situ by direct reaction of the heated metal wire with Cl2 gas, as shown in gure 5. Prior to intercalation, the two-zone ampoule is back-lled with Cl2 gas to encourage staging (Dzurus and Hennig 1957b, Metz and Hohlwein 1975b); without the Cl2 gas, only weight uptake is achieved and staging is inhibited (Rudor and Zeller 1955, Hooley 1972, 1973, Underhill et al. 1979). The presence of Cl2 gas is also necessary for the

Intercalation compounds of graphite

Figure 5. Schematic diagram of a system used for the preparation of graphiteFeCl3 . (a) System used to prepare crystalline FeCl3 in situ by passing Cl2 gas over a heated Fe wire. This closed system is advantageous because FeCl3 is highly hygroscopic. (b) Two-zone ampoule containing a highly oriented pyrolytic graphite sample (HOPG) in one zone and the crystalline FeCl3 to be intercalated in the other zone.

preparation of many other metal chloride intercalation compounds (Stumpp 1977), generally increasing the amount of intercalant uptake, reducing the threshold pressure, and encouraging good staging to occur. Whether or not there is a signicant uptake of excess chloride in such reactions has not been fully resolved (Herold 1979). Once intercalation has been completed, the reaction chamber and its contents are quenched. To prevent exfoliation during this cooling process, the reaction chamber is rst quenched on the intercalant side, away from the graphite, so that the vapour pressure is suitably reduced and condensation of metal chloride vapour on the sample is avoided (Hooley 1973). Once cool, the ampoule containing the intercalation compound is sealed o . Reducing Tg Ti in the presence of a high intercalate vapour pressure can result in very rapid intercalate uptake, the formation of mixed stages, and the introduction of large strains which are usually relieved by rapid (explosive) exfoliation of the sample. Exfoliation can also cause problems in handling samples. In some cases, exposure of intercalated graphite to vacuum causes exfoliation, which is prevented by use of an inert gas atmosphere in the encapsulating ampoule. For example, graphiteAsF5 samples are encapsulated in dry nitrogen gas, rather than in a vacuum (Falardeau et al. 1978). For the case of halogen growth, typical growth temperatures are close to room temperature, owing to the high threshold vapour pressure of the halogens. The only diatomic homopolar halogen molecule that intercalates readily into graphite is Br2 , though intercalation with ICl and IBr can also be carried out. For Br2 intercalation, the lowest stage that has been reported is a stage 2 compound. GraphiteBr2 compounds are readily prepared using the growth conditions Tg 208C and 308C < Ti < 208C, by inserting liquid bromine into a temperature-controlle d (refrigerated) alcohol bath (Sasa et al. 1971). To illustrate the staging conditions (regions of stability) for stage 2, 3, 4 and 5 compounds, the adsorption isotherm for graphiteBr2 is shown in gure 6. Also evident in this gure is the threshold vapour

10

M. S. Dresselhaus and G. Dresselhaus

Figure 6. Isotherms of bromine uptake by highly oriented pyrolytic graphite (HOPG) at 208C as measured by Sasa et al. (1971) for the bromination cycle (open circles) and the debromination cycle (closed circles). Below the threshold pressure pt , no intercalation occurs. Note also that when p 0 on the debromination cycle, a nonvanishing bromine uptake remains, forming a `residue compound.

pressure for intercalation, pt 0:1p0 , where p0 is the pressure where the saturation compound (second stage C16 Br2 ) is formed. The open circles give the experimental points of Sasa et al. (1971) for bromination (adsorption) and the solid circles for debromination (desorption) . Because of the instability of these compounds (seen as hysteresis in the adsorptiondesorption curves), they must be encapsulate d in ampoules for storage and properties measurements. We note from this gure that in the absence of an ambient Br2 vapour pressure (p=p0 0), about 30% of the Br2 uptake at saturation remains in the resulting residue compound. The two-zone vapour transport method is widely applied to the growth of wellstaged intercalation compounds and for many systems is the preferred growth method. However, for specic systems, other intercalation methods are advantageous as described below. 2.1.3. Other intercalation methods Isothermal vapour transport is another method used for the preparation of acceptor compounds. This method is for example applied to the preparation of the high conductivity graphiteAsF5 compounds (Falardeau et al. 1978). For this system, the growth takes place at room temperature using an AsF5 overpressure of 3 atm. The growth time is the principal parameter used to control the stage index, and well-staged compounds have been prepared for 1 n 5. During the growth process, Falardeau et al. monitored the stage index visually by measurement of the increase in c-axis sample thickness, which exhibits distinct steps for each of these low stage compounds that is produced (gure 7). Another modication of the basic vapour transport technique to control the stage index is the use of restricted amounts of intercalant in the reaction chamber. For example, a number of well-staged compounds have been prepared using a twozone vapour transport system but with restricted quantities of crystalline AlCl3 , the smaller the quantity, the higher the stage that results (Gualberto et al. 1980).

Intercalation compounds of graphite

11

Figure 7. Plot of c-axis expansion ratio, t=t0 , versus reaction time for a typical intercalation of AsF5 into graphite (from Falardeau et al. 1978). Note the stability regions of t=t0 corresponding to each of the stages 1 n 5:

Post-intercalatio n methods have also been used successfully to change the intercalate concentration or oxidation state. For example, using the wash method, the preparation of dilute compounds from a more concentrated material can be carried out by washing with a solvent, as for example using acetone to wash graphiteFeCl3 (Hooley and Soniassy 1970). While this method can provide very dilute compounds, the resulting samples are not well staged. Chemical reactions with the intercalate species to change the oxidation state is in some cases also possible after intercalation is completed. For example, FeCl3 is readily intercalated into graphite by the two-zone method, but FeCl2 has not been successfully intercalated directly. However, FeCl3 can be reduced to FeCl2 in the intercalation compound using H2 at 3758C (Hooley et al. 1968). In this case, the oxidation state of the intercalant was identied using Mossbauer spectroscopy, which shows distinctly di erent spectra for FeCl3 and FeCl2 . It is also of interest that the reduction to FeCl2 could be carried out essentially completely, so that no characteristic Mo ssbauer spectrum for FeCl3 was observable. It has also been claimed that complete reduction of a metal chloride can be carried out to obtain two-dimensional metal sheets between graphite layers (Klotz and Schneider 1962, Novikov et al. 1971, Volpin et al. 1975). Though a number of workers have found evidence for the presence of -Fe upon reduction of graphite FeCl3 , using X-ray, Mossbauer and other techniques, it has not been established that staged intercalate metal layers can be prepared by reduction methods (Bewer et al. 1977). Studies of the di erences of H2 and CO adsorption isotherms on unactivated and activated (by heat treatment in H2 ) stage 1 graphiteFeCl3 have been interpreted to indicate that Fe metal is formed on the sample surface, but is not intercalated (Parkash et al. 1978). Another interesting chemical reaction that has been carried out within an intercalation compound is the reduction of various graphite metal chlorides with alkali metals to produce a catalyst for NH3 synthesis (Ichikawa et al. 1972a). The electron di raction patterns taken by Evans and Thomas (1975) of the graphiteFeCl3 K system show evidence for the presence of KCl and free iron, in agreement with Mossbauer studies (Tricker et al. 1974), also showing free iron. It has been proposed (Evans and Thomas 1975) that the catalytic activity of this system is due to the presence of dispersed free iron on the graphite surface.

12

M. S. Dresselhaus and G. Dresselhaus

Somewhat related to the use of chemical rections to change the oxidation state of the intercalant is the state of the intercalant before and after intercalation. For example in the case of graphiteHNO3 compounds, Forsman et al. (1978) have concluded that the intercalant is present in the form of neutral HNO 3 molecules admixed with charged NO ions by identifying NO 2 as a by-produc t of the 3 intercalation process. It is believed that similar phenomenon may occur in other acceptor compounds such as with the intercalants AsF5 and SbCl5 . Another useful technique for the preparation of certain intercalation compounds is the use of liquid intercalants. For the case of the alkali metal donor lithium, rst stage C6 Li samples can be prepared by degassing the graphite and then immersing it in molten lithium in a stainless steel crucible in an argon atmosphere containing less than 1 ppm H2 O and O2 (Zanini et al. 1978a, Basu et al. 1979). Using a solution of molten lithium and sodium (3.8 wt% Li) stage 2 samples can be produced, with only Li being intercalated. After intercalation, the external sample surfaces containing a metal lm are removed by cleavage. Graphitelithium compounds with stages 1 n 4 have also been prepared by compressing lithium powder (1020 kbar) with crushed natural graphite in an argon atmosphere (Gue rard and Herold 1975). Somewhat related to the method of intercalation from the melt is intercalation from solution. This method can for example be applied to the preparation of graphitebromine compounds using CCl4 as a solvent (Hennig 1952a, Hennig and McClelland 1955, Saunders et al. 1963). In this case the intercalate concentration in the compound is controlled by the intercalate concentration in solution, its temperature and the immersion time. Though intercalation from solution provides a convenient intercalation method, it is di cult to prepare well-staged compounds using this technique. Some other examples of intercalant/solvent mixtures are FeCl3 in acetone and FeCl3 in nitro-methane (Hooley 1972), and in nitroethane (Ginderow and Setton 1963), and other metal chlorides in SOCl2 , SO2 Cl2 and CCl4 (Stumpp 1977). Gas solutions have also been used to intercalate species that do not alone intercalate readily. For example, metal chloride vapours such as AlCl3 and FeCl3 have been used successfully to intercalate other metal chloride species such as CoCl2 and NiCl2 into graphite (Stumpp 1977). In the case of donor intercalants, liquid ammonia has been used by Rudor et al. (1955) and other workers as a solvent for the intercalation of metals M such as Li, Na, K, Rb, Cs, Co, Sr, Ba, though in this solution growth process a ternary compound results, as for example stage 1 compounds with stoichiometry C12 M(NH 3 )2 . Other workers (Stein et al. 1966, Ginderow and Setton 1970, Beguin and Setton 1975) have used organic solvents for intercalating metals, and some dilute donor compounds have been prepared with this technique. The preparation of ternary intercalation compounds by either cointercalation or sequential intercalation has also been pursued. Some examples of interest are cointercalation of alloys of the alkali metals K, Cs, Rb with Na and of mixed halogen compounds of Br2 with I2 and Cl2 . Preparation of such alloys and mixtures provides a method for the insertion of materials that do not readily intercalate by themselves. For example, Na has been intercalated into graphite as a binary compound only at very high stages (Asher 1959), but Na can be readily intercalated as an alloy with Cs and K (Billaud and Herold 1974, Billaud et al. 1980). Likewise, chlorine does not intercalate as a binary compound but can be intercalated as a mixture with bromine (Furdin et al. 1970, Furdin and Herold 1972) and with iodine (Bach and Herold 1963, 1968). One explanation that has been proposed for the

Intercalation compounds of graphite

13

failure of Na and Cl2 to intercalate in low stage compounds is the lattice mismatch of layers of graphite and of solid Na (Dresselhaus and Dresselhaus 1979), and of solid Cl2 (Hooley 1973). In order to prepare a compound by this type of cointercalation process it is necessary for one of the elements to intercalate by itself and for the two intercalants to be miscible (He rold 1979). In these cases, cointercalation improves the lattice match, thereby promoting the intercalation process. Another example of cointercalation is the introduction of intercalants with di erent chemical properties. For example, graphiteAlBr3 Br2 can be prepared as a stage 1 compound with stoichiometry C9 AlBr3 Br2 , containing a higher bromine density than the saturated stage 2 compound C16 Br2 (Sasa et al. 1972). In general, the preparation of metal bromide compounds is similar to that of metal chloride compounds in that the presence of Br2 gas is usually required for intercalation and the preparation of well-staged samples (Stumpp 1977). However, for the case of the metal bromides, the presence of Br2 gas results in the growth of ternary compounds because of Br2 uptake. Whereas the alloy compounds mentioned above are intercalated at the same time, sequential intercalation is also commonly used to insert materials that are not readily intercalated as binary compounds. For example, hydrogen which does not intercalate by itself, can be inserted into an alkali metal compound. The introduction of hydrogen into a rst stage C8 K compound forms an ordered second stage compound C16 K2 H4=3 in which the intercalant resides in a triple layer sandwich formed by two highly electropositive K layers between which is inserted a less electropositive H layer (Gue rard et al. 1977b, Lagrange and Herold 1978). A model proposed by Lagrange and He rold (1978) for the intercalation process that takes 2C8 K into C16 K2 H4=3 is shown in gure 8(a), and the corresponding layer arrangement is shown in gure 8(b). The intercalate sandwich, including a hydrogen layer between two potassium layers, has a thickness of 5.18 A. Subsequent intercalation of C16 K2 H4=3 with an alkali metal M K, Rb or Cs yields another stage 2 compound with stoichiometry C8 K 2 H4=3 , C8 M and a unit cell of Ic 8:53 A IM as shown in gure 8(b). The intercalation of higher stage alkali metal compounds can also be carried out, and use of such compounds for H2 storage and as molecular sieves for H2 has been discussed (Lagrange and Herold 1978). Ternary compounds with HgK and HgRb have been similarly prepared by Lagrange et al. (1980). Large organic molecules can also be intercalated using the sequential intercalation technique starting with a graphiteK binary compound. For example, benzene can be intercalated into stage 2 or stage 3 graphitepotassium, but not into rst stage C8 K (Merle et al. 1977, Bonnetain et al. 1977). The use of cointercalation and sequential intercalation techniques greatly expands the number of possible intercalation compounds that can be prepared. These techniques furthermore form a basis for the use of graphite intercalation compounds as catalysts for carrying out chemical reactions. Such applications have been made to the preparation of a number of organic compounds (Setton 1977) and to the use of graphite layers to promote polymerization (Gole 1977). Electrochemical techniques are also useful for the preparation of graphite intercalation compounds, particularly for strong acid intercalants such as sulphuric, perchloric, nitric and triuoroacetic acids. The graphite is oxidized anodically by placing the graphite specimen in a platinum cap suspended in concentrated acid, and using a second platinum counterelectrode. In this case, the stage formation is controlled by the electrode voltage (Rudor and Siecke 1958, Bottomley et al.

14

M. S. Dresselhaus and G. Dresselhaus

Figure 8. Model for ternary graphitealkali metal compounds containing hydrogen as proposed by Lagrange et al. (1978). (a) Schematic model for the transformation of rst stage 2C 8 K to second stage C16 K2 H4=3 . (b) Layer arrangement for pristine graphite, rst stage C8 K, second stage C16 K2 H4=3 and second stage C8 K2 H4=3 , C8 M where the thickness of the alkali metal sandwich is IM 1:97, 2.27, 2.58 A depending on whether M K, Rb or Cs.

1963b, Horn and Boehm 1977, McRae et al. 1980a). The electrochemical preparation method provides an excellent technique for study of the thermodynamic changes accompanying a change in stage during the growth process. This technique has been applied to both the system of donor compounds graphiteK (Aronson et al. 1968) and the acceptor compounds based on the intercalant H2 SO4 (Aronson et al. 1971), as discussed in section 2.3. Because of the colour change associated with di erent stages of certain graphite intercalation compounds (for example, alkali metal compounds), it is possible to use electrochemical means to produce a change in stage (and colour), thereby yielding an electrochromic e ect. Such an electrochromic e ect has been proposed for display device applications (Puger et al. 1979). 2.1.4. Some prototype intercalants For various physical measurements, specic intercalation compounds are of particular interest from the point of view of chemical stability, the simplicity of the intercalate structure, and specic properties such as high electrical conductivity or anisotropy, magnetic ordering and superconductivity . We discuss below some prototype intercalation compounds that are particularly suitable for specic applications. Because of their stability in air (Lazo and Hooley 1956), graphiteFeCl3 samples may be safely removed from their encapsulating ampoules for short periods of time, and therefore provide prototype materials for properties measurements and exploratory investigations. Curiously, pristine FeCl3 is itself highly reactive and hygroscopic, in contrast with the behaviour of its intercalation compounds. It has, however, not yet been demonstrated that one can prepare a stage 1 graphiteFeCl3

Intercalation compounds of graphite

15

sample without inclusions of pristine graphite, as was long ago noted by Cowley and Ibers (1956). On the other hand, stage 1 graphiteAlCl3 is readily prepared, although it is found to be very unstable when removed from its growth ampoule and exposed to air (Rudor and Zeller 1955, Dzurus and Hennig 1957a). Because of the chemical similarity between graphiteFeCl3 and graphiteAlCl3 , stage 1 samples of graphite AlCl3 are sometimes used in conjunction with higher stage graphiteFeCl3 compounds for properties measurements. Although stage 1 graphiteAlCl3 is highly reactive and requires encapsulation, high stage (n 4) graphiteAlCl3 samples are of comparable stability to graphiteFeCl3 samples of similar (high) stage. Environmental stability has also been reported for graphiteSbCl5 compounds for n 2 by Murthy et al. (1980). Alkali metal compounds are used frequently as prototype materials for many property measurements. Of all intercalation compounds these materials are most easily prepared, exhibit the highest degree of order and are best understood from a structural point of view. Compounds formed with the heavy alkali metals K, Rb and Cs form one class of compounds with many similar properties, which can be compared to compounds formed with the alkali metal donor Li. The graphiteLi compounds di er from the heavy alkali metal compounds with regard to structural ordering and intercalate ionic size, and therefore exhibit somewhat di erent properties. Other donor intercalants exhibiting the same in-plane ordering as Li are the alkaline earth metals Ba and Sr (Guerard and He rold 1974) and the lanthanides Eu, Yb, Sm and Tm (Lagrange et al. 1980). Among the acceptors, the simplest compounds are the graphiteBr2 compounds, where the intercalant is a simple homopolar diatomic molecule having its molecular axis aligned within the intercalate layer. Another simplifying feature of the graphite Br2 system is the formation of commensurate compounds, in which the ordering of the bromine intercalate layer is in registry with that in the graphite layers (Eeles and Turnbull 1965). The graphiteBr2 system however does not form stage 1 compounds (Sasa et al. 1971). On the other hand, the halogen ICl does form stage 1 compounds, and has the further advantage of having a dipole moment unlike Br2 , thereby providing increased coupling of the intercalant to an applied electromagnetic eld probe. The failure to prepare stage 1 compounds is common for many acceptor compounds, as for example with certain metal halides (Stumpp 1977), and the reason for this di culty is not generally understood. There is also much evidence in the literature for di culties in the preparation of certain low stage compounds with specic intercalants, as for example PdCl2 which formed only stage 3 compounds (Novikov et al. 1973). Intercalation in some cases permits study of a chemical species that is otherwise unstable. For example, TlBr2 does not exist by itself but must be kept in liquid Br2 . However, stable graphite intercalation compounds can be formed with TlBr3 (Niess and Stumpp 1978). Low stage acceptor compounds prepared with the highly reactive Lewis acid AsF 5 have been prototype materials for conductivity studies because of reports of a room temperature in-plane conductivity exceeding that of copper and an anisotropy between in-plane and c-axis conductivity of 106 (Foley et al. 1977, Falardeau et al. 1977). Signicant advances in materials preparation of graphiteAsF5 compounds (Falardeau et al. 1978) have also contributed to the widespread interest in these compounds.

16

M. S. Dresselhaus and G. Dresselhaus

The graphiteFeCl3 compounds are prototype magnetic materials with antiferromagnetic ordering reported below 3.6 K for stage 1 and ferromagnetic ordering below 8.5 K for stage 2 (Karimov et al. 1971, Ohhashi and Tsujikawa 1974a) while the stage 1 alkali metal compound C8 K is an intercalation compound in which superconductivity can be studied below 1 K (Hannay et al. 1965, Koike et al. 1978). As graphite intercalation compounds nd new application areas, it is likely that other intercalation compounds will become prototypes for these applications. 2.2. Materials characterization A number of techniques are exploited for sample characterization of graphite intercalation compounds, including visual inspection, weight uptake, chemical analysis, c-axis dilatation, di raction measurements and electron microscopy. Di raction studies yield the stage index and information on staging delity and on in-plane order, and electron microscopy provides information on the microstructure and submicrostructure. In this section we review some of the most important sample characterization techniques. Several simple characterization methods are used to provide useful qualitative information. The sample colour observed by visual inspection gives qualitative information on the stage: for example, for the alkali metal compounds a yellow, gold or red colour is characteristic of stage 1 compounds, steel blue for stage 2, dark blue for stage 3 and graphitemetallic for higher stages (Hennig 1959, Herold 1979). For acceptor compounds, stage 1 is often blue and higher stage compounds graphitemetallic. Direct chemical analysis gives the chemical formula for the intercalation compound. Weight uptake gives information on the sample stage if the chemical formula for the compound is known and stoichiometry is assumed. However, this information is qualitative because of sample inhomogeneity, the presence of intercalate vacancies and the preferential accumulation of intercalant in the vicinity of crystal defects. Because of sample expansion along the c-axis due to the intercalation process, dilatation along the c-axis, observed using a travelling microscope, also provides qualitative information on the sample stage. This method is subject to large systematic errors because of sample inhomogeneities, the formation of microcracks and the tendency for exfoliation especially near the sample edges where the length measurements are usually made. However, for the case of graphiteAsF5 , c-axis sample expansion provides a good indicator of stage, as shown in gure 7. For this system, c-axis expansion is used as a diagnostic for stopping the growth process at a given stage (Falardeau et al. 1978). Since many properties measurements are strongly dependent on the stage index, it is important to characterize samples to be used for property measurements with regard to stage index and stage delity. This type of characterization of graphite intercalation compounds is provided by X-ray di raction using (00l) reections. Sample characterization for stage index n and repeat distance Ic has been carried out for large numbers of compounds (more than 50) and the results are given in various review articles (Hennig 1959, Ebert 1976, Stumpp 1977) and in table 6. It has become common for property measurement studies on graphite intercalation compounds to include staging information as obtained by (00l) di ractograms. Accurate staging determinations can be made with a system such as is shown in gure 9 (Leung et al. 1981a). Mo K radiation is used in order to minimize the X-ray absorption by the glass encapsulating the samples. A typical set of (00l) di ractrograms is shown in

Intercalation compounds of graphite

17

Figure 9. X-ray system for (00l) di ractometer scans. K radiation from a Mo X-ray source is incident on the sample and the di racted beam is detected by a cooled Li-drifted silicon detector. This detector permits high resolution energy discrimination of the di racted beam. The energy windows of the single channel analyser are set so that only signals corresponding to K1 and K2 radiation are processed. The multichannel analyser is used for data acquisition of (00l) di ractograms (from Leung et al. 1981a).

gure 10 for graphite and graphiteRb samples of stages 1, 3 and 6. From these di ractograms, the di raction angles l corresponding to the various (00l) reections are determined, and using Braggs law l 2Ic sin l ; 2:1

Ic , the repeat distance (see gure 2), is accurately evaluated. Since the graphite interlayer separation is essentially una ected by intercalation, the stage index n is found from the relation Ic nc 0 di n 1c0 ds ; 2:2

where c0 is the distance between adjacent graphite layers (c0 3:35 A) and ds c0 di is the distance separating two graphite layers between which an intercalate layer is sandwiched. Analysis of (00l) di ractograms show that for a given intercalant, ds and c0 are essentially independent of stage (Hennig 1959, Rudor 1959), as illustrated in table 2. The validity of (2.2) provides strong evidence that the n graphite layers remain essentially unperturbed by the intercalation process. From (2.2) it follows that the increase in length along the c-direction t relative to the length prior to intercalation t0 is given by

18

M. S. Dresselhaus and G. Dresselhaus

Figure 10. X-ray stage characterization using (00l) di ractograms for stages 1, 3 and 6 graphiteRb compounds and for pristine graphite (from Leung et al. 1980a). Note the correlation in 2 of the high intensity reections for the high stage intercalation compounds with the occurrence of (00l) reections in graphite.

Table 2. Identity period or repeat distance Ic for graphitepotassium compounds. Stage Ic (measured) A Rudor a Parryb (measured) A (measured) A Underhillc Ic (calculated) 5:41 3:35n 1 A n 1 5.41 5.35 5.32 5.41 2 8.77 8.72 8.74 8.76 3 12.12 12.10 12.07 12.11 4 15.49 15.45 15.44 15.46

References: a Rudor and Schulze (1954); b Nixon and Parry (1968); c Private communication.

di t : nc 0 t0

2:3

Since the separation of neighbouring (00l) di raction peaks becomes smaller as the stage index increases, the energy discrimination technique provides an invaluable tool for stage determination of dilute compounds (for example, stage n > 6). In the analysis of staging for dilute compounds, the greatest sensitivity for determining the stage index is achieved with low index reections (small 2 l ). However, the intensities of the (00l) reections for small l due to the superlattice structure in high stage compounds are usually very low because of the mismatch of these 2 l values with the maximum (at 2 128) in the envelope function (see gure 10) due to the strong

Intercalation compounds of graphite

19

graphite (002) reection. However, for high stage compounds (n 6), high intensities are again obtained for 2 < 18, through matching with the envelope provided by the strong graphite (000) peak. Stage determination for very dilute compounds is made possible by the narrow linewidths of the di racted patterns and the ability to resolve di raction peaks within 0:58 of the direct beam (Underhill et al. 1980). For example, the full widths at half maximum (FWHM) intensity for the alkali metal intercalation compounds are roughly equivalent to those of pristine graphite; for 2 128 (the position of the (002) graphite peak which is the most intense peak), typical FWHM values for graphite are 0.28. Using the two-zone vapour transport growth technique described in the previous section, it is possible to prepare single-staged alkali metal compounds as dilute as n 8 (Underhill et al. 1980), and using a di ractometer system such as in gure 9 to characterize them for stage index and stage delity. Stage delity is established by the absence of satellite di raction peaks due to small quantities of admixed stages, and by accurate distinction between stages n and n 1 in the analysis of the di ractometer scans. The absence of broadening of the di raction lines at large di raction angles provides strong evidence that the samples are single staged, and not an average over a distribution of stages (Metz and Hohlwein 1975b). The X-ray pattern obtained from most intercalation compounds during a staging phase transition is of the type discussed by Metz and Hohlwein and can be explained in terms of the domain model for staging proposed by Daumas and He rold (1969) and discussed in section 2.4. In discussing stage indelities it is useful to distinguish two types of stage indelity. A mixed stage sample contains macroscopic regions that exhibit stage n and other macroscopic regions with di erent stages, usually stages n 1 or n 1. In this case the X-ray (00l) di ractograms show reections that are broadened relative to those from a single-staged sample, some reections that can be identied with each constituent stage, and others that are shifted because of unresolved 2 l components. In contrast, a randomly staged material will contain a random arrangement of regions with various stages. If one of these stages is dominant, a well-dened di raction pattern will result but with reections showing extensive line broadening (Metz and Hohlwein 1975b). The random stacking of intercalate layers has been studied in detail by Metz and Hohlwein (1975a , b) for the graphiteFeCl3 system, but no systematic study of this phenomenon has been carried out for any other intercalant. Further sample characterization is provided by analysis of the intensity of the ^ (00l) reections, by relating the square root of the observed intensities I00l to the magnitude of the structure factor F00l X sin n"l l 2:4 fX exp 2i"X l; 1 jF00l j fC sin "l
X

where fC and fX are atomic scattering factors for carbon and for the intercalant on layer X, " c0 =Ic , and "X zX =Ic , where zX is the distance of an intercalate layer relative to the centre of the intercalate sandwich, and 1= is the intercalate density. Since the structure factor determines the relative intensity of the (00l) reections, the ^ ^ index l of the reection with maximum peak intensity (00l ) approximately determines the stage n according to the relation ^ l n mi ; 2:5

20

M. S. Dresselhaus and G. Dresselhaus

Figure 11. Plot of relative X-ray integrated intensity versus x for stage 2 graphiteAlCl3 , where x is the dimensionless quantity x c0 l=Ic (see text). The solid and dashed curves represent Cl3 Al2 Cl3 and AlCl6 Al intercalate stacking respectively. The t to the experimental measurements (closed circles) of Gualberto et al. (1980) is much better for the Cl3 Al2 Cl3 intercalate stacking arrangement than for the AlCl6 Al stacking (Leung et al. 1980b).

where mi is the nearest integer to the ratio di =c0 (Leung et al. 1980b). For multilayer intercalants (for example, FeCl3 and AlCl3 ), structure factor calculations can provide valuable information for determining the intercalant stacking arrangement ^ when compared to experimental I00l measurements. An example of such a determination is the identication of the Cl3 Al2 Cl3 stacking arrangement of the intercalate layer sandwich, as shown in gure 11 (Leung et al. 1980b). Integrated intensity measurements of (00l) reections also provides an important characterization technique through determination of the intercalate in-plane density 1=. While gravimetric measurements depend on the total weight uptake during intercalation, the structure factor depends only on the fraction of intercalate arranged in ordered stages. For compounds where the in-plane structure is known, the in-plane intercalate density can be used to obtain the fractional site occupation in the intercalate layer (Leung et al. 1979). Such an analysis applied to a rst stage graphiteRb sample, which forms a p2 2R08 superlattice (Rudor and Schulze 1954, Nixon and Parry 1968, Kambe et al. 1980a) and assuming the stoichiometry, C8 Rb, shows 15% vacancies at room temperature (Leung et al. 1979). Real space electron micrographs of thin graphiteRb specimens show the presence of more than one phase at 300 K for compounds with stages n 2 (Kambe et al. 1980a). These experiments suggest that even in single-stage d materials with long-range c-axis ordering, defects and multiple phases within the intercalate layer are common for n 2 compounds. As discussed in section 3, di raction measurements also provide the principal technique for structural studies of the in-plane ordering in graphite intercalation compounds. The electron microscope has also provided useful information on the microstructure and ultramicrostructur e of graphite intercalation compounds. An especially vivid application of this technique was made by Evans and Thomas (1975) and by Thomas et al. (1976), who applied high resolution transmission electron microscopy (T.E.M.) in the lattice imaging mode to `see individual graphite and intercalant lattice planes directly. Such a bright eld lattice image of a graphite FeCl3 samples is shown in gure 12(a) and the schematic diagram illustrating the interpretation given by Thomas et al. (1976) of this photograph is presented in gure 12(b). This work demonstrate s the staging phenomenon directly with real image

Intercalation compounds of graphite

21

(a)

(b) Figure 12. High resolution transmission electron microscopy applied to lattice imaging of individual graphite and intercalate lattice planes in graphiteFeCl3 by Thomas et al. (1976). (a) Bright-eld lattice image of a graphiteFeCl3 sample (average stage n 2). The light striations on the micrograph delineate the sheets of FeCl3 intercalant, and from their separation the stage indices may be identied. (b) Schematic diagram illustrating the interpretation of (a) in terms of the various stages that are contained in the photograph.

22

M. S. Dresselhaus and G. Dresselhaus

photographs and gives evidence for regions of material with large repeat distances (sixth stage). These authors were also able to relate the observed layer thicknesses in the intercalation compound with those in the parent materials. Furthermore, the sample region shown in gures 12(a) and (b) indicates a predominance of stage 2 with some admixtures of stage 1 and trace amounts of stages 5 and 6. In view of the di culty in preparing a-face samples (see section 4.3), the staging delity shown in gure 12(a) is impressive. Similar latttice image results have also been presented for pristine graphite and for graphiteK compounds (Evans and Thomas 1975). The T.E.M. technique is di cult to apply in practice because very thin samples (600 A) are required to permit penetration by the electron beam and because the intercalate tends to desorb after several hours under the high vacuum conditions of the electron microscope column. Nevertheless, the use of high resolution transmission electron microscopy under dynamic conditions could provide extremely valuable information on the intercalation mechanism. Very high resolution studies using scanning transmission electron microscopy (S.T.E.M.) have been carried out on the heavy metal atom migration on thin amorphou s carbon lms, allowing the motion of single atoms to be studied (Isaacson et al. 1979). Such studies could also provide very interesting information on intercalated graphite. Also of interest is the S.T.E.M. study of commercially produced `graphimet intercalates by Fischer et al. (1979) showing very thin metal clusters of about 2060 A in diameter. Transmission electron microscopy has also been applied very successfully by Heerschap et al. (1964) and Heerschap and Delavignette (1967) to study the microstructure associated with in-plane imperfections (stacking faults and dislocations) in graphiteBr2 , graphiteICl and graphiteFeCl2 compounds. Their real image micrographs provide evidence for the existence of isolated dislocations bounding the edge of an intercalated layer (see gure 13) in the case of all three intercalant systems. The usual hexagonal layer stacking of pristine graphite is ABAB as shown in gure 14, while rhombohedral graphite exhibits the ABCABC stacking shown in gure 15. The Burgers vectors for the dislocations in the graphiteBr2 and graphiteICl systems are equivalent to the Burgers vector of the normal partial dislocations in hexagonal pristine graphite. This result is interpreted in terms of an interlayer shift of the graphite bounding layers to achieve identical crystallographic positions, shown in gure 14 as an `A over A stacking arrangement of the graphite layers about the intercalant (AjA), in agreement with X-ray di raction results for

Figure 13. Schematic cross-section of a graphite crystal containing a dislocation bounding an intercalated layer of a reactant. Intercalation often causes a shift of the adjacent carbon layers to yield a fully symmetrical arrangement relative to the intercalant layer X. For clarication of the A, B, C interlayer arrangement of graphite layers see gures 14 and 15.

Intercalation compounds of graphite

23

Figure 14. Structure of hexagonal graphite showing ABAB stacking (from Wycko

1964).

Figure 15. Structure of rhombohedral graphite showing ABCABC stacking (from Wycko 1964).

these materials (Rudor 1941). On the other hand, for the graphiteFeCl3 compounds the Burgers vector is found to be perpendicular to the graphite planes so that the graphite preserves its ABAB stacking upon intercalation with FeCl3 , in agreement with X-ray results of Rudor and Schulz (1940). It is reasonable to expect commensurate intercalate structures to exhibit a symmetrical AjA stacking, while incommensurat e structures have no symmetry reason for disturbing the normal graphite stacking. The pinning of dislocation loops of 1 mm diameter at crystal imperfections is also demonstrated in the micrographs obtained by Heerschap and Delavignette (1967). Based on these observations it is concluded that upon intercalate desorption to form residue compounds (Hennig 1959), intercalate islands can be trapped in

24

M. S. Dresselhaus and G. Dresselhaus

Figure 16. Model of a residue compound showing intercalate islands trapped between dislocation loops.

dislocation loops, as shown in gure 16. Evidence for these trapped intercalate islands is also provided by in-plane X-ray and electron di raction patterns from residue compounds (Maire and Mering 1959, Chung et al. 1977). The preferential trapping of the intercalant at crystal defects is observed in the T.E.M. micrographs of Heerschap et al. (1964) and has also been identied by X-ray dispersive analysis using scanning transmission electron microscopy (Chung 1977). A combination of electron di raction patterns and real image electron micrographs have shown that epitaxial layers form on the graphite surface under certain growth conditions. For example, when a graphiteBr2 compound is exposed to potassium vapour, Evans and Thomas (1975) have shown that KBr forms epitaxially on the surface, and they have interpreted their results in terms of an extraction of bromine from the intercalation compound. The formation of epitaxial layers of potassium and caesium on the surface of graphiteK and graphiteCs compounds when heated (to 2008C) has also been noted by Halpin and Jenkins (1970) and by Chung et al. (1977), and this surface growth has been related to intercalate desorption from the bulk. Scanning electron microscopy (S.E.M.) also provides information on the microstructures of the intercalation compounds. Such observations on an a-face (containing both an a-axis and a c-axis) give evidence for microcrack formation resulting from intercalation (Chung 1977). These S.E.M. obervations further show that microcracks perpendicular to the c-direction are common in the HOPG host material and in single crystal akes separated by acid treatment from the carbonate ore-bearing rocks, but such microcracks are almost absent in single crystal akes that are separated mechanically. 2.3. Kinetics, thermodynamic s and the intercalation process Important insight into the intercalation process has been provided by studies made while intercalation is in progress. Such time-dependent studies involve both kinetic and thermodynamic considerations as described below. An important set of experiments relevant to intercalation kinetics was carried out by Hooley et al. (1965) on the graphitebromine system using a cylindrical HOPG host sample. In the rst of these experiments, it was found that no intercalation occurred for bromine vapour pressures less than the threshold pressure for intercalation pt , where pt 170 torr at 208C for Br2 intercalation. As p was increased above pt , intercalation rst began into the layers near the terminal free surfaces. By coating the outer cylindrical surfaces of the graphite sample with impervious material but leaving the end planes uncoated, and then coating the end planes but leaving the cylindrical surfaces uncoated, Hooley et al. (1965) concluded that interaction of the intercalant with both the exposed basal plane surfaces and the

Intercalation compounds of graphite

25

exposed cylindrical surfaces is necessary for the initiation of the intercalation process. From these experiments Hooley et al. inferred that intercalation sets up a tension between the outer surface region and the inner core region. If these tension forces are small, the reaction front proceeds to the inner core and the samples become fully intercalated. If, however, these forces are large, then the intercalation process stops and the tension breaks up the cylinder into discs. The results of the Hooley experiments have been interpreted to indicate that a strain energy is introduced by separating two graphite layers through the insertion of an intercalant. The connection of a strain mechanism with staging is discussed in the following section. As shown in gure 6, intercalation does not begin unless the intercalant vapour pressure exceeds the threshold pressure pt . It is believed that a threshold pressure is required to unpin lattice dislocations and to relieve lattice strain since the intercalation process changes the atomic stacking sequence and requires the motion of dislocations (Ubbelohde 1968a, Dowell 1977). The intercalation threshold depends sensitively on intercalate species, on temperature and on the characteristics of the graphite host materials (Hooley et al. 1965). With regard to sample size, pt is lower for smaller sample thicknesses, so that thin samples intercalate more readily than thick samples. With regard to sample perfection, pt is lowest for single crystal akes, higher for HOPG host materials and yet higher for carbon bres (Hooley and Bartlett 1967, Hooley 1977a) . Measurement of intercalation isotherms for graphite bres by Hooley and Dietz (1978) with the intercalants Br2 and ICl show that the high initial value for pt is signicantly reduced (for example, by a factor of 2) after an intercalationdeintercalation cycle, and the intercalation rate is correspondingly increased. This behaviour is attributed to the formation of microscopic cracks in the interlocking amorphous carbon networks between the graphitized domains of the bres, and this explanation is consistent with the observed dependence of pt on the graphitization temperature of the bres. Values for pt relative to p0 , the pressure at maximum intercalate uptake (lowest stage that can be prepared), are shown in gure 6. Values for pt =p0 can be varied over more than three orders of magnitude with pt =p0 < 104 for materials that intercalate easily (K, Rb, Cs, AlCl3 ) and pt =p0 > 0:1 for materials that are di cult to intercalate (MoCl5 , WCl6 , HgCl2 ). Intercalants with high (pt =p0 ) values tend to have lower intercalate uptake (higher stages) at saturation. For example, the lowest stage that has been prepared with WCl6 having pt =p0 0:5 is stage n 5 (Hooley 1977a). Furthermore pt can in some cases be reduced by the presence of other chemical species, as for example in the case of graphiteFeCl3 , the presence of Cl2 gas signicantly lowers pt . A decrease in pt is generally achieved by raising the temperature of the graphite host material. The initiation of intercalation at the edges of the graphite crystal as discussed above is also supported by a number of other experiments, including studies of successive sorptiondesorption cycles in a bromine vapour atmosphere by Marchand et al. (1973) which were explained by a simple two-dimensional di usion process along the basal planes of the graphitic crystallites. The importance of the di usion mechanism was previously recognized by Aronson (1963), who studied the exchange of normal gaseous Br2 gas with radioactive bromine (Br82 with a half-life of 35.7 hours) that has previously been intercalated into natural graphite powders. The average concentration of radioactive bromine in the intercalation compound and in the exchange gas was determined from the decay. From measurement of the

26

M. S. Dresselhaus and G. Dresselhaus

bromine exchange rate at 303 and 321 K, Aronson concluded that self-di usion of the bromine in the graphite rather than an exchange between adsorbed and gaseous bromine at the graphite surface was dominant in the intercalation process. Because of the distribution of particle sizes and shapes, a fully quantitative study of the di usion process could not be carried out. Although intercalation is initiated at the edges of a graphite crystal, kinetic studies on the graphiteFeCl3 system by Metz and Siemsgluss (1978) using gravimetric (weight uptake) techniques have indicated that an approximately constant macroscopic distribution of intercalant is achieved in the intercalate layers when the total intercalate uptake is only 20 to 30% of its saturation value. This absence of a concentration gradient in the intercalated layers has been interpreted in terms of intercalate nucleation near the edge of the graphite crystal, rapid growth of the nucleus to an island having approximately the nal structure of the intercalate layer at saturation intercalate uptake. Thus Metz and Siemsgluss concluded that even at these low intercalant uptake values, the intercalate islands are uniformly distributed on a macroscopic scale between two sequential graphite layers. The slope of an Arrhenius plot of the intercalation rate constant is interpreted as an activation energy for nucleus formation at the edges of the graphite crystal (25 kcal/mol for the graphiteFeCl3 system) and this activation process limits di usion into the bulk graphite (Metz and Siemsgluss 1978). The activation energy for nucleus formation is much larger than the activation energy for di usion which was found to be 23 kcal/ mol by Barker and Croft (1953). Aronson (1963) also noted that more energy was required to initiate intercalation between two graphite host layers than to sustain subsequent di usion into the host. The main conclusions of the MetzSiemsgluss study on the graphiteFeCl3 system are in agreement with kinetic di usion studies by Dowell and Badorrek (1978) on the intercalation of HNO3 , Br2 and PdCl2 into an HOPG host material. In this study the di usion coe cient D was determined by measuring the weight gain Mt =M1 and its time derivative dMt =M1 =dt as a function of time, and using the equation for di usion into a cylinder Mt 4 Dt 1=2 M1 r2 1=2 Dt 1 Dt 3=2 2 1=2 2 ; r r 3 2:6

where r is the radius of the cylinder, and Mt and M1 are respectively the weight gain at time t and at very long times. These measurements, taken at constant vapour pressure on samples with di erent radii and as a function of temperature, show that the results for both Mt =M1 and dMt =M1 =dt can be explained by equation (2.6) in terms of a single di usion coe cient which is operative over the entire composition range. The absence of change in di usion coe cient as the sample passes from one stage to another, as for example from stage 4 to stage 3 on the way to the saturated stage 2 compound, implies that only minor crystal rearrangements occur upon stage transformation . Dowell and Badorrek showed that the magnitudes of D vary widely from one intercalant to another. For example, for Br2 at 303 K and p 258 torr, D 1:47 106 cm2 /min, while for HNO3 at the same temperature and p 81 torr, D 203 106 cm2 /min. The value for D obtained by Dowell and Badorrek (1978) for bromine di usion into an HOPG host material of specied cylindrical shape is in qualitative agreement with measurements by Aronson (1963) on intercalated natural

Intercalation compounds of graphite

27

graphite powders. The high di usion coe cient for HNO 3 has been attributed by Dowell and Badorrek (1978) to the liquid-like HNO3 intercalate phase at room temperature, whereas the Br2 intercalant at room temperature is arranged in an ordered phase. An activation energy of 5 kcal/mol for Br2 di usion in a HOPG host material has been measured by Mukaibo and Takahashi (1963), and 12 kcal/mol for Br2 di usion in natural graphite by Aronson (1963). The kinetics of the intercalation process depends intimately on thermodynamic considerations. A detailed determination of the thermodynamic parameters connected with changes in stage has been carried out for the case of the graphitealkali metal compounds with the intercalants K, Rb and Cs (Salzano and Aronson 1966a, b, c and Aronson et al. 1968). Relatively little attention has been given to the corresponding thermodynamic data for the acceptor compounds, though some work has been done on the graphiteH2 SO4 system (Aronson et al. 1971). Referring to the isobaric growth diagram for the graphiteK system presented in gure 4, we identify temperature (Tg Ti ) regions of phase stability where a single stage is dominant (for example, for n 1, 2, 3), and other temperature regions where two stages are in equilibrium. The transition from one stage to another occurs in these equilbrium regions and the thermodynamic studies to be described measure the changes in free energy, enthalpy and entropy associated with a stage transformation . Thermodynamic data can be obtained directly from calorimetric measurements, calculations based on the temperature dependence of equilibrium pressures, and electrochemical measurements. For the conditions used to prepare most intercalation compounds, energy is absorbed upon the stage transition n ! n 1 and the entropy decreases. While calorimetric measurements provide the most direct determination of thermodynamic data, this method is di cult to apply in practice to the study of stage transitions in intercalated graphite. This di culty arises from the sensitivity of the intercalation process to crystalline faults as demonstrated by the kinetic studies of Hooley and Dietz (1978) and Dowell (1977) on graphite host materials of varying perfection. On the other hand, direct calorimetric methods can be readily applied to the study of structural phase transitions or orderdisorder transformation s in graphite intercalation compounds as discussed in section 3.4. Using a solid state electrochemical cell method on the graphiteK system (Aronson et al. 1968), and a Knudsen e usion method in conjunction with a radioactive tracer technique for the graphiteRb and graphiteCs systems (Salzano and Aronson 1966a, b, c), values for the enthalpy and entropy changes associated with stage transformation s were obtained and the results are given in table 3. In the electrochemical method, Aronson et al. employed the e.m.f. cell shown in gure 17, which consisted of a liquid potassium Km anode and an intercalated graphite cathode. These electrodes were coupled by a solid potassiumglass electrolyte which provided transport of K ions from anode to cathode. The electrochemical reaction at the anode was therefore Km e K ; 2:7

while at the cathode a stage change n ! n 1 occurred. For example, for the transformation from stage 3 to stage 2, the reaction at the cathode was 2C 36 L K e 3C24 K to yield an overall reaction 2:8

28

M. S. Dresselhaus and G. Dresselhaus

Table 3. Thermodynamic properties of graphitepotassium, graphiterubidium and graphite caesium compounds.a


Graphite potassium H8 (cal/mol K) 27 400 38 000 24 000 27 800 30 000 30 600 S8 (cal/mol K 8K) 25.7 44 24 20.6 20.7 20.8 Graphite rubidium H8 S8 (cal/mol (cal/mol Rb) Rb 8K) Graphite caesium H8 (cal/mol Cs) S8 (cal/mol Cs 8K)

Equilibrium reaction 1a 1b 2 3 4 5 6 7 8
a

1=3C 24 Ms Ms C8 Ms 4C 10 Ms Mg 5C8 Ms 5=7C 24 Ms Mg 12=7C 10 Ms 2C 36 Ms Mg 3C24 Ms 3C 48 Ms Mg 4C36 Ms 4C 60 Ms Mg 5C48 Ms 5C 72 Ms Mg 6C60 Ms 6C 84 Ms Mg 7C72 Ms 7C 96 Ms Mg 8C84 Ms

33 900 25 300 27 200 29 500 31 100 31 800

42.2 19.2 17.0 17.6 18.3 18.1

43 800 29 600 32 700 34 200 34 900 35 800

43.4 19.6 18.7 18.6 18.6 18.8

31 700

20.9

From Aronson et al. (1968).

Figure 17. Apparatus used by Aronson et al. (1968) for e.m.f. measurements on cells of the type K (molten) jKglass j graphiteK. Based on the temperature dependence of these e.m.f. measurements the enthalpy and entropy changes associated with a stage transformation were determined.

2C36 K Km 3C24 K:

2:9

The free energy of the reaction Gcell is then obtained from the measured e.m.f. of the cell, E, Gcell mFE; 2:10 where F is Faradays constant (F 96 450 C) and m is the number of equivalents of potassium that are transferred. Using the thermodynamic relation between the Gibbs free energy, enthalpy and entropy, Gcell Hcell TScell ; 2:11

Intercalation compounds of graphite

29

Figure 18. Phase diagram by Aronson and Salzano (1966a) for several stage transformations in the graphitecaesium system.

the entropy term can be found from the temperature dependence of Gcell according to the thermodynamic relation Scell @Gcell =@T . Using this value of Scell , the enthalpy change is determined by Hcell Gcell TScell : 2:12

The resulting values for H cell and Scell are given in table 3 for several stage transitions that were studied in the graphiteK system. Using an e usion method, Salzano and Aronson (1966a, b, c) measured the equilibrium pressure versus temperature for a number of stage transitions in the graphiteRb and graphiteCs systems and obtained the results shown in gure 18. From the slopes of these curves, the enthalpy change was calculated and from the 1=T ! 0 intercept, the entropy change was determined. He rold (1979) has pointed out that the stoichiometry of the system changes during the stage transformatio n so that improved values for H8 and S8 are obtained by integration over the limits of the concentration range associated with the change in stage. He rold further noted that these corrections are especially important for the low stage transitions and may explain the anomalous H8 and S8 values in table 3 for the low stage transitions. However, the thermodynamic results obtained by Salzano and Aronson are believed to be qualitatively correct. The data in table 3 show S8 to be approximately independent of stage and H8 to decrease in magnitude with decreasing stage, at least for n 2. It should be noted that at high temperatures Salzano and Aronson observed an additional rst stage phase C10 M which is less dense than C8 M and is purple in colour. This observation illustrated that di erent inplane structures and in-plane intercalate densities are possible for a compound of a given stage.

30

M. S. Dresselhaus and G. Dresselhaus

Table 4. Heats and entropies of formation of graphitepotassium, graphiterubidium and graphitecaesium compounds.a Graphite potassium H f 8 (cal/mol K) 28 500 27 000 30 400 31 700 32 300 32 800 S f 8 (cal/mol K 8K) 24.0 22.5 20.7 20.7 20.7 20.8 Graphite rubidium H f 8 (cal/mol Rb) 28800 27 500 30 600 32 200 33 100 33 600 S f 8 (cal/mol Rb 8K) 23.3 18.5 17.6 18.0 18.1 18.1 Graphite caesium H f 8 S8 (cal/mol (cal/mol Cs) Cs 8K) 34200 32 200 34 800 35 800 36 400 36 700 24.0 19.2 18.7 18.7 18.8 18.8

Reaction 8Cs Mg ! C8 Ms 10Cs Mg ! C10 Ms 24Cs Mg ! C24 Ms 36Cs Mg ! C36 Ms 48Cs Mg ! C48 Ms 60Cs Mg ! C60 Ms
a

FromAronson et al. (1968).

Listed in table 4 are the overall enthalpies and entropies of formation for each of the compounds listed in table 3, based on the starting reactants pristine graphite and the intercalant in the gas phase. These entropies and enthalpies of formation are found by summing the corresponding entropies and enthalpies for each previous stage transition, taking proper account of the number of carbon and metal atoms for each transition (Aronson et al. 1968, He rold 1979). It is of interest to note that these values for the enthalpies of formation are in good agreement with the corresponding values calculated from (1) the measured heats of vaporization for the alkali metals, and (2) the measured enthalpies of formation of C8 K, C8 Rb and C8 Cs from the liquid metal using calorimetric techniques (Saehr 1964). Salzano and Aronson (1966a) also related these thermodynami c studies to a determination of the energy required to separate two graphite layers to innity against the van der Waals interplanar binding forces which hold the graphite layers together. This separation energy provides an estimate for the potential barrier that must be overcome when an intercalate layer is introduced into the graphite host. The value obtained by Salzano and Aronson (1966a) for this separation energy is 1.23 kcal/g mol carbon, which is in rough agreement with a number of other estimates of this quantity obtained on the basis of totally di erent techniques. By considering the stage dependence of the thermodynamic data in terms of an electrostatic model for the binding of the intercalate metal layer to the adjacent graphite layers, Salzano and Aronson (1966a) concluded that the amount of charge transfer to the graphite layers is one electron per intercalant atom ( f 1) for the K, Rb and Cs systems, at least for stage n 2, and close to f 1 for the rst stage compounds. This conclusion is consistent with other experiments relevant to the electronic structure (see section 4). On the other hand, analysis of the thermodyamic data for the acceptor intercalant H2 SO4 (Aronson et al. 1971) indicated that f 0:3, also in agreement with studies of the electronic properties of acceptors (Hennig 1959, Ubbelohde and Lewis 1960). Thermodynamic arguments have also been applied to other aspects of the intercalation process. For example, Dzurus et al. (1960) used thermodynamic arguments in an attempt to explain why Na does not intercalate into graphite to form low stage compounds by plotting the free energy of formation for the K, Rb

Intercalation compounds of graphite

31

and Cs intercalation compounds against the ionization potential of these metals and extrapolating to a positive free energy of formation for a rst stage Na compound. It would be of interest to explore whether there are thermodynamic constraints which inhibit the formation of certain low stage compounds, as, for example, rst stage graphiteBr2 . For the graphiteBr2 system, the lowest stage compound that has been prepared is stage 2. 2.4. The staging phenomenon A remarkable feature of graphite intercalation compounds is the occurrence of staging, in contrast to other intercalation compounds such as the transition metal dichalcogenides, which generally do not exhibit accurate staging. The existence of staging has been well documented by X-ray (00l) di ractograms as discussed in section 2.2, showing that the resulting superlattice unit cell extends over many atomic layers; for example, in stage 8 alkali metal compounds, Ic 30 A. Furthermore, the measured linewidth of the di ractograms indicates that a single stage sample can be established over macroscopic dimensions (1000 A). It should further be noted that although stacking faults in intercalated graphite are separated by as little as 30 A (see section 3.4), the staging phenomenon is long range. Staging is very general, occurring in both donor and acceptor compounds, regardless of whether the intercalate in-plane structure is commensurate or incommensurat e with the graphite host lattice. The existence and extent of staging does not seem sensitive to the amount of charge transfer between the graphite and intercalate layers, which is large for the alkali metal donor compounds and considerably smaller for the molecular acceptor compounds (Ubbelohde 1976). Furthermore, models for the c-axis charge distribution (Pietronero et al. 1978), the electrical conductivity (Bok 1978), and the interpretation of Raman and infrared spectra (Nemanich et al. 1977c, Dresselhaus et al. 1977b, Underhill et al. 1979), provide strong evidence that a single graphite bounding layer e ectively screens the intercalate from the graphite interior layers, thereby inhibiting long-range electrostatic interactions in the intercalation compounds. These observations on the e ective screening of the intercalate layer by the adjacent graphite bounding layers suggest that the staging phenomenon is related to a long-range lattice strain interaction rather than an electrostatic e ect. The expansion of the lattice constant in graphiteK compounds as a function of reciprocal stage (1=n) (Nixon and Parry 1969) and the (1=n) dependence of the lattice mode upshift in acceptors and downshifts in donors also strongly suggest a strain mechanism for staging (Underhill et al. 1979). Raman experiments, indicate that donor intercalants cause the in-plane graphite layers to expand as electrons are added, and acceptors to contract as electrons are removed. The Raman experiments also imply that the electron transfer causes a stage-dependent sti ening (acceptors) or softening (donors) of the carboncarbon bonds in the bounding layer plane. Whether the intercalate bonding is ionic, covalent or metallic, there is strong evidence that the long-range forces associated with staging are directly related to the strain energy of the solids. The kinetic studies of the intercalation process described in section 2.3 are also consistent with a strain model for intercalation. These kinetic studies show that once the exposed surface graphite planes have interacted with the intercalate species, the intercalant is introduced into the bulk host material as layers close to the exposed end surfaces, and sequentially into layers increasingly distant from the exposed end

32

M. S. Dresselhaus and G. Dresselhaus

surface. Initiation of intercalation at these extremal layers minimizes the elastic energy that must be supplied to produce the dilatation of the graphite host accompanying the intercalation process. The strain model can also account for the observation that thin samples intercalate more readily than thick samples because of the lower elastic impedance of thin samples. Molecular alignment studies in liquid crystals by de Gennes (1974) show that the elastic force eld yields a decrease in strain energy by clustering and aligning neighbouring rod-like molecules. Application of this concept to graphite intercalation compounds implies that the strain energy is minimized by the clustering of intercalate atoms to form platelets between a single set of graphite planes, in which intraplanar forces are extremely strong. Thus if a random distribution of intercalate atoms were to be introduced into graphite initially, the strain energy of the system would be lowered by the clustering of all intercalate atoms between a single set of graphite planes. However, each intercalate layer can accommodate no more than some maximum number dened by the close-packed condition. Thus the minimum strain energy is achieved by the closest intercalant packing within the intercalate layer, and by the largest separation between intercalate layers consistent with a given intercalate concentration. Although the introduction of an intercalate layer results in maximum displacement of the atoms in the graphite bounding layers, displacements must also occur in the graphite interior layers to maintain their commensurate interrelation with the graphite bounding layer. Thus the elastic interactions are long range and the resulting strain energy must be shared by all the graphite layers, or else the crystal would cleave. If one assumes that intercalate clustering and stage changes occur without intercalant hopping or di usion across graphite layers inside the crystal, or without migration around graphite layers at the crystal edge, one is led to the staging domain model or pleated layer model of gure 19 proposed by Daumas and Herold (1969),

Figure 19. (a) Single domain model for a stage 3 compound with every third layer occupied by the intercalant. (b) DaumasHerold domain model for a stage 3 compound with intercalant contained between all graphite planes. (c) Interchange of domains of fourth and third stage regions as might occur during a stage transformation.

Intercalation compounds of graphite

33

and recently observed directly by Thomas et al. (1980) in graphiteFeCl3 using lattice imaging with a transmission electron microscope. According to the pleated layer model the average number of intercalate atoms between any two graphite layers of the macroscopic crystal is the same. This domain model has been used by Metz and Hohlwein (1975b), and by Schoppen et al. (1977) to explain the observation of di erent intercalation isotherms for ordered and disordered sequences of intercalate layers in graphiteFeCl3 compounds. These authors have also used the model to explain the observations of hysteresis e ects in growth isotherms, depending on the number of intercalationdeintercalation cycles, the sample geometry and the reaction time. The pleated layer model has also been used to explain why X-ray di ractograms taken on a sample during intercalation show the sample to pass through a sequence of stages, since the domain structure allows a sample to undergo stage transitions (from n to n 1) without large spatial movement of the intercalate atoms. The DaumasHe rold domain model has not only been used to explain stage transitions n ! n 1 in binary intercalation compounds, but this model has also been applied by He rold (1979) to explain sequential intercalation of hydrogen into rst stage 2C8 K to form a second stage compound C16 K2 H4=3 as shown in gure 8. Dowell and Badorrek (1978) have argued that the absence of a change in di usion coe cient during a stage transformation provides strong support for the DaumasHe rold domain model, because this model does not require large intercalate migration to produce a stage transformation . McRae et al. (1980a) have noted that for stage 15 graphiteK, the electrical resistivity is three to four times lower than for graphite, indicating that intercalation corresponds to an ordered state, not merely to the random addition of intercalant. Clarke et al. (1979a) have interpreted hysteresis e ects in their high temperature (500 < T < 650 K) X-ray data for low stage graphiteCs compounds in terms of the DaumasHe rold model, identifying the high temperature phase with the melting of the Cs intercalate layer at 608 K. On the basis of their measured X-ray linewidths, they concluded that on freezing, the intercalate layer forms clusters of intercalate islands with a c-axis stacking coherence greater than 200 A. No direct measurement of the strain energy associated with a domain wall has been made, nor has the size of the domains been determined quantitatively . Small angle X-ray scattering measurements may provide a useful technique for the determination of the mean domain diameter in bulk samples (Berney and Underhill 1980). The connection between staging and lattice strain has recently received theoretical consideration by several authors. The relation between long-range ordering interactions and strain was considered by Kittel (1978) in relation to the long-range periodicity which is observed in certain magnetic systems. Safran and Hamann (1979, 1980) have discussed the application of these strain considerations as one of the driving forces for staging in graphite intercalation compounds according to the DaumasHerold domain model. These authors have calculated the strain-induced forces between two intercalate atoms in a layered graphite host material, showing that intercalate atoms between the same two graphite layers are attracted to one another to form two-dimensional islands, while intercalate atoms on di erent layers interact repulsively. By forming intercalate islands, the compound lowers its strain energy by 0.1 eV per intercalate atom, and the free energy is minimized by the formation of a pure stage conguration from a mixed or randomly staged crystal. The arrangemen t of intercalate islands in a mixed-stage compound is illustrated in

34

M. S. Dresselhaus and G. Dresselhaus

gure 19, and it is seen that the strain is localized in the boundary region between adjacent islands. (Refer to section 2.2 for the distinction between mixed-staged and randomly-stage d materials.) The domain model of Daumas and Herold furnishes a compellingly simple model for the growth characteristics of the K, Rb and Cs compounds, which show consecutive stage formation on growth (Nixon and Parry 1968, He rold 1979). A simple scenario (but not fully veried) for the intercalation process might be as follows. High temperature growth might start with the random intercalation of isolated atoms between all the graphite planes. As the temperature is lowered, the formation of two-dimensional islands or clusters between the graphite planes is favoured on the basis of elastic energy considerations (de Gennes 1974). For intercalation at lower temperatures, there could be a nucleation of islands near the sample edge, followed by a migration toward the centre of the sample. The domain walls at the boundaries of these islands have Burgers vectors to account for slipping of the graphite A plane relative to the B plane to achieve the AA stacking (see gure 13). Burgers vectors associated with dislocation loops have been identied in graphiteFeCl3 by Heershap et al. (1964, 1967) and in graphiteBr2 by Eeles and Turnbull (1965). Once again, elastic energy considerations favour the nesting of intercalate islands in order to yield equal and opposite Burgers vectors and thus to minimize the longrange strain associated with these islands. A consequence of this model is that the thickness of the intercalate layer di must be greater than the Burgers vector for the p stacking fault or di a0 = 3. Na and Li fail to satisfy this criteria (see table 5) and would not be expected to stage in an AjA conguration by the mechanism outlined here. This mechanism suggests a domain size which is inversely proportional to the stage index. The pairing of the Burgers vectors results in the stacking arrangement AjABjBCjCA for stage 2 compounds, in agreement with experimental observations on compounds with intercalants such as K, Rb, Cs, HNO 3 . It should be mentioned in this connection that rhombohedral graphite, having the stacking sequence ABCABC (see gure 15) is metastable and can be transformed into the commonly occurring hexagonal graphite (gure 14) by heating to high temperature or by

Table 5. Intercalate layer thickness di for various intercalants. Intercalant K Rb Cs Li HNO3 SbCl5 SbF5 Br2 AlCl3 FeCl3 NiCl2 AsF5 Thickness di (A) 2.05 2.35 2.66 0.35 4.44 6.01 5.06 3.69 6.13 6.10 5.98 4.80

Intercalation compounds of graphite

35

undergoing intercalationdesorption cycles with an intercalant such as potassium (Crespin et al. 1978). The arguments just presented assumed that the intercalate layer was commensurate with the graphite. Since the acceptor compounds are not generally commensurate, there is no reason for the bounding graphite layer to align as AjA. Thus the DaumasHerold domain model would apply in cases of incommensurate inter calants without the Burgers vector constraint on the thickness of the intercalate layer. Also graphiteLi compounds, which do not satisfy the stacking fault p constraint di a0 = 3, have in fact been grown with an AjA stacking arrangemen t (Guerard and Herold 1975, Basu et al. 1979). This observation suggests that these compounds might be formed by a di erent mechanism (possibly the small size of the Li ion allows di usion of Li between adjacent graphite planes). It is not clear at the present time that a strain mechanism can by itself, fully account for intercalation and staging. In fact, some authors have stressed the importance of electrostatic interactions for specic aspects of the intercalation process. Pioneering work in this area was done by Salzano and Aronson (1966a), who modelled the intercalation process by considering explicitly the interaction of charged alkali metal layers with the surrounding graphite bounding layers, which were treated as metallic layers. The bonding energy produced by this interaction was calculated, using an image force model and the results were compared with thermodynamic data (see section 2.3). Since good agreement was obtained, a number of workers subsequently tried to explain intercalation and staging in terms of electrostatic interactions. For example, Young (1977) argued that for bromine intercalation, a critical degree of ionic Br adsorption on the exposed basal surface (related to the inter2 calation threshold) was a necessary prerequisite for intercalation. Using elipsometric techniques to study bromine adsorption on a basal graphite c-face, Hibbs and Young (1978) found that the composition of a saturated adsorbed monolayer was Br 2Br2 . 2 According to Youngs model, the intercalation process sets up a long wavelength charge density wave with a maximum charge transfer upon initiation of the intercalation process, and a decreasing charge transfer as intercalation proceeds. He argued that the long-range interaction (700 A) is provided by the charge density wave-periodic lattice distortion which induces Br ions to seek positive maxima in 2 the oscillatory screening charge density wave. Safran and Hamann (1980, 1981) considered both elastic and electrostatic interactions on the intercalation process for a general conguration of intercalate layers, but neglected c-axis band dispersion. From this calculation they concluded that elastic interactions were dominant for the kinetics of intercalation, while electrostatic processes were more important for determining the equilibrium state. Because the ThomasFermi screening of a charged intercalate layer by the surrounding graphite medium is non-exponential , but follows a power law, Safran and Hamann argued that the electrostatic interaction could be long range. They further showed that the energy was minimized when the intercalant was arranged in a pure stage conguration, thereby showing that staging corresponds to the equilibrium state. Although many careful experimental and theoretical studies of the staging transformation have already been carried out, further detailed e ort is needed to clarify the roles of elastic and electrostatic interactions in intercalation and staging.

36

M. S. Dresselhaus and G. Dresselhaus

3. Structural properties 3.1. Classication of types of crystalline order The most important structural characteristic of graphite intercalation compounds is the occurrence of separate graphite and intercalate layers due to the very strong intraplanar binding and the very weak interplanar binding. Thus the graphite layers retain the basic properties of pristine graphite, and the intercalate layers behave similarly to the parent intercalate material. These layers can in addition exhibit varous types of ordering, the most important of which is the stage ordering along the c-direction (see gure 2), discussed in the previous section. The stage index, n, which denotes the number of graphite planes between adjacent layers, is readily determined from (00l) X-ray di ractograms (see section 2.2). Other types of ordering prevalent in graphite intercalation compounds include stacking order of the graphite layers, in-plane ordering in the graphite and intercalate layers as well as the interlayer correlation between these layers, commensurate and incommensurate molecular ordering, and interlayer intercalate stacking order. In the following, these types of ordering are described and examples of intercalation compounds exhibiting each ordering type are given. The graphite layers themselves exhibit denite stacking arrangements in the intercalation compounds. There are two basic types of graphite stacking arrangements: type I where the graphite layer arrangement about the intercalate layer is the same (AjA) and type II where it is di erent (AjB). Examples of several stacking arrangements within each category have been reported. In some cases the stacking arrangement is temperature dependent, and in other cases, more than one stacking arrangement can coexist at the same temperature. An example of the type I stacking is found in the stage I alkali metal compounds with the intercalants K, Rb and Cs, where the stacking arrangement is AjAjA, while for stage 2 compounds, the stacking arrangement is ABjBCjCAjAB, where the vertical line denotes an intercalate layer and A, B and C denote graphite layers (Nixon and Parry 1968). For higher stage compounds, the graphite layers between two intercalate layers will alternate ABAB or ACAC or BCBC in accordance with the usual stacking sequence in pristine graphite. This stacking sequence can be seen in gure 14 which illustrates the graphite crystal structure (space group P63 /mme). A rhombohedral form of graphite also occurs, exhibiting the ABCABC stacking arrangement and corresponding to the space group R32/m (Wycko 1964), as shown in gure 15. For these compounds the insertion of an intercalate layer results in a rhombohedral slippage across intercalate layers, though adjacent graphite layers retain their usual hexagonal ABAB stacking sequence (see gure 2). An example of the ABjBAjAB stacking sequence is found in graphiteSO3 compounds. In the case of graphiteHNO3 , Herold (1979) has also reported an ABjBAjAB stacking though Nixon et al. (1966) reported an ABjBCjCA stacking arrangement for this system. Examples of the type II graphite layer stacking arrangement AjB are found in a number of acceptor compounds. For example, the stage 2 graphiteSbF5 compound exhibits the hexagonal type II arrangement ABjABjAB (Herold 1979) while for stage 4 graphiteMoCl5 both the hexagonal AjBABAjBA . . . and rhombohedral AjBCABjABCAjB . . . stacking arrangements have been reported (SymeJohnson 1967). For some intercalants such as H 2 SO4 both type I and type II graphite stacking arrangements have been observed, as, for example, for stage 2 graphiteH2 SO4 where both AjABjBAjABjB . . . and AjBAjBAjBA . . . stacking arrangements have been reported (Herold 1979). On the other hand, one cannot

Intercalation compounds of graphite


Table 6. Metal sandwich thickness for several low stage compounds. Metal sandwich thickness ds in A Intercalant K Rb Cs Li HNO3 SO3 HClO4 Cl2 O7 SbCl5 SbF5 Br2 AlCl3 FeCl3 AsF5 Author Parry Underhill Underhill Guerard Rudor Fuzellier Rudor Fuzellier Me lin Me lin Sasa Rudor Rudor Markiewicz Stage 1 5.35 5.65 5.94 3.706 7.84 7.96 7.94 7.98 9.42 8.46 9.54 9.37 Stage 2 5.39 5.70 6.01 3.71 7.79 7.90 7.77 7.78 9.36 8.41 7.04 9.48 9.45 8.15

37

Stage 3 5.40 5.70 3.70 7.79 7.90 7.60 7.70 9.36 8.41 7.04 9.51

rule out the possibility of intercalation compounds (for example, with large intercalate layer thickness di ) where there is no stacking correlation between graphite layers on either side of the intercalate layer (or intercalate sandwich for the large molecular intercalants). It is also of interest to point out that the intercalation process can be used to stabilize the hexagonal graphite stacking arrangement. For example, when metastable rhombohedral graphite is intercalated and subsequently desorbed, the stable hexagonal stacking results (Herold 1979). The basic symmetry within the graphite interior layers remains una ected by intercalation. Not only is the ABAB hexagonal stacking preserved, but the spacing between graphite layers in the intercalation compounds is essentially the same as in pristine graphite c0 3:35 A (Hennig 1959). Small variations in interplanar spacing are expected for graphite layers adjacent to the intercalate layer (the graphite bounding layers) because of the di erent charge distributions at this interface, but such variations in interplanar spacing have only been explored for a few cases. In this connection table 6 gives data for several donor and acceptor intercalants for the dependence on stage index of ds , the distance between graphite layers between which an intercalate layer is sandwiched, or more succinctly ds is called the intercalate sandwich thickness. Thus the repeat distance, as measured by (00l) X-ray di ractograms, can to a good approximation be related to c0 and ds by Ic ds n 1c0 ; 3:1 where ds di c0 (see table 5). The results of table 6 show that for most donor compounds the intercalate sandwich distance ds increases from stage 1 to stage 2 but decreases for most acceptor compounds. For both donors and acceptors, ds is in most cases essentially independent of stage for n 2. These conclusions are in agreement with preliminary results on the charge distribution about the graphite bounding layers as calculated from (00l) X-ray intensity measurements, showing that for donor compounds the bounding layer charge distribution shifts towards the intercalate layer and thus decreases the bonding between the graphite boundary layer and the

38

M. S. Dresselhaus and G. Dresselhaus

graphite interior layer, whereas the acceptors show the opposite e ect, namely an increase in bonding between the bounding layer and the interior layer (Leung et al. 1981a). Also consistent with these conclusions are EXAFS (extended X-ray absorption ne structure) measurements on graphiteK samples by Caswell et al. (1980). These experiments are sensitive to the carbonpotassium distances and conrm the increase in ds from stage n 1 to stage n 2 compounds. The in-plane ordering in the graphitic planes is the same as in pristine graphite, with approximately the same in-plane lattice constant a0 2:46 A, or a nearest neighbour carboncarbon distance of 1.42 A (Wycko 1964). A small in-plane expansion due to intercalation has, however, been found for graphiteK donor compounds by Nixon and Parry (1969), with an approximate linear dependence of the nearest-neighbou r carboncarbon distance on reciprocal stage (1=n) given by dC-C 1:4203 0:0113=n A :

3:2

Later work by Guerard et al. (1976) on other donor stage 1 compounds yielded a relation between dC-C and the ionic radius r and the valence Z of the metal (Z 1 for alkali metals, 2 for alkaline earths and 3 for lanthanide metals) dC-C 1:420 0:032 66Z=r A dC-C 1:420 0:019 58Z=r A

f or f or

C8 X; C6 X:

3:3 3:4

For all the donor intercalants that were examined, K, Cs, Li, Ba, Sr, Eu, Yb, intercalation resulted in a lattice expansion. The dependence of dC-C on 1=n is consistent with a sharing between graphite layers of the lattice strain resulting from intercalation. The observed dependence of dC-C on Z is also consistent with the expected increase in strain associated with increased intercalate layer charge, though the larger increase in dC-C for C8 X as compared with C6 X structures is not understood. Raman scattering experiments suggest that a small in-plane contraction occurs in a number of acceptor compounds (Underhill et al. 1979, Gualberto et al. 1980) in contrast with the in-plane expansion reported for the donor compounds. Such a contraction in acceptor compounds has recently been veried by direct X-ray measurement in NiCl2 (Flandrois et al. 1980), in AsF5 (Markiewicz et al. 1980b) and in FeCl3 and AlCl3 (Krapchev et al. 1981) compounds, and the contraction has been shown to be of much smaller magnitude in acceptor compounds as compared with the expansion in donor compounds of the same stage. In the low temperature phase the intercalant is arranged in an ordered network, not necessarily commensurate with the in-plane graphite site ordering, and often exhibiting coexisting ordered structures (Berker et al. 1980). In a few cases the ordering is simple and has been deduced directly from (hk0) di raction data, such as the p2 2R08 structure for the stage I alkali metals C8 K, C8 Rb and C8 Cs shown in p p gure 20 (Ru dor and Schulze 1954), the p 3 3R308 structure for C6 Li shown p p in gure 21 (Gue rard and He rold 1975), or the p 7 7R 19:18 structure shown in gure 23 and observed at low temperatures for higher stage graphiteRb and graphiteCs compounds (Parry 1977, Kambe et al. 1980). In the notation used for the in-plane superlattices, p denotes a primitive unit cell, (i j) denotes the unit vectors measured in units of a0 2:46 A in the hexagonal superlattice primitive cell, and R denotes the angle of rotation of the unit vectors of the superlattice relative to the graphite unit vectors. For other intercalation compounds, the di raction patterns are very complex and have not been fully interpreted. Some of the

Intercalation compounds of graphite

39

Figure 20. Schematic diagram of an in-plane unit cell for the p2 2R08 structure showing the carbon atoms at the corners of the hexagons and the intercalants as the open balls. In this diagram the intercalate and graphite layers are projected into a single plane. For this structure there are four equivalent intercalate sites labelled ; ; ; and on a given intercalate layer only one of these sites is occupied. The dashed lines represent the in-plane unit cell for pristine graphite. For C8 K and C8 Rb, the intercalant is arranged on sequential layers on ; ; , sites while for C8 Cs, the intercalant reportedly follows the ; ; sequence.

p p Figure 21. Schematic diagram of an in-plane unit cell for the p 3 3R308 structure showing the carbon atoms at the corners of the hexagons and the intercalants as the open balls. In this diagram the intercalate and graphite layers are projected onto a single plane. There are three equivalent intercalate sites labelled ; ; . On a given intercalate layer only one of these sites is occupied. At low temperatures the lithium intercalant in C6 Li is arranged on sequential ; ; sites, whereas for C6 Eu the stacking is reported to be .

40

M. S. Dresselhaus and G. Dresselhaus

p Figure p Schematic diagram of an in-plane unit cell for the honeycomb h 12 22. 12R308 structure. Of the twelve possible intercalate sites on a given intercalate layer, two are occupied as indicated by the open circles on B and C , and the unoccupied site A is used as a layer designation. Equivalent intercalate arrangements are achieved by cyclic permutation of the A , B and C sites. In addition, there are four equivalent arrangements for each of the A , B , C sites (labelled by ), so that in all there are 12 equivalent intercalant arrangements. No known examples of this superlattice exists.

p p Figure 23. Schematic diagram of an in-plane unit cell for the p 7 7R19:18 structure. The seven possible intercalate sites are labelled.

complexity of the observed di raction patterns may be due to the coexistence of multiple ordered phases in equilibrium (Berker et al. 1980). For some intercalation compounds and in certain temperature ranges, the site ordering between the intercalate layer and the adjacent graphitic layers is correlated or commensurate, as for example for the alkali metal compounds exhibiting the p p p p p2 2R08, p 3 3R308 and p 7 7R19:18 structures enumerated above. For certain temperature ranges and stages, some of the alkali metal compounds exhibit incommensurate in-plane structures as discussed further in section 3.4.

Intercalation compounds of graphite

41

For molecular intercalants, the intercalate layers may be commensurate as in the case of graphiteBr2 (Eeles and Turnbull 1965) or incommensurate as in the case of graphiteFeCl3 (Cowley and Ibers 1956). It is also thought that, in analogy with three-dimensional solids, there is an ordering of the molecular axes superimposed on the intercalate site ordering, where in this case the intercalate site is identied with the molecular centre of mass. Since no complete structural determination has been carried out for a molecular intercalation compound, the ordering of the molecular axis is incompletely specied. A correlation in the site ordering of the intercalant in sequential intercalate layer planes has also been observed in a number of compounds, particularly in stage 1 compounds. Because of the relatively large size of the intercalant ions or molecules relative to the carbon atoms, the correlation of the site arrangement of the intercalant on sequential intercalate layers allows for closer packing of the intercalate layers in the three-dimensional stacking of the crystal. Even when longrange correlation is absent, these close-packing considerations imply that the placement of the intercalant on a particular site on one layer tends to exclude the placement of the intercalant on a similar site in the nearest neighbour intercalant layers. A well-documented example of interlayer intercalate stacking order is the stacking of sequential intercalate layers in well-annealed stage 1 compounds C8 K and C8 Rb, as originally observed by Rudor and Schulze (1954) and layer conrmed by many other workers. This stacking sequence is illustrated in gure 20 showing a projection of the in-plane p2 2R08 structure and in gures 1 and 2, showing the interplanar stacking. For the case of the rst stage C8 Cs compound, which also exhibits a p2 2R08 in-plane superlattice structure, the interlayer intercalate stacking order has been reported to have the stacking sequence (Parry 1971, Ellenson et al. 1977, Gue rard et al. 1978) rather than the sequence, presumably because of the large size of the Cs ions. Another example of the interplanar intercalate stacking order is found in the low temperature phase of C6 Li p p which has the p 3 R308 in-plane ordering and an interlayer intercalate stacking ordering as shown in gure 21 (Kambe et al. 1979). It should be noted that the above discussion on interlayer intercalate exclusion argues against an interplanar intercalate stacking sequence for C6 Li at 300 K, as reported by Guerard and Herold (1975). Another example where not all sites on the intercalate p p layer are occupied sequentially is C6 Eu, which exhibits a p 3 3R308 superlattice but has intercalate layer stackings (El Makrini et al. 1980). For compounds exhibiting interlayer intercalate stacking order, the unit vector in the c-direction is a multiple of Ic , the intercalant repeat distance. For example, stacking implies a unit vector in the c-direction of 3Ic Close-packing geometrical considerations also appear to be relevant to the interlayer intercalate stacking of higher stage compounds. For the case of second stage alkali metal compounds in phases where the intercalant exhibits an in-plane ordering commensurate to the graphite bounding layers, long-range interlayer intercalate stacking order has been reported. For example, the low temperature (T 9 98 K) phase of stage 2 graphiteK has been reported to exhibit in-plane intercalate ordering commensurate to the graphite layers, and long-range intercalate stacking order which has not been completely identied (Parry and Nixon 1967, Hastings et al. 1979, Zabel et al. 1979). At higher tempertures 98 9 T < 126 K, where the in-plane structure becomes slightly incommensurate with the graphite bounding layers (Hastings et al. 1979, Berker et al. 1980), interlayer intercalate

42

M. S. Dresselhaus and G. Dresselhaus

stacking correlations are still found over a distance of 30 A (estimated from the linewidth of the di raction peak), but the long-range ordering is no longer present (Hastings et al. 1979, Zabel et al. 1979). At yet higher temperatures T > 126 K, the range of these interlayer stacking correlations is greatly reduced, and in this region the intercalate layers have consequently been called `two-dimensonal by Hastings et al. and Zabel et al., though interactions between the intercalant and the graphite bounding layers persist. The in-plane intercalate structure represents a coexistence of ordered structures both commensurate and incommensurate with the graphite and of disordered regions. The second stage graphiteRb and graphiteCs compounds also exhibit long-range interplanar intercalate stacking order at low temperatures, though the precise stacking arrangements have not been fully established (Parry et al. 1969). Also of interest is the interlayer intercalate stacking correlation in the case of intercalants with in-plane ordering that is incommensurate with the graphite bounding layers. For example, the incommensurate graphiteFeCl3 compounds exhibit such interlayer intercalate stacking correlations (Herold 1979), presumably due to the large size of the intercalate molecules which imposes restrictions excluding intercalant stacking in order to yield a closer packing of the three-dimensional structure. 3.2. X-ray, electron and neutron di raction for structural studies X-ray, electron and neutron di raction techniques provide the most powerful tools for studying the symmetries associated with the ordered structures found in graphite intercalation compounds, and for determining the critical temperature s for the structural and orderdisorder transformation s that occur in these materials. Although similar in many ways, X-ray, electron and neutron di raction also provide complementary information on graphite intercalation compounds. Various X-ray scattering techniques have been applied to structural studies of graphite intercalation compounds, ranging from DebyeScherrer powder patterns, to -2 di ractometer scans to obtain both (hk0) and (00l) reections, to rotating single crystal structure determinations, and di use X-ray scattering studies. Except for the rotating single crystal structure determination, all of these techniques can be applied to intercalation compounds based on a HOPG (highly oriented pyrolytic graphite) host material, which has a random arrangement of crystallites of 1 mm diameter in the layer planes, but with c-axis alignment to better than 18 (Moore 1973). The Ic measurements, stage index determinations, and in-plane density measurements are most sensitively made using (00l) reections which can be equally well carried out on samples based on HOPG, as discussed in section 2.2. On the other hand, the in-plane ordering is most sensitive to (hk0) reections, while the correlated graphiteintercalate stacking order is most sensitive to scans of l values for xed non-vanishing h and k values. The geometry for the -2 scans used to determine the (00l) and (hk0) reections for an HOPG sample are contrasted in gure 24. Whereas the (00l) measurements are made in reection, the (hk0) measurements are made on transmission through thin samples. If the sample is a single crystal, then it must be rapidly rotated perpendicular to the scan direction in order to satisfy the Bragg condition during the -2 scan. For samples based on HOPG, such rapid rotation is not needed because of its polycrystallinity with regard to the X-ray beam. When the proper for a di raction peak is determined in single crystal experiments, then a slow rotation about the c-axis is used to determine the multiplicity of each spot and the angular distances of the reections. Using such X-

Intercalation compounds of graphite

43

Figure 24. Geometries for measurement of X-ray (00l) and (hk0) reections from intercalation compounds based on single crystal and highly oriented pyrolytic graphite host materials.

ray techniques, several workers have been able to obtain good X-ray spot patterns for intercalated graphite samples based on an HOPG host material (Nixon et al. 1966, Parry 1979, He rold 1979). Di use X-ray studies have also been made on graphite intercalation compounds to study short-range order e ects and interlayer intercalate correlation e ects (Hastings et al. 1979, Zabel et al. 1979, Clarke et al. 1980). Application of X-ray di raction techniques to the analysis of (hkl) reections using single crystal host materials provides the simplest and most direct method for a complete structural determination. This technique has, however, been applied only to a very few compounds for a full structural determination. The compounds that have been studied in great detail are the stage 1 alkali metals C8 K, C8 Rb, C8 Cs (Rudor and Schulze 1954) and C6 Li (Gue rard and Herold 1975) and for these a complete structural analysis has been made. Detailed X-ray studies have also been carried out for a number of other systems. The most extensively studied have been the higher stage compounds with K, Rb and Cs (Rudor and Schulze 1954, Nixon and Parry 1969, Parry and Nixon 1967, Nixon et al. 1969, Parry 1977) where attention has been given to the in-plane intercalate ordering, the stacking order of the graphite layers and of the intercalate layers, and the correlation between the graphite and intercalate layers. These systems are currently under intense study using X-ray di raction techniques with particular regard to study of the variety of structural and orderdisorder transformation s that occur in these systems as a function of temperature (see section 3.4). Attention has also been given to the higher stage compounds with the donor Li (Guerard and Herold 1975), but a full structural determination has not yet been completed for the higher stage compounds with Li. Molecular intercalation compounds have also been studied extensively using Xray di raction, but no complete structural determinations have been reported. In most cases the molecular intercalants are non-planar so that the structure of the molecular intercalate layer or intercalate sandwich can be quite complex. For a number of molecular systems the stacking order of the graphite layers has been studied (see section 3.1 for some examples) and the correlation between the intercalate layers and the graphite bounding layers has likewise been studied. Little is known about interlayer intercalate stacking order, except that to accommodate intercalants in a close-packed arrangement the interlayer intercalant exclusion argument given in section 3.1 is expected to apply. The specic systems that have

44

M. S. Dresselhaus and G. Dresselhaus

been studied to date are reviewed in some detail by Ebert (1976) and by Herold (1979). These systems include the intercalants: Br2 (Eeles and Turnbull 1965, Sasa et al. 1971), FeCl3 (Rudor and Schulz 1940, Cowley and Ibers 1956), MoCl5 (Syme Johnson 1967), CrCl3 (Vange listi and He rold 1976), SbCl5 (Me lin and Herold 1975), H2 SO4 (Rudor and Ho man 1938), HNO3 (Rudor 1939, Nixon et al. 1966, Fuzellier et al. 1977b and He rold 1978), alkali metals with benzene and toluene (Merle et al. 1977, Bonnetain et al. 1977). For electron di raction the electron beam diameter can be made comparable to the in-plane crystallite size in HOPG, and thus electron di raction has some distinct advantages for probing the in-plane intercalate ordering in these materials. Because of the small electron beam size, electron di raction e ectively probes HOPG-based samples as if they were single crystal. However, the relatively small penetration depth of electron beams requires the samples to be thin (1000 A). Therefore the electron di raction technique is most suitable for study of in-plane structure using thin c-face samples. Yet the samples must be defect-free and characteristic of the bulk intercalation compound. Because of multiple di raction e ects (Hirsch et al. 1965), the electron di raction patterns can be di cult to interpret quantitatively if the site ordering on the intercalate and graphitic planes is incommensurate. In this case the X-ray patterns are simpler to interpret because of the absence of multiple di raction e ects. The X-ray and electron di raction techniques provide complementary information because of the di erence in wavelength of typical electron and X-ray beams. In order to have su cient bulk penetration, electron beam energies are typically 50 80 keV, resulting in a very small electron wavelength and a very large Ewald di raction sphere (radius 2=). Thus for electron di raction with the beam incident along the c-axis, the Ewald sphere e ectively intersects a single plane in reciprocal space (l 0) as shown in gure 25, unlike the case of X-ray di raction where the Laue relation ki ks G between the incident and scattered wave vectors is satised for points on di erent planes in reciprocal space normal to the X-ray beam direction. Solutions to the Laue condition are satised both for G 0 and for

Figure 25. Reciprocal space construction of the Ewald sphere for the electron di raction experiment showing reciprocal lattice points and the incident and scattered wave vectors for the electron beam.

Intercalation compounds of graphite

45

reciprocal lattice points on the Ewald sphere. Because of the qualitatively di erent wavelengths for the X-ray and electron beams, these two techniques sample somewhat di erent points in reciprocal space. Electron di raction has been applied primarily to studies of the in-plane structure as a function of temperature, thereby yielding information on structural and orderdisorder transformation s (see section 3.4). Because of the vanishing of the structure factor for certain intercalate layer stacking sequences (such as for the p p p 3 3R308 structure, and for the p2 2R08 structure), electron di raction has provided a valuable tool for the investigation of interlayer intercalate stacking order (Kambe et al. 1979, 1980a). Intercalation compounds that have been studied with electron di raction include the intercalants K (Halpin and Jenkins 1970, Evans and Thomas 1975, Chung et al. 1977, Berker et al. 1980), Rb (Chung et al. 1977, Kambe et al. 1980), Cs (Chung et al. 1977), Li (Kambe et al. 1979), Br2 (Eeles and Turnbull 1965, Evans and Thomas 1975, Chung et al. 1977), FeCl3 (Cowley and Ibers 1956, Evans and Thomas 1975), SbCl5 (Me lin and Herold 1975) and MoCl5 (Syme-Johnson 1967). Of these systems, those that were studied using combined Xray and electron di raction include Br2 (Eeles and Turnbull 1965), FeCl3 (Cowley and Ibers 1956) and MoCl5 (Syme-Johnson 1967). In principle, neutron di raction experiments could also be used to provide structural information on graphite intercalation compounds. Since the neutron di raction technique requires large sample volumes (Brockhouse et al. 1963), this technique has only been applied to compounds based on HOPG (Ellenson et al. 1977). Because of the polycrystalline in-plane arrangement in HOPG, neutron di raction has been primarily useful for studying c-axis structure, such as staging, the graphitic layer stacking order, the intercalate layer stacking-order, and order disorder transformation s associated with interlayer stacking order (Ellenson et al. 1977). Neutron di raction experiments have thus far only been reported for the compounds C8 Cs and C8 Rb (Ellenson et al. 1977). In the case of C8 Cs, the intercalate stacking sequence was conrmed while neutron di raction experiments on C8 Rb produced di erent c-axis stacking patterns depending on the measurement temperature and on previous sample treatment history. This is illustrated in gure 26, where the neutron counting rate is plotted against the dimensionless wave vector c for a C8 Rb sample prepared from an HOPG host material, where c kz ds =2 and ds is the interlayer separation between two sequential graphitic layers between which the intercalate is sandwiched. The upper trace in gure 26, taken at 290 K, is characteristic of the stacking by successive Rb layers. For this stacking sequence, the structure factor vanishes for the di raction peaks at c integer. The high temperature phase at 721 K (middle trace) is characteristic of a well-ordered phase for which the intercalate layers have the stacking arrangement. The bottom trace, taken at 290 K, is the spectrum observed from an as-grown C8 Rb sample in which the interlayer intercalate stacking is not well ordered. The upper two traces were obtained on samples which were annealed for 2 days at 747 K. At this elevated temperature, the Rb layer was reported to exhibit a liquid-like or highly disordered in-plane structure, while maintaining the stage 1 ordering along the c-axis. The di raction pattern for the well-ordered interlayer intercalate stacking sequence was observed only after the sample had been annealed at high temperature (for example, 721 K) and subsequently cooled (Ellenson et al. 1977).

46

M. S. Dresselhaus and G. Dresselhaus

Figure 26. Neutron di raction patterns of Ellenson et al. (1977) for rst stage C8 Rb as a function of dimensionless wave vector c k2 ds =2. (a) Di raction pattern at 290 K is consistent with a well-annealed sample with an ordered ; ; ; intercalate layer stacking sequence. (b) Di raction pattern at 721 K is consistent with an ; intercalate layer stacking sequence. (c) The pattern at 290 K obtained from an `as grown sample shows disorder with regard to intercalate stacking.

3.3. Relation between the structure of intercalation compounds and their parent materials An intercalation compound can be considered as a superposition of a graphitic and intercalate skeleton (Hennig 1959). This view introduces a major simplication in treating the structural properties of intercalation compounds by relating their constituents to their parent materials from which they originate. This modelling of the structure of an intercalation compound is best illustrated by the alkali metal compounds which have been studied most extensively. In this section it is rst shown that for the stage 1 alkali metal compounds the preferred in-plane superlattice structure of the intercalation compound is the one that requires minimal distortion of the metallic parent structure to form a layer plane in the intercalation compound. It is then shown that for the higher stage alkali metal compounds, the in-plane intercalate density in the intercalation compound is determined by the in-plane density in the parent metal. Using a model where the diameters of the carbon and intercalant atoms are taken as the nearest-neighbou r distances in the parent materials, it is shown that the c-axis distances in the intercalation compound are consistent with a close packing of spheres, in which the intercalate site locations favour a uniform lling of space with intercalant. A discussion is then given of geometrical considerations applicable to molecular acceptor compounds. For the case of the stage 1 alkali metals an argument is presented to show that the preferred in-plane superlattice structure in the intercalation compound follows from the geometrical relations of the parent metallic solid. To achieve the threefold symmetry of the intercalate layers, we consider layer planes normal to the 111 axis

Intercalation compounds of graphite

47

Table 7. Characteristic distances in b.c.c. alkali metals and in their intercalation compounds. X Li Na K Rb Cs
a b

8 aXA 3.51 4.29 5.23 5.59 6.05

p2

3 aXA

3 c 8 aXA 2

Ic A d Stage 1 3.71 4.50 5.35 5.65 5.94

2.87 3.50 4.27 4.56 4.94

3.04 3.72 4.53 4.84 5.24

B.c.c. lattice constants aX taken from Wycko (1964),m where X denotes the intercalant. 8 p Nearest-neighbour distance projected on a (111) plane. For the p2 2R08 structure 2=3aX is to be 8 p p p compared with 2a0 4:91 A , while for the p 3 3R308 structure, 2=3aX is to be compared with 8 a0 2:46 A . c Nearest-neighbour b.c.c. distance, to be compared with 2a0 4:91 A for the intercalants forming the p2 2R08 structure. d For all intercalants except Na, Ic for stage 1 compounds is measured directly (see table 6). For Na, measurements of Ic n for higher stage compounds are used to infer Ic 1 from the relation Ic n Ic 1 n 1c0 , since no stage 1 graphiteNa compound has been prepared.

of the parent body-centred-cubi c (b.c.c.) alkali metal crystal. The site arrangemen t on each of the (111) lattice planes of the b.c.c. system is a triangular net that is readily represented by a hexagonal coordinate system. For the b.c.c. structure the p nearest-neighbou r alkali metal distance is equal to 3aX 8=2, where aX is the b.c.c. 8 metallic lattice constant (see table 7). If we consider three sequential (111) planes and project all three on to a single plane, we obtain a network with hexagonal symmetry p and a nearest-neighbou r distance of 2 aX. Table 7 shows that for metallic K, Rb 8 3 p2 and Cs, the values of 3 aX are close to 2a0 4:91 A, the nearest-neighbour distance 8 for the p2 2R08 in-plane structure for these stage 1 compounds. Also listed in p this table are values for 3aX=2, the nearest-neighbou r distances before projection 8 p 8 on to a coplanar arrangement. The similarity of 2 aX to 2a0 for K, Rb and Cs as 3 shown in gure 27, and the dissimilarity of these quantities to 2a0 for Na and Li has been used to explain why stage 1 compounds with K, Rb and Cs form in-plane p2 2R08 superlattices, and Na and Li do not (Dresselhaus and Dresselhaus 1979). A similar argument has been used to explain why Li forms an in-plane p p p 3 3R308 superlattice, where the nearest-neighbou r distance is p 3a0 4:26 A, which is roughly equal to the nearest-neighbou t distance in the p (111) plane of the metal 2aX 4:96 A . The lattice constant of b.c.c. metallic Na 8 (Wycko 1964) is intermediate between that of Li and those of the heavier alkali metals K, Rb and Cs, so that neither the condition for the p2 2R08 superlattice p p p p (aX 6a0 ) nor the condition for the p 3 3R308 superlattice (aX 2 a0 ) is 8 8 3 satised approximately (see gure 27). Thus neither superlattice can be realized for the sodium intercalant without major distortion. In this connection it has been noted that Na does not readily intercalate into graphite (Asher 1959). It is of interest to note that the intercalants K, Rb and Cs are more densely packed in the stage 1 compounds than in the b.c.c. parent metals, as shown in gure 27. To account for this reduction in molar volume, an ionization of the metal intercalant has been proposed (Setton 1964, Herold 1979). It is also signicant that the intercalate sandwich distance ds is of comparable magnitude to the b.c.c. metallic lattice constants and to the nearest-neighbou r separations in the metal.

48

M. S. Dresselhaus and G. Dresselhaus

Figure 27. Projections of the b.c.c. alkali metal (111) lattice planes on a hexagonal graphite network centred at the middle of the indicated unit cells. The groups marked (0), () and () correspond to the three planes which pass through the body centre (0), or 8 8 above () or below () the median plane at a distance 2aX, where aX is the b.c.c. lattice constant. (a) Projection of the b.c.c. (111) planes of K, Rb, Cs onto the graphite network. If the (0), () and () sites are projected onto a single plane, and the atoms are slightly displaced, the p2 2R08 structure is obtained, and the corresponding unit cell is indicated on the gure. (b) Projection of the b.c.c. (111) planes of Na and Li onto the graphite network. By considering only one of the (111) p p Li lattice planes, as for example the (0) plane, the p 3 3R08 structure is obtained with only slight displacement of the Li ions, and the corresponding unit cell is indicated on the gure. In the case of Na, a0 is such that the (111) lattice planes Na cannot easily be made commensurate with the graphite network.

The in-plane structures for the higher stage alkali metal compounds with K, Rb and Cs exhibit a lower in-plane density than for stage 1 compounds, and this is closely related to geometrical relations found in the parent metal compound. For p p these compounds a honeycomb h 12 12R308 structure with two atoms per unit cell and a nearest-neighbou r distance of 2a0 4:91 A was rst proposed by Rudor and Schulze (1954), but not conrmed by later X-ray and electron di raction measurements on higher stage graphiteK compounds (Parry and Nixon 1967, Hastings et al. 1979, Zabel et al. 1979, Berker et al. 1980). In fact, evidence has been given in support of coexisting phases in second stage graphiteK at room temperature. For the higher stage compounds with Rb, the less dense p p p p 7 7R19:18 structure with a nearest-neighbou r distance of 7a0 6:51 A was found by Kambe et al. (1980) using electron di raction techniques to be the dominant structure, in good agreement with the measured in-plane density for stage 2 graphiteRb (Leung et al. 1979). In the following, it is shown that the geometrical relations in the b.c.c. parent metals are also consistent with these in-plane structures. In table 8 the metallic atomic volumes Vm aX3 =2 are given for each of the b.c.c. alkali metals. In the 8 stage 1 intercalation compounds, space is lled by unit cells with volumes Vi Ai Ic . Note that for the K, Rb and Cs intercalants, the in-plane packing of the metal atoms

Intercalation compounds of graphite


Table 8. Metallic volumes and intercalant volumes for stage 1 alkali metal compounds a;b . B.c.c. metal Li Na K Rb Cs
a b

49

Vm A 21.62 39.48 71.53 87.34 110.72

Ai A

Vi A

p 3 3 2 a0 15:72 2 p 2 3a2 20:96 0 p 2 3a2 20:96 0 p 2 3a2 20:96 0

58.32 112.14 118.42 124.50

Metallic atomic volume Vm 1 aX3 . 8 2 Ai is the area of an in-plane unitp in the intercalation compound. For C6 Li, the cell p in-plane unit cell has the p 3 3R308 structure (see gure 21), while for C8 K, C8 Rb, C8 Cs the in-plane structure is p2 2R08 (see gure 20). No stage 1 graphiteNa compound has been prepared to date.

Table 9. Metallic in-plane areas in relation to areas of unit cells for higher stage alkali metal compounds.
Dominant in-plane structure n 2 p 3 p p 3R308 Intercalant in-plane area Ai b A 2 p 3 3 2 a 15:72 2 0 p 2 2 3a0 20:96 p 2 7 3a0 36:68 2 Calculatedc from Am and chemical formula C X 5.43 8.10 12.05 13.88 16.13 Measuredd from (00l) integrated intensities 10.5 14.0 16.4

B.c.c. metal Li Na K Rb Cs
a

Metallic area Am a A 2 14.22 21.22 31.58 36.10 42.26

p2 2R08 Mixed p p p 7 7R19:18 Incommensurate

p p Metallic area Am Vm = 3=4aX 2aX 2 = 3. Values for Vm and aX are found in tables 8 and 7 8 8 8 respectively. b The intercalant in-plane area Ai refers to the superlattice unit cell. c The calculated reciprocal in-plane density in the chemical formula C X is found directly from the metallic area Am . d From Leung et al. 1979.

is somewhat tighter in the intercalation compound than in the metal (see gure 27). However, the volume available to the conduction electrons in the intercalation compound Vi is greater than that in the metal Vm . Using the distance between p neighbouring (111) planes of 3aX=4, an e ective metallic in-plane area Am 8 p p Vm = 3aX=4 2aX2 = 3 is found and given in table 9, incorporating the values 8 8 for aX given in table 7. Also given in table 9 are reported in-plane structures for 8 higher stage (n 2) alkali metal compounds, and from these structures the area per intercalant Ai is readily found using the formulae indicated in table 9. Excellent agreement is obtained between Am and Ai for the second stage Rb compound. For the case of Na, Asher (1959) has reported a stage 8 compound C64 Na which suggests the p2 2R08 in-plane structure, and gives excellent agreement between Am and p 2 3a2 . For the case of the higher stage Li compound Guerard and Herold (1975) 0

50

M. S. Dresselhaus and G. Dresselhaus

p p have found the p 3 3R308 structure which gives an in-plane area of p 3 3a2 =2 15:72 A 2 which is comparable to Am 7:11 A 2 . This agreement 0 between Am and Ai argues against the stability of a C18 Li stage 2 compound, as suggested by Basu et al. (1979). For the case of the high stage Cs compounds, table 9 suggests that the dominant in-plane structure is either incommensurate with the graphite lattice and corresponds to an in-plane density of C16:4 Cs (Leung et al. 1979), or consists of coexisting commensurate phases which average to the proper in-plane density (see table 9). An incommensurate in-plane structure has in fact been found in stage 2 graphiteCs by Clarke et al. (1979, 1980); this incommensurate structure has p p a nearest-neighbou r distance of 5.95 A, is less dense than the h 12 12R308 structure, but is more dense than the Am values in table 9 would suggest. It is of p p interest to note that the h 12 12R308 structure proposed by Rudor and Schulze for stage 2 graphiteK, but not conrmed by later di raction work, yields p Ai 3 3a2 31:44 A 2 in excellent agreement with Am in table 9. The experi0 mental di raction results indicate that closed-packed rather than honeycomb inplane intercalate structures are favoured. The possible commensurate close-packed structure in order of increasing Ai and decreasing density are p1 1R08, p p p p p p p 3 3R308, p2 2R08, p 7 7R19:18, p3 3R08, p 12 12R308. A geometrical argument is given below to account for the observed intercalate sandwich thicknesses ds , based on a close packing of the intercalant within the graphite network. A `billiard ball model is used to compute the radii of the carbon and intercalate atoms based on the nearest-neighbou r distances in the parent materials, so that the nearest-neighbou r metallic distance denes the intercalate diameter. The nearest-neighbou r distance in the intercalation compound dci is found from dci ds =22 1:422 1=2 A ;

using a0 2:46 A and the intercalate locations indicated in gures 20, 21 and 28. The sum of the radii rci is found from the nearest-neighbou r distances in the parent materials, as, for example, for the b.c.c. metals p 3 aX 0:71 A ; 8 3:6 rci 4 where the nearest-neighbou r distance of two carbon atoms is 1.42 A (see gure 28(a) for the geometrical relations for rst stage C6 Li and gure 28(b) for rst stage C8 K). Results comparing dci and rci for a variety of intercalants are given in table 10, from which it is seen that the intercalant is closely packed with respect to the graphite host. Values for the model intercalate sandwich thickness ds based on a close-packed intercalate arrangement can be calculated from the relation
^ ds 2r2 1:422 1=2 A ci

3:5

^ and it is seen in table 10 that the calculated ds is in good agreement with the observed ds for a variety of intercalants, thereby supporting the close-packing model for c-axis intercalate stacking. The justication for use of the metallic nearest-neighbou r distance rather than the ionic radii for the intercalant is that the graphite bounding layers, which have a high electron carrier density, overlap the metallic intercalate balls, thereby providing an overlap of the free electron gas plasma with the metallic intercalant. Setton (1964) and He rold (1979) have also been able to account for the observed intercalate sandwich distances ds in terms of a `billiard ball model where

3:7

Intercalation compounds of graphite

51

Figure 28. (a) Hard sphere model for rst stage C6 Li using the metallic Li radius. The upper gure shows an x-z cross-section of metallic Li balls and carbon balls having the diameter of the nearest-neighbour in-plane carbon distance. A close packed arrangement of carbon and Li balls yields the Ic distance of 3.72 A (Basu et al. 1979). The distance of c0 =2 1:68 A is indicated. The lower gure shows anp in-plane (x-y) p projection of the carbon and lithium balls. The unit cell for the p 3 3R308 structure is indicated. (b) Hard sphere model for rst stage C8 K using the metallic K radius. The upper gure shows an x-z cross-section of metallic K balls and carbon balls having the diameter of the nearest-neighbour in-plane carbon distance. The repeat distance of Ic 5:35 A is consistent with a close-packed arrangement of metallic K and carbon balls. The lower gure shows an in-plane (x-y) projection of the carbon and potassium balls. The unit cell for the p2 2R08 structure is indicated.

52

M. S. Dresselhaus and G. Dresselhaus

Table 10. Nearest-neighbour intercalatecarbon distance dci in relation to model for closepacked intercalant.
b c

Intercalant Li Na K Rb Cs Ca Sr Ba Sm Eu Yb
a b

Metallic radiusa 1.52 1.86 2.26 2.42 2.62 1.97 2.15 2.22 1.81 1.99 1.94

rci (A) 2.23 2.57 2.97 3.13 3.33 2.68 2.86 2.93 2.52 2.70 2.65

dci (A) 2.34 2.66 3.05 3.16 3.29 2.70 2.85 2.98 2.70 2.82 2.69

^ ds d Calculated (A) 3.44 4.28 5.22 5.58 6.02 4.55 4.97 5.13 4.16 4.59 4.47

ds e Observed (A) 3.72 4.50 5.35 5.65 5.94 4.60 4.94 5.25 4.58 4.87 4.57

p Metallic radius is nearest-neighbour distance in metal, as, for example, 3=4aX for the b.c.c. metals. 8 rci atomic radius 0:71 A (see gure 28). c The observed nearest-neighbour intercalantcarbon distance dci is found from equation (3.5). d ^ ds is calculated from equation (3.7). e The observed ds is found from Ic ds n 1c0 .

the diameters of the carbon atoms were taken to be equal to the c-axis graphite layer thickness c0 3:35 A, and for the intercalants, the ionic radii were used. One advantage of the model based on the metallic nearest-neighbou r distances is that the same size `billiard balls are used to explain both in-plane and c-axis distances observed in the intercalate compounds. Using the extended X-ray absorption ne structure (EXAFS) technique on graphiteK compounds of stages n 1, 2, 3, 4, Caswell et al. (1980) determined values for dci , obtaining dci 3:055 A for n 1 and 3.14 A for n 2, 3, 4. We note that the increase in ds from stage 1 to higher stage compounds is considerably larger than found from analysis of (00l) reections, as given in table 6. From the small linewidths of the EXAFS structure Caswell et al. conclude that the K are localized over the centres of the graphite hexagons, in agreement with electron di raction results of Berker et al. (1980). On this basis, Caswell et al. (1980) postulated that the K intercalant is a lattice gas, though other experimental work has not conrmed this model. Puckering distortions of the graphite bounding layer atoms nearest to the intercalant were suggested on the basis of the EXAFS measurements but the extent of puckering was not quantitatively established. Although a great deal less is known about the structure of the large number of molecular intercalation compounds that are formed, it is well established that wellstaged compounds can be prepared and that the intercalate layers are closely related to layer planes in the parent crystalline materials. The molecular intercalants tend to occupy relatively large volumes, resulting in large intercalate sandwich thicknesses ds (see table 6) that have been measured for many compounds and have been summarized in several review articles (Ebert 1976, Stumpp 1977, He rold 1979). The intercalant in most cases retains its molecular identity, though in other cases, changes in species occur upon intercalation. For example, the halogens (Br2 , ICl and

Intercalation compounds of graphite

53

Figure 29. Projection of a lattice plane of solid Br2 on the hexagonal graphite network. The dashed lines indicate the unit cell for solid Br2 and the molecular axes lie in this lattice plane. This gure shows that with only slight displacement of the Br2 molecules, the intercalate species can be placed on C8 X intercalate sites (small closed circles).

IBr) and some metal chlorides such as FeCl3 and AlCl3 are believed to maintain their molecular identity, while highly reactive acid intercalants such as HNO3 , AsF5 , SbCl5 and H2 SO4 are believed to transform upon intercalation (Ebert 1986, Herold 1979). Perhaps the simplest and best understood of the molecular intercalation compounds is the graphiteBr2 system. For this system it has been well established that many features of the intercalate layers relate closely to solid bromine. This molecular solid crystallizes in a layered orthorhombic structure with the unit cell within the (010) layers having a molecular site ordering which is almost commensurate with the in-plane graphite lattice as shown in gure 29. (In describing the ordering of molecules, the term `site order refers to the centre of mass of the molecule.) Thus in the case of the Br2 intercalant, site ordering in the intercalation compound can be achieved with very little distortion from the intermolecular spacing of solid Br2 (Wycko 1964). A potential minimum is achieved when each of the bromine atoms lies over a centre of a hexagon in the graphite network. It has been suggested by Hooley (1973) that the Br2 , IBr and ICl halogens intercalate readily because their intermolecular distances (2.27, 2.49 and 2.40 A respectively) are close to the distance between the graphite hexagons, whose centres are separated by 2.46 A . Also supporting this idea are the halogens that do not readily intercalate. For example, Cl2 which is much smaller (2.02 A ) and I2 which is much larger (2.68 A), have not to date been intercalated to form a binary compound. Neither the site ordering nor the molecular alignment of bromine in the intercalation compound have been established in detail, nor has the complete structure been determined for any other molecular intercalation compound. Electron, X-ray di raction and EXAFS studies have been carried out on graphite bromine and a number of conclusions have been reached (Eeles and Turnbull 1965,

54

M. S. Dresselhaus and G. Dresselhaus

Sasa et al. 1972, Evans and Thomas 1975, Chung et al. 1977, Heald and Stern 1978). These conclusions can be summarized as follows. The intramolecular bromine bromine distance is found to di er by less than 3% in the isolated molecule, the molecular solid and the intercalation compound, indicating that the intramolecular binding is the dominant binding force in all three cases. Eels and Turnbull (1965) have further concluded that the intermolecular brominebromine distance is also very similar in solid bromine and in intercalated bromine, and that the bromine molecules in the intercalation compound are arranged in chains, but with a site ordering commensurate with the graphite layers. The bromine molecules lie down in layer planes both in the solid and in the intercalation compound, a result also consistent with an analysis of EXAFS data for graphitebromine by Heald and Stern (1978). It has been suggested that the additional structure observed in the 77 K electron di raction patterns as compared with the 300 K patterns may be due to molecular alignment e ects, though this has not been fully substantiate d (Eeles and Turnbull 1965, Chung et al. 1977). At yet higher temperatures (379 K), the complicated superlattice di raction structure associated with the bromine intercalate suddenly disappears, leaving the simple graphite di raction pattern. The high temperature phase has been identied as a disordered or melted phase on the basis of electron di raction measurements (Chung et al. 1977) and di erential scanning calorimetry (Culik and Chung 1979). Although transport measurements indicate that the intercalate layer is partially ionized, structural studies have not clearly identied the nature of the concentration of the ionized species. Raman spectra identied with the bromine intercalant show no clear evidence for a charged intercalant species (Eklund et al. 1978), which is also consistent with the observed diamagnetic susceptibility, showing no evidence for unpaired spins (Delhaes 1977). The extended X-ray absorption ne structure (EXAFS) technique has been applied by Heald and Stern (1978, 1980) to study the BrBr intramolecular distances and the BrC nearest-neighbou r distances, using bromine gas and CBr4 as reference materials. The experiments were carried out using grafoil and grafoam (an expanded form of grafoil) to provide high surface area for adsorption studies. On the basis of polarization studies, Heald and Stern showed that the Br2 molecules lie parallel to the basal plane for the case of both the adsorbed (coverage between 0.6 and 0.9 monolayers) and intercalated molecules. The BrBr distance expands from 2.283 A for the adsorbed species, increasing by 0.03 A to in the free molecule to 2.31 A accommodate part of the lattice mismatch between the free bromine molecule and the 2.46 A distance between graphite hexagons. A further expansion of 0.03 A occurs for the intercalated molecule. upon intercalation, yielding a separation of 2.24 A Analysis of the rst shell lineshape and the second shell EXAFS gives evidence that the bromine atom lies above the hollow of the graphite hexagons. In addition, evidence is presented for an expanded BrBr intramolecular distance of 2.53 A which could be associated with a Br molecule or with a Br2 molecule having considerable 3 charge transfer ( f 0:6), the authors preferring the latter explanation. An estimate of f 0:16 is made for the charge transferred to the intercalated species (BrBr distance of 2.34 A) on the basis of the weakening of the BrBr bond by charge transfer. For the adsorbed species, the average BrC distance is found to be 2.9 A, as compared with 2.55 A for the intercalated species, yielding a distance of 2.12 A for the separation between the bromine molecule and the adjacent graphite bounding layers. This separation would suggest a 4.24 A sandwich thickness for graphiteBr2

Intercalation compounds of graphite

55

Figure 30. Schematic diagram of stage 1 graphiteFeCl3 showing intercalate layers with the bulk FeCl3 structure sandwiched between graphite planes. The intercalate layers are incommensurate with the graphite and the pertinent lattice constants are indicated. The octahedral coordination of the ferric ion is shown in the lower sandwich and the two inequivalent ferric ions are indicated on the upper sandwich.

compounds, which Ic measurements show to be ds 7:04 A. This apparent discrepancy remains to be resolved. As an example of a multilayered intercalant, the metal chloride FeCl3 crystallizes into a layered solid composed of Cl3 Fe2 Cl3 intercalate sandwiches that are incommensurat e with the graphite host. In the intercalation compound the intercalate layer tends to have the same sandwich structure (Rudor and Schulz 1940, Cowley and Ibers 1956, Metz and Hohlwein 1975a, b, Vangelisti and He rold 1977) as in the molecular solid (Wycko 1964). A typical molecular sandwich structure is illustrated in gure 30. A similar intercalate sandwich structure also occurs in the graphiteAlCl3 compounds (Leung et al. 1980b, 1981a), and in the graphiteCrCl3 compounds (Vangelisti and He rold 1976). Such composite intercalate layers are frequently found in intercalation compounds, other examples being graphiteMoCl5 (Syme-Johnson 1967), graphiteSbCl5 (Melin and He rold 1975), various graphite metal uoride compounds (Ebert 1976), graphiteH2 SO4 (Rudor and Ho mann 1938), and graphiteHNO3 (Rudor 1939, Nixon et al. 1966, and Fuzellier et al. 1977b). For many intercalation compounds the basic molecular group is preserved upon intercalation, though for some compounds the species in the intercalation compound di ers from the intercalant, as for example AsF and AsF 3 species are 6 found in the intercalation compound grown from AsF5 . 3.4. Phase transitions Graphite intercalation compounds exhibit many di erent ordered structures or phases, some of which were classied in section 3.1. Various types of transitions between these phases have been reported, including structural transitions associated with changes of stage, changes in interlayer ordering and changes in intralayer or in-

56

M. S. Dresselhaus and G. Dresselhaus

plane ordering, magnetic transitions, and superconducting transitions. Structural phase transitions have been induced by variation of the temperature, pressure, and in some cases by variation of the vapour pressure of the intercalant. To date, only a few of the observed phases have been completely interpreted, and correspondingl y few transitions between these phases have been carefully studied. Complete phase diagrams as a function of temperature and intercalate density have not been determined for even the simplest compounds (for example, graphitealkali metal compounds). One class of phase transitions is represented by the change in stage n n 1. Such transitions have been documented on a static basis using (00l) X-ray di ractograms, showing that changes in growth conditions can produce a transition from one distinct stage to another, as predicted for example by a growth diagram, such as is shown in gure 4 for the alkali metal compounds with K. During the stage change, the two stages are in equilibrium. Dynamically, transitions between stages n n 1 have been identied on the basis of stable regions in the intercalation isotherms (Sasa et al. 1971, Hooley 1977a) and in dilatationtime plots as in gure 7 (Falardeau et al. 1978). Thermodynamic changes associated with the staging transition n n 1 have been measured by Aronson et al. (1968) who determined both entropy and enthalpy changes for a number of staging transitions in compounds containing the donor intercalants K, Rb and Cs, as well as the acceptor H2 SO4 (Aronson et al. 1971). The kinetics and thermodynamics associated with the staging transition are discussed in section 2.3 and the staging transition itself in section 2.4. These studies show that that staging transition is rst order and exhibits hysteresis during intercalationdeintercalation cycles. Another type of phase transition relevant to c-axis ordering relates to the intercalate interlayer stacking ordering observed in low stage alkali metal compounds. A simple example of such a phase transition has been observed by Kambe et al. (1979) in C6 Li where a reversible transition is found between the low temperature phase exhibiting intercalate interlayer stacking order (for T < 220 K) as shown in gure 21 to another phase above 220 K with either random interlayer stacking order, or stacking order as suggested by Guerard and Herold (1975) for the room temperature intercalant arrangement. A similar reversible transition has also been reported on the basis of electron di raction studies by Kambe et al. (1980) for rst stage C8 Rb from the low temperature phase in which the interlayer stacking sequence is to a higher temperature phase above 300 K which lacks this stacking order. Ellenson et al. (1977) using the neutron di raction technique on C8 Rb have reported the observation of intercalate stacking order at 290 K, but intercalate stacking at 721 K. The disappearanc e of a well-dened superlattice pattern above 747 K was interpreted by Ellenson et al. in terms of the onset of a liquid-like or highly disordered intercalate arrangement. Phase transitions associated with changes in interlayer intercalate stacking order are also found in stage 2 alkali metal compounds, but in these cases the interlayer stacking transition is also accompanied by a change in the in-plane intercalate ordering, as discussed below. Transitions associated with a commensurate to incommensurate ordering relative to the graphite layer provide another class of transitions. Such transitions have been observed in connection with structural transitions of adsorbed rare gases on graphite. For example, epitaxial monolayers of solid krypton on a graphite surface p p exhibit a second-order phase transition from a commensurate p 3 3R308 structure to a denser incommensurate structure as the krypton vapour pressure is

Intercalation compounds of graphite

57

increased above a critical value (Chinn and Fain 1977, Stephens et al. 1979). Commensurateincommensurate transitions have also been observed in the stage 2 alkali metal compounds as described below, but these transitions are also reported to be accompanied by a change in intercalate interlayer stacking order. Another example of a commensurateincommensurate phase transition that may occur in intercalated graphite is the charge density wave transition. Arising from a strong electronphonon interaction, charge density waves form incommensurate superlattices whose periodicity is governed by Fermi surface wave vectors rather than reciprocal lattice vectors. Inoshita et al. (1977) have pointed out that the stage 1 Fermi surface in C8 K is theoretically favourable for the occurrence of charge density waves. From the experimental side, several tentative identications of observed phenomena with charge density wave transitions have been made but in no case has the evidence been compelling. Certain phase transitions in graphiteBr2 and in graphiteCs compounds observed by electron di raction by Chung (1977) have been tentatively identied with charge density waves, though no clear connection of the transition wih the Fermi wave vector has been demonstrated. More recently Clarke et al. (1979b, 1980) have identied certain sideband X-ray reections with a static distortion wave (SDW), associated with the lattice strain introduced by an incommensurat e intercalant. They have, however, been careful to distinguish their mechanism for superlattice formation from a Fermi surface instability which gives rise to charge density waves (CDW). In addition, certain DHVA frequencies observed in graphiteBr2 have been attributed by Rosenman et al. (1979) to CDW-induced orbits. Phase transitions involving changes of in-plane ordering provide a second main category of phase transitions in graphite intercalation compounds, and such phase transitions have been both considered theoretically and observed experimentally. The theoretical work of Ostlund and Berker (1980) for interacting adsorbed rare gas overlays on graphite surfaces show phase diagrams as a function of adsorbate concentration with many in-plane structures commonly found in intercalated graphite. These phase diagrams indicate that coexisting phases occur at most temperatures and adsorbate concentrations. In the intercalation compounds, phase transitions from one in-plane intercalate arrangement to another have received considerable attention recently in the donor compounds, particularly the series of phase transitions occurring in the stage 2 alkali metal compounds with K, Rb and Cs (Parry and Nixon 1967, Parry et al. 1979, Hastings et al. 1979, Clarke et al. 1979a, b, Zabel et al. 1979, Kambe et al. 1980, Berker et al. 1980). In addition to the structural changes discussed in these references, in-plane intercalate density X-ray measurements have been made by Leung et al. (1979), which when combined with the suggested superlattice structure, imply the stoichiometry of the intercalation compound. Since the adsorbed phases on graphite are very sensitive to the adsorbate concentration (Lander and Morrison 1967), it seems likely that the same is true of the intercalate layers. In the intercalation compounds both the in-plane order and the interplanar order (stacking order) are expected to be sensitive to the in-plane intercalate density. Some of the di erences in structure reported by the various workers may be related to the fact that though the compounds are of the same stage, the in-plane densities may di er. For example, stage 1 graphiteK has been observed with an in-plane p p p 3 3R308 structure which corresponds to C6 K stoichiometry (Evans and Thomas 1975), the usual p2 2R08 structure corresponding to C8 K (Rudor and

58

M. S. Dresselhaus and G. Dresselhaus

Figure 31. Measurements by Onn et al. (1977) of resistivity versus temperature for stage 2 graphiteK at zero magnetic eld (open circles) and for H 1:5 tesla (closed circles). The transition temperatures of Tl 95 K and Tu 124 K are indicated. Similar resistance anomalies are observed for stage 2 compounds with the intercalants Rb and Cs, at transition temperatures indicated in table 11.

Schulze 1954), and a high temperature presumably mixed and/or incommensurate C10 K purple phase (Aronson et al. 1968). Because of the variety of in-plane phase transitions that are found, the stage 2 alkali metal compounds have in fact provided an important means to study the physics of phase transitions. An overview of these transitions is in some sense provided by the temperature dependent resistivity measurements by Onn et al. (1977, 1979) on stage 2 compounds with K, Rb and Cs (see gure 31), revealing similar resistivity anomalies for all three intercalants, and dening the lower and upper transition temperatures Tl and Tu respectively for each compound. Values for Tl and Tu obtained by a number of workers is given in table 11. Several structural studies of these phase transitions have also been carried out and in most cases the reported transition temperatures are in good agreement with those obtained from the resistivity anomalies. It is thus meaningful to discuss the structure of those compounds in terms of three regions of behaviour: (1) T < Tl ; (2) Tl < T < Tu ; and (3) T > Tu . Although signicant structural di erences occur from one intercalant to another, some common behaviour patterns are found. At low temperatures (T < Tl ), there is interlayer intercalate stacking order, while at high temperatures (T > Tu ), there is no such stacking order (Parry and Nixon 1967, Parry et al. 1979). A high degree of in-plane order accompanies the presence of long-range interlayer intercalate ordering, and correspondigly less in-plane order is found above Tu where the interlayer ordering disappears. The low temperature behaviour for second stage graphiteK di ers in detail from that observed with the intercalants Rb and Cs. For T < Tl , the in-plane ordering for stage 2 graphiteK has been tentatively identied by Parry and Nixon (1967) and

Intercalation compounds of graphite

59

Table 11. Upper and lower transition temperatures for stage 2 alkali metal compounds. Material K Tl (K) 95 98 93 86 106 106 110 180 163 165 178e Tu (K) 124.0 126.0 122.9 130.0 172.0 162.0 165.0 170.0 230.0 228.0 225.0 Techniques Resistivity X-ray X-ray X-ray Electron di ractiona Resistivity Resistivity b Neutron di raction Electron di ractionc Resistivity X-ray X-rayd Electron di ractionf Authors Onn et al. (1977) Parry and Nixon (1977) Hastings et al. (1979) Zabel et al. (1979) Kambe et al. (1981) Onn et al. (1977) Suematsu et al. (1980a) Suzuki et al. (1980) Kambe et al. (1980a) Onn et al. (1977) Parry et al. (1969) Clarke et al. (1979a, b, 1980) Kambe et al. (1981)

Rb

Cs

a b

An additional transition was observed at 700 K. An additional transition was reported in the range 280300 K. A second sample gave four resistivity anomalies at 60, 110, 163 and 202 K. c An additional transition was observed at 620 K (see text). d Additional transitions were observed at 50 and 608 K (see text). e On heating. f An additional transiton was observed at 625 K.

p p Parry (1977) as the honeycomb h 12 12R308 structure (see gure 22) with AABBCCA 0 AB 0 BC 0 C intercalate ordering in the c-direction. Recent electron di raction studies on stage 2 graphiteK by Berker et al. (1980) have indicated that this honeycomb structure is not the dominant phase. Moreover, the presence of such long-range c-axis ordering is somewhat inconsistent with the work of Hastings et al. (1979) who inferred from analysis of the lineshape of their measured X-ray intensity proles that the c-axis ordering below Tu is interrupted by the presence of stacking faults in the graphite layers. They estimate that in stage 2 graphiteK based on an HOPG host material, stacking faults occur with an average interval of 22 A, estimate for stacking faults in stage which is within a factor of 2 of the 40 A graphiteCs, made by Clarke et al. (1979b) and based on a similar analysis. In this connection it is of interest to note that for the high stage alkali metal compounds (for example, n 8) the repeat distance Ic exceeds the mean stacking fault distance. For Tl < T < Tu , the electron di raction results on stage 2 graphiteK of Berker et al. (1980) show a powder pattern indicative of an incommensurate in-plane structure and with di raction q vectors in good agreement with those observed in the X-ray studies of Hastings et al. (1979) and Zabel et al. (1979). As T is increased toward Tu , the intensities of certain X-ray lines decrease and eventually approach zero at Tu . A study of the critical behaviour of the intensity of the 1.70 A1 X-ray line, the dominant one for T < Tu , shows a decrease in intensity (see gure 32) with little change in linewidth, which was interpreted as providing evidence for a secondorder phase transition at Tu . From the detailed analysis of the intensity measurements shown in gure 32, Zabel et al. (1979) have been able to deduce a power law dependence of the intensity I aTu T =Tu 2 , with a critical exponent 0:18 0:01, and a transition temperature Tu 122:9 0:5 K.

60

M. S. Dresselhaus and G. Dresselhaus

Figure 32. Scan of X-ray scattering prole by Zabel et al. (1979) in the l direction for qhk 1:7 A1 taken below Tu at 98.4 K (lled circles) and above Tu at 125 K (crosses), where Tu 122:9 0:5 K. The dashed line measures the background scattering, well away from qhk 1:7 A1 . From the linewidth of the X-ray line, an is obtained for the c-axis interlayer correlation distance for estimate of 30 A intercalant stacking.

Figure 33. Intensity of the superlattice qhk 1:70 A1 line at ql 0 versus reduced temperature Tu T=Tu . A least square t of the data to the relation I 2 aTu T=Tu yields 0:18 0:01 and Tu 122:9 0:5 K (from Zabel et al. 1979).

Also in the temperature interval Tl < T < Tu , the range of the intercalate interlayer c-axis stacking order is greatly reduced (Parry and Nixon 1967, Hastings et al. 1979, Zabel et al. 1979), but a short-range correlation remains. The extent of the correlation near Tl can be roughly estimated as 30 A from the linewidth of the 1 X-ray line shown in gure 32 (Zabel et al. 1979). Above T , the X-ray 1.70 A u results of gure 33 show that c-axis interlayer intercalate ordering vanishes. Though the extent of the c-axis ordering below Tu has not been quantitatively established, it is generally agreed that the c-axis ordering vanishes above Tu . With regard to the in-plane arrangement for T > Tu , the observed electron di raction spot patterns by Berker et al. (1980) imply coexisting commensurate,

Intercalation compounds of graphite

61

(a)

(b

Figure 34. Electron di raction pattern observed by Kambe et al. (1980) from stage 2 graphite at T 110 K. (a) Photograph of the di raction pattern observed with a transmission electron microscope. (b) Schematic representation of the di raction p p pattern of (a) showing the p 7 7R19:18 superlattice (dark spots) associated with the intercalant. The graphite layers give rise to the spot di raction pattern p1 1 indicated by the larger open circles.

incommensurat e and disordered phases, while the X-ray studies have been interpreted in terms of a disordered phase. Since a number of distinct phases are known to exist in various temperature ranges, a tentative phase diagram has been proposed by Berker et al. (1980) for graphitepotassium which exhibits temperature regions where more than one phase can coexist. In an attempt to explain their observed X-ray intensity distribution data for T > Tu , Zabel et al. (1980) considered the possibility of a lattice gas phase, where the intercalant exhibits a random occupation of sites between the centres of hexagons on graphite bounding layers in (AjA) conformation. Although the lattice gas model was not successful in explaining the observed di raction data, it is believed that lattice gas phases should exist in some range of temperature in graphite intercalation compounds. Bak and Domany (1979) have discussed theoretically the transition from a high temperature ordered phase to a lattice gas using the Landau theory for phase transitions and the symmetry properties of the order parameters. They nd this type of transition to be rst order, and have speculated that for the rst stage compounds with Rb and Cs the observed transitions near 750 K (Ellenson et al. 1977) and 550 K (Clarke et al. 1979a) respectively are transitions to a lattice gas from a phase with interlayer intercalate stacking order for Rb and stacking order for Cs. For the case of Rb, electron di raction studies for T < Tl p have p shown the superlattice pattern of gure 34, consistent with an in-plane p 7 7R19:18 structure that is commensurate with the graphite layers (Kambe et al. 1980). The observation of a series of di raction rings provides evidence for an incommensurate phase in the temperature range Tl < T < Tu . In the temperature range T < Tu , Parry et al. (1969) have reported an AABBCC intercalate interlayer stacking order, though the precise meaning of the intercalate stacking sequence has not been specied. Above Tu , electron patterns indicate coexisting in-plane orientationally

62

M. S. Dresselhaus and G. Dresselhaus

locked and unlocked incommensurate phases, while X-ray studies by Parry indicate no c-axis intercalate stacking order above Tu . The work of Kambe et al. (1980) shows that the phase transition at Tu is also accompanied by a dramatic change in the real space image of the sample. At very high temperatures (T 0 620 K), a very complicated electron di raction spot pattern has been reported, but has not yet been explained (Kambe et al. 1980). For the case of Cs, single crystal X-ray di raction measurements give evidence for p2 2 and p3 3 ordered structures below 50 K, while for 50 K < T < Tl , a ring pattern, modulated by 12 sidebands, was reported and interpreted in terms of an in-plane incommensurate ordering (Clarke et al. 1979b, 1980), though c-axis intercalate interplanar AABBCC ordering and commensurate in-plane ordering were also reported (Parry et al. 1979, Parry 1977) for this temperature range. Above Tl , short-range c-axis correlation and long-range orientational order were reported by Clarke et al. (1980). For the Cs compound Clarke et al. also found a second-order transition at Tu similar to that shown for potassium in gure 32, but with 0:23 0:04 (where is dened in gure 32). An intercalate melting temperature of T 608 K was also reported by Clarke et al. (1979a). The transition temperatures associated with structural transitions can also be obtained by studying anomalies in transport and other properties of intercalated graphite. Phase changes are often accompanied by changes in scattering mechanisms which give rise to changes in transport properties. For the case of the alkali metal donor compounds, Onn et al. (1977, 1979) observed resistivity anomalies at Tl and Tu which yielded transition temperatures in good agreement with values obtained from structural studies. A summary of their results for donor compounds with stages n 2, 3, 4 is given in tables 11 and 12. For n 2 the stage dependence of the transition temperatures is weak, in general. These authors, however, found no resistivity anomalies for stage 1 alkali metal compounds, where the phase transitions, as observed by electron di raction, involve only a change in interlayer intercalate stacking order and no change in the close-packed in-plane arrangement. At low temperatures (below 1 K) transitions to a superconducting state have been observed for stage 1 compounds with K, Rb and Cs (Hannay et al. 1965). Detailed studies of the superconducting transition in C8 K by Koike et al. (1980a, b) show evidence for type I and type II superconductivity , depending on the direction of the applied magnetic eld relative to the c-axis. The anisotropic properties of the superconducting transition are discussed in section 4.8. Phase transitions in molecular acceptor compounds have also been studied. In most cases, phase transitions in acceptor compounds give rise to larger changes in the pertinent observables than do donor compounds. Also for the case of acceptors, phase transitions have been detected by more diverse techniques than for the donor compounds. On the other hand, di raction studies on donor compounds have yielded more detailed information than for the acceptors. The acceptor system that has received more attention in this connection is the graphiteHNO3 system. By measuring the intensity of (hkl) X-ray reections Nixon et al. (1966) were able to reach a number of conclusions concerning the unit cell dimensions, the appropriate space group, the in-plane intercalate ordering, the stacking order of the graphite layers (same as shown in gure 2), and the interlayer intercalate correlation. In the low temperature phase (T < 252 K), they found the basic intercalate in-plane structure to be commensurate with the graphite layers and to exhibit some interlayer intercalate correlation in the c-direction. Above 252 K,

Intercalation compounds of graphite


Table 12. Phase transition reported for various donor and acceptor compounds.a Transition temperatures (K) 87, 112 95, 250 92, 250 200, 230 220 190, 230 253 250 249 250 30, 252 30, 250 6 252 5 253 232 250 250 1 254 1 252.0 0.2, 261.5 0.2 379 10 266, 277, 297, 375 2 333 10 316 10 309, 312, 5 310.5, 314 314 2 175 180 200 21, 165, 225

63

Intercalant K K K K Rb Cs HNO3

Stage 3 3 4 5 4 5 1, 2, 3 2, 3, 4 1, 2 1 Residue Residue Residue 3 Residue 2 Dilute Dilute

References Suematsu et al. (1980a) Onn et al. (1979) Onn et al. (1979) McRae et al. (1980) McRae et al. (1980) McRae et al. (1980) Bottomley et al. (1964) Nixon et al. (1966) Ubbelohde (1968a, b) Ubbelohde (1968b) Kawamura et al. (1975) Kawamura et al. (1975) Kawamura et al. (1975) Avogadro and Villa (1977) Inagaki et al. (1977) Khanna et al. (1978) Dworkin and Ubbelohde (1978) Dworkin and Ubbelohde (1978) Chung et al. (1977) Culik and Chung (1979) Chung et al. (1977) Chung et al. (1977) Tashiro et al. (1978) Tashiro et al. (1978) Culik and Chung (1979) Weinberger et al. (1978a) Khanna et al. (1978) Zeller et al. (1979) Zeller et al. (1979)

Method Resistivity Resistivity Resistivity Resistivity Resistivity Resistivity Thermal expansion X-ray Thermopower Resistivity Resistivity Hall e ect Magnetoresistance Nuclear magnetic resonance Specic heat Electron spin resonance Di erential scanning calorimetry (exothermic) (endothermic) Electron di raction Di erential scanning calorimetry Electron di raction Electron di raction Specic heat Specic heat Di erential scanning calorimetry Nuclear magnetic resonance Electron spin resonance Resistivity Resistivity

Br2 IBr ICl

Residue Residue Residue Residue 2 5 2 1 1 1 2

AsF5

Phase transition temperatures for the stage 2 alkali metal compounds with K, Rb and Cs are given in table 11.

both the interlayer correlation and the in-plane ordering were found to disappear. For T > 252 K, the HNO3 intercalant was described as liquid-like (Nixon et al. 1966), and the phase transformatio n was classied as an orderdisorder transformation. The transition to a liquid-like state is also consistent with motional narrowing e ects observed by Avogadro et al. (1977) and by Avogadro and Villa (1977) in the proton N.M.R. line, giving rise to sharp, discontinuous jumps in the spinlattice relaxation time T1 , and in the spinspin relaxation time T2 . Closely related to these discontinuous changes in T1 and T2 , is the large increase in the linewidth of the E.S.R. line observed by Khanna et al. (1978) as T was lowered below 250 K. Also associated with this orderdisorder transition in graphiteHNO3 is an anomaly in the thermal expansion, reported by Bottomley et al. (1964) and corresponding to a discontinuous decrease in c-axis lattice constant of 0.044, 0.048

64

M. S. Dresselhaus and G. Dresselhaus

and 0.052 A for stages n 1, 2, 3 respectively, as the sample was cooled through the transition temperature, which showed little stage dependence. The larger c-axis spacing above the transition temperature might be necessary to stabilize the large molecular intercalant in the liquid zone. An anomalous peak in the thermoelectric power at the transition temperature was found by Ubbelohde (1968b) and attributed to additional scattering of the hole carriers. In addition, large anomalies were later reported near 250 K in the electrical resistivity, magnetoresistanc e and Hall e ect (Kawamura et al. 1975). Using the di erential scanning calorimetry technique, Dworkin and Ubbelohde (1978) were able to provide additional information about the thermodynamics of this orderdisorder transition. They showed that there are in fact two distinct anomalies closely spaced in temperature, the exothermic peaks occurring at 250 1 K and 254 1 K and the endothermic peaks at T1 252:0 0:2 K and T2 261:5 0:2 K, both peaks being associated with rstorder transitions. Their analysis yielded values for the change in enthalpy and entropy of H 1 370 3 cal/mol, H2 46 1 cal/mol, S1 1:43 0:01 cal/ mol K, and S2 0:16 0:01 cal/mol K, where it was noted that S1 R ln 2, to within experimental error. Low temperature resistivity and Hall e ect measurements on graphiteHNO 3 compounds by Kawamura et al. (1975) have identied an additional sharp anomaly slightly below 30 K, which they attributed to a structural phase transition between two ordered phases. The explicit identication of the structure of the low temperature phase awaits further di raction studies. For the graphiteAsF5 system, a broad phase transition is observed in the vicinity of 200 K for stage 1 and 220 K for stage 2 using susceptibility (Weinberger et al. 1978a), electron spin resonance (Khanna et al. 1978) and resistivity (Zeller et al. 1979) measurements. A sharp phase transition was also observed in the graphite AsF 5 system at low temperature (21 K) as a resistivity anomaly by Zeller et al. (1979). A phase transition has also been reported by McRae et al. (1980a) in the graphiteH2 SO4 system at 199 K through observation of an anomaly in the resistivity. The graphitehalogen compounds represent another group of compounds where phase transformation s have been identied. The disappearance of the superlattice structure in the electron di raction patterns by Chung et al. (1977) was interpreted in terms of an orderdisorder transformation , yielding transformation temperatures for the intercalants Br2 , IBr and ICl of 379, 333 and 316 K respectively. These phase transformation s have been established as rst order through thermal measurements. Specically, di erential scanning calorimetry studies by Culik and Chung (1979), yielded a transition temperature of 375 2 K for graphiteBr2 and 314 5 K for graphiteICl in good agreement with the electron di raction results. Similar results were obtained for the graphiteICl system by Tashiro et al. (1978) using specic heat measurements, yielding two closely spaced specic heat packs for C16 ICl at 309 and 312.5 K and for C40 ICl at 310.5 and 314.0 K. The enthalpies associated with these transitions were found by Culik and Chung (1979) to be 108 15 cal/mol in graphiteBr2 , and 460 50 cal/mol in graphiteICl. An additional rst-order transition has been observed in graphiteBr2 by Culik and Chung at 277 2 K, which they tentatively identied with an orderdisorder transition associated with an intralayer intercalate molecular orientational ordering. This identication has not yet been conrmed by direct structural studies. Magnetic phase transitions have been observed in stage 1 and stage 2 graphite FeCl3 and graphiteFeCl2 using both magnetic susceptibility measurements (Kar-

Intercalation compounds of graphite

65

imov et al. 1972) and Mossbauer spectroscopy (Ohhashi and Tsujikawa 1974, Hohlwein et al. 1974). Magnetic order has also been reported in stage 2 NiCl2 and in compounds of CoCl2 with various stages (Karimov 1973, 1974, 1975). In the magnetic susceptibility study of FeCl3 , a magnetic balance was employed using the Faraday method, and magnetic measurements were made in the range 1:5 < T < 300 K. In the Mossbauer study, the magnetic environments of Fe2 and 3 ions were probed through measurement of hyperne elds. A transition to an Fe antiferromagneticall y ordered state was reported in stage 1 graphiteFeCl3 with a Neel temperature of TN 3:8 K, while in stage 2 graphiteFeCl3 and in stages n 1, 2 graphiteFeCl2 the transitions are reported to a ferromagnetically ordered state with Curie temperatures of Tc 8:5, 15.5 and 14.5 K respectively. For both the antiferromagneti c and ferromagneticall y ordered states, the magnetic moments associated with a single magnetic layer are aligned ferromagnetically within that layer. The interactions between successive magnetic layers are very small, except for the case of the stage 1 compounds, where for graphiteFeCl3 the interlayer interaction is antiferromagneti c and gives rise to an opposite spin alignment on successive magnetic layers. An Ising model has been used to explain the magnetic properties of this prototype two-dimensional magnetic system. Both the magnetic susceptibility of Mossbauer studies show that the magnetic properties of the intercalation compounds are closely related to the behaviour of the magnetic Fe2 and Fe3 ions in pristine FeCl2 and FeCl3 . For example, the temperature dependence of the internal magnetic elds for stage 1 and stage 2 graphiteFeCl3 is well described by a mean eld model for spin S 5=2, an internal magnetic eld at T 0 of 50 1 T and a magnetic moment associated with the Fe3 ions of 4:5 0:2 , all values in good ageement with well-established results for Fe3 in an octahedral chlorine environment. The NiCl2 and CoCl2 intercalants also show at low temperatures a ferromagnetic in-plane spin alignment. Two distinct magnetic orderings at Tc1 and Tc2 have been reported. The transition from the paramagnetic high temperature state to a magnetic state showing no long-range magnetic correlations occurs at Tc1 , which is 20.3 K for NiCl2 (stage 2) and 9.05 K for CoCl2 . At Tc1 the magnetic susceptibility undergoes a discontinuity, abruptly becoming innite. The specic heat also shows a maximum at Tc1 . At the lower temperature Tc2 , given by 18.1 K for NiCl2 and 8.1 K for CoCl2 , long-range magnetic order sets in, though no singularity in the specic heat is observed. The relation between these ordering transitions, theoretical models and critical exponents for low dimensional magnetic systems is discussed by Karimov (1975). Thus graphite intercalation compounds provide an interesting opportunit y to study the properties of `two-dimensional magnetic systems.

4. Electronic properties 4.1. Electronic structure of graphite and graphite intercalation compounds 4.1.1. Introduction A major simplication is made in treating the electronic structure of graphite intercalation compounds by recognizing the strong similarity of the structural and electronic properties of graphite intercalation compounds to those of their parent constitutive materials, the graphite host and the intercalant. The physical basis for this identication arises from the strong intralayer bonding in both graphitic and

66

M. S. Dresselhaus and G. Dresselhaus

intercalate layers, and the relatively weak interlayer bonding between graphite intercalate and graphitegraphite layers in the intercalation compounds. A close relationship between the graphite intercalation compounds and their parent materials implies that in the dilute intercalate concentration limit, the electronic structure is closely related to that of pristine graphite. The discussion of the electronic structure for the graphite intercalation compounds therefore begins with a review of the graphite electronic structure and its application in dilute limit models for the intercalation compounds. The e ect of zone folding connected with the superlattice periodicity introduced by the intercalant is then considered. This is followed by a discussion of rst principles band structure calculations for the rst stage alkali metal compounds C6 Li and C8 K, and a discussion of phenomenological models for higher stage compounds. 4.1.2. Electronic structure of graphite Because of the large anisotropy of the crystal structure, most models for the electronic structure start from a two-dimensional approximation, treating the intraplanar interaction between the 2s, 2px , 2py atomic orbitals to form the strongly coupled bonding and antibonding trigonal orbitals. These trigonal orbitals give rise to three bonding and three antibonding -bands in the two-dimensional graphite band structure. In these models, the weakly coupled pz atomic wavefunctions give rise to two -bands, which are degenerate by symmetry at the six Brillouin zone corners at point P, through which the Fermi level passes. The points P, Q and of the two-dimensional zone correspond respectively to HKH, LML and AA of the three-dimensional zone shown in gure 35. A large number of calculations of the two-dimensional graphite electronic structure have been made (Wallace 1947, Coulson 1947, Corbato 1956, 1959, Bassani and Parravicini 1967, Doni and Parravicini 1969, Painter and Ellis 1970). The dispersion relations for a twodimensional band structure are shown in gure 36, where the calculations of Painter and Ellis (1970), Zunger (1978) and Bassani and Parravicini (1967) are compared. The degenerate -bands at point P are of particular interest since the Fermi level goes through this P-point degeneracy, and the Ek relation away from the P-point is linear in k for the -bands. We further note that at the P-point the three antibonding -bands lie far above the -bands in energy, and the three bonding -bands lie far below. Three-dimensional band models conrm that (1) the Fermi surface is located near the Brillouin zone edges HKH and H 0K 0 H 0 (shown in gure 35), and (2) the width of the -bands in the vicinity of the Brillouin zone edges is much less than the

Figure 35. Graphite Brillouin zone showing several high symmetry points and a schematic version of the graphite electron and hole Fermi surfaces located along the HK axes.

Intercalation compounds of graphite

67

separation between the -bands and the bonding and antibonding -bands. Therefore two-dimensional band models have been extensively applied to the qualitative interpretation of much experimental data on graphite. Graphite, however, is a three-dimensional solid with AB stacking of the graphite layers, giving rise to four carbon atoms per unit cell as shown in gure 14. Although the interlayer interaction is small, it has a profound e ect on the four -bands near the Brillouin zone edges, causing a band overlap that is responsible for the semimetallic properties of graphite, whereas the two-dimensional model gives a zero gap semiconductor for graphite. Detailed models for the dispersion relations for the four -bands have been developed by Slonczewski and Weiss (1958) and by McClure (1957, 1960). This model, known as the SlonczewskiWeissMcClure (SWMcC) band model, has been applied extensively to explain the transport, quantum oscillatory, optical and magneto-optical properties dependent on the electronic structure near the Fermi level. An excellent summary of the electronic properties has been given by Spain (1973) and aspects relevant to the properties of graphite intercalation compounds are briey reviewed in this section. Whereas the two-dimensional band models give the in-plane dispersion relations for the two -bands for a planar graphite unit cell, the three-dimensional SWMcC model gives three-dimensional dispersion relations for the four -bands corresponding to the full graphite unit cell containing four crystallographicall y distinct atoms based on the ABAB stacking sequence (gure 14). The e ect of di erent stacking arrangements of the graphite layers (AAA, ABAB, ABCABC) on the electronic structure has been considered by McClure (1969) and by Samuelson et al. (1980), who showed that each stacking arrangement results in somewhat di erent energy levels near EF . It should be emphasized that the mathematical form of the threedimensonal SWMcC model is general and is based on the space group symmetry of graphite. The basic SWMcC model has also been extended to include explicitly the spinorbit interaction (Dresselhaus and Dresselhaus 1965). The SWMcC model gives a phenomenological treatment of the electronic structure based on crystal symmetry. The most general form of the hamiltonian consistent with crystal symmetry is developed for k values in the vicinity of the Brillouin zone edges. In the kz direction, a Fourier expansion is made, and rapid convergence is obtained because of the weak interplanar binding. In the layer planes, a k p expansion is made since the extent of the Fermi surface in the basal planes is small compared with Brillouin zone dimensions. The SWMcC model is commonly written in terms of the (4 4) hamiltonian for the -bands, 0 1 E1 0 H13 H13 B 0 E2 H23 H23 C C B 4:1 HB C @ H13 H23 E3 H33 A H13 H23
H33

E3

where the band edge energies are given by

E 1 1 1 5 2 ; 2 E 2 1 1 5 2 ; 2 E 3 1 2 2 2 and the interaction terms are

4:2 4:3 4:4

68

M. S. Dresselhaus and G. Dresselhaus H13 21=2 0 4 exp i; H23 21=2 0 4 exp i; H33 3 exp i; 4:5 4:6 4:7 4:8 4:9 4:10

in which is the angle between k and the K direction, and 2 cos kz c0 = 1 2 p 3a0 ; and the dimensionless wave vectors along kz and in the basal plane are given by and

where c0 3:35 A and a0 2:46 A and is the in-plane wave vector measured from the Brillouin zone edges. Each of the seven parameters (0 ; . . . ; 5 ; ) of the SWMcC model can be identied with overlap and transfer integrals within the framework of the tight binding approximation, but in practice they are evaluated experimentally. The tight binding identication of the seven SWMcC band parameters together with their numerical values are summarized in table 13. The eigenvalues of the SWMcC hamiltonian (equation (4.1)) yield the energy dispersion relations which are schematically illustrated in gure 37. Along the Brillouin zone edge HKH, two of the four solutions are doubly degenerate and are labelled by E3 . The remaining two solutions are non-degenerate and are denoted in this gure by E1 and E2 . The degeneracy of the two E3 levels is lifted as we move away from the zone edge, and this is indicated on the left-hand side of the gure with reference to the plane dened by 0. At the H point ( 1=2), the levels E1 and E2 are degenerate, and the double degeneracy of these levels is maintained throughout the planes 1=2, as shown on the right-hand side of the gure. The SWMcC hamiltonian has simple solutions in certain special cases. If 3 is neglected, the four solutions E 1 E1 E3 1 E1 E3 2 0 4 2 2 1=2 ; 2 4 E 1 E2 E3 1 E2 E3 2 0 4 2 2 1=2 2 4 4:11 4:12

are obtained. These solutions have been applied to the interpretation of a large variety of experiments relevant to the electronic structure and Fermi surface of graphite (McClure 1971, Spain 1973). It is of interest to note that whereas twodimensional graphite is a zero-gap semiconductor, three-dimensional graphite is semimetallic with a band overlap of 22 (0.040 eV) and a bandwidth along the Brillouin zone edge of 41 (1.56 eV). Since the Fermi surface of the intercalation compounds occupies a much larger volume of the Brillouin zone than for pristine graphite, the extension by Johnson and Dresselhaus (1973) of the SWMcC model to yield dispersion relations for the bands throughout the Brillouin zone is of particular interest. Using symmetry requirements to specify the form of the hamiltonian, a three-dimensional Fourier expansion was used for the basis functions. The band parameters of the model were evaluated using (1) Fermi surface data in the vicinity of the Brillouin zone edges HKH and H 0 K 0 H 0 , (2) ts to the optical data below 6 eV, and (3) the requirement that the dispersion relations reduce to those of the SWMcC model in the vicinity of

Intercalation compounds of graphite

69

Table 13. SlonczewskiWeissMcClure band parameters for graphitetheir magnitudes and physical signicance.
Order of magnitude (eV) 3:16 0:05 0:39 0:01

Band parameter 0 1

Physical origin Overlap of neighbouring atoms in a single layer plane. Overlap of orbitals associated with carbon atoms located one above the other in adjacent layer planes. Width of -bands at point K is 41 . Interactions beween atoms in next nearest layers and from coupling between - and -bands. Band overlap is 22 . Majority de Haasvan Alphen frequencies determined by 2 . Coupling of the two E3 bands by a momentum matrix element. Trigonal warping of the Fermi surface is determined by 3 . Coupling of E3 bands to E 1 and E2 bands by a momentum matrix element. Determines inequality of K-point e ective masses in valence and conduction bands. Interactions between second nearest layer planes. Introduced in E1 and E2 to be consistent with E3 in the order of the Fourier expansion. Di erence in crystalline elds experienced by inequivalent carbon sites in layer planes. Volume of minority hole carrier pocket sensitive to . The Fermi level is measured with respect to the H-point extremum (see gure 37) and is xed by the condition that the electron density hole density.

Reference Toy et al. (1977) Misu et al. (1979)

0:020 0:002

Soule et al. (1964)

0:315 0:015 0:044 0:024

Doezema et al. (1979)

Mendez et al. (1980)

0:038 0:005

Misu et al. (1979)

0:008 0:002

Toy et al. (1977)

EF

0:024 0:002

the Brillouin zone edges. The resulting dispersion relations for the -bands along several high symmetry directions are shown in gure 38. The extension of this threedimensional Fourier expansion model to describe dispersion relations in graphite intercalation compounds (Dresselhaus and Leung 1980) is described in section 4.1.4. First principles three-dimensional calculations of the electronic band structure of graphite have also been carried out using a variety of techniques (Van Haeringen and Junginger 1969, Nagayoshi et al. 1973, 1976, Zunger 1978), and yielding good agreement with the widely used phenomenological SWMcC model. These rst principles calculations have more recently been extended to consider the e ect of pressure on the graphite electronic structure (Nagayosh i 1977), and to treat low stage graphite intercalation compounds (Inoshita et al. 1977, Holzwarth et al. 1977, 1978a, b, Ohno et al. 1979) as discussed in section 4.1.4. A dilute limit model for graphite intercalation compounds based on rigid SWMcC bands for the graphite -bands was originally proposed by McClure (1960) for application to graphitebromine compounds. According to this model,

70 M. S. Dresselhaus and G. Dresselhaus


Figure 36. Band structure calculations for two-dimensional graphite along the PQ directions of the two-dimensional zone, in which P corresponds to the HK axis and the Q corresponds to the LM axis of the three-dimensional zone. The left panel (a) is from the work of Painter and Ellis (1970), the centre panel (b) is from Zunger (1978) and the right panel (c) from Bassani and Pastori Parravicini (1967). The solid curves are for the -bands and the dashed curves for the -bands. The Fermi level passes through the degenerate P levels, giving rise to a dispersion relation 3 linear in for the P levels. 3

Intercalation compounds of graphite

71

Figure 37. Electronic energy bands near the HK axis in three-dimensional graphite as obtained from the SWMcC band model. Along the HK axis (centre gure) the E3 band is doubly degenerate. This degeneracy is lifted as we move away from the axis, as is illustrated in the left-hand gure for k vectors ? c-axis, but in the 0 plane. For k ? c-axis and in the plane 1=2 (at the Brillouin zone hexagonal boundary), the four -bands occur in doubly degenerate pairs, as illustrated in the right-hand gure.

Figure 38. Graphite -bands along several high symmetry directions calculated by Johnson and Dresselhaus (1973) on the basis of a three-dimensional Fourier expansion of Ek. The Fourier expansion coe cients are evaluated by tting Ek to the SWMcC model along the HK axis and by comparing the calculated frequency dependent dielectric function "! to optical reectivity measurements.

the Fermi level is allowed to rise for donors to accommodate the additional electron carriers, and for the case of acceptor compounds, the Fermi level is lowered to accommodate the additional holes. Further application of this model to intercalated graphite was made by Dresselhaus et al. (1977a). The basic assumption of this model is the observation that the intercalate layer is e ectively screened by a single graphite bounding layer. Thus for the graphite interior layers in intercalation compounds

72

M. S. Dresselhaus and G. Dresselhaus

with stage index 04 or 5, the electronic structure should be basically graphitic, and therefore could be described by bands of the SWMcC form, with only minor modications to the band parameters of the model. Justication for this assumption has come from magnetoreection (see section 4.6), Shubnikovde HaasFermi surface studies (see section 4.5), Raman spectroscopy (see section 5), and optical reectivity (see section 4.3) studies of intercalated graphite. In some cases the form of the SWMcC model has been applied to treat in an approximate way the graphite bounding layers adjacent to the intercalant. The dilute limit model makes use of the simple analytic form for the energy bands given by (4.11) and (4.12). Thus the formulae for the Fermi surface cross-sectional area for electrons and holes can be obtained directly from (4.11) and (4.12). The result for electrons is given by Ae 4 E2 EF E3 EF ; 2 1 2 3a2 0 0 4:13

in which 24 =0 cos and is the dimensionless wave vector given by equation (4.9). Because of the dependence of the cross-sectional area on (or kz ) for the roughly ellipsoidal Fermi surface, the electron carrier density ne EF as a function of Fermi energy is then obtained by a kz integration of Ae given by 1 1=2 Ae d: 4:14 ne EF c0 1=2 The corresponding expression for the cross-sectional area of the hole surface is obtained from equations (4.13) and (4.14) by the substitutions E2 ! E1 and 1 2 ! 1 2 . From these expressions, the dependence of the electron and hole carrier densities for the graphite interior layers as a function of Fermi energy are obtained and the results are plotted in gure 39. When the Fermi level rises above the H-point E3 band extremum or falls below the K-point E3 band extremum, a single carrier regime results, as seen in gure 39. Di erentiation of the carrier densities with

Figure 39. Carrier density obtained from the SWMcC band model for various Fermi level shifts relative to the graphite Fermi level E8 where the electron and hole carrier F densities are equal. In the band overlap region (see gure 37) both electron and hole carriers are present.

Intercalation compounds of graphite

73

Figure 40. de Haasvan Alphen frequencies (in tesla) obtained from the SWMcC model for various Fermi level shifts. The indicated DHVA frequencies correspond primarily to extremal cross sections around the K and H points. For a small range of (EF E8 ), F extremal Fermi surface cross sections (dotted curve) are also found at non-extremal values in the range 0 < < 1=2.

respect to energy yields the density of states in the graphite interior layers as a function of Fermi level. The position of the Fermi level relative to the K-point band edge is found experimentally, as, for example, from magnetoreection measurements (see section 4.6). The dilute limit model can thus be applied to yield the contribution of the graphite interior layers to transport and other electronic properties. The dependence of the (kz 0) Fermi cross-sectional areas on Fermi level is found from (4.13) and is shown in gure 40. These cross-sectional areas have a direct application to the interpretation of de Haasvan Alphen (DHVA) frequencies that are associated with graphitic parts of the Fermi surfaces. Higuchi et al. (1980) have made use of the dilute limit model to explain their observed DHVA frequencies for stage 3 and 4 graphiteK compounds. The e ect of staging is incorporated into the dilute limit model by these authors through c-axis zone folding as discussed in section 4.1.3. Since there are no graphite interior layers for stage n 1 and n 2 compounds, the dilute limit model is not directly applicable in these cases. Even for stage 3 compounds, where there is a single graphite interior layer, signicant perturbations to the SWMcC model are expected. Band parameters associated with the graphite interior layers are determined through the magnetoreection experiment discussed in section 4.6. A version of this dilute limit model has been developed by Blinowski et al. (1980) and applied to the interpretation of the optical properties of graphite acceptor compounds for stages n 1 and n 2. For the stage 1 acceptors, a two-dimensional version of (4.11) and (4.12) was used, obtained by neglecting all interactions other than the nearest-neighbou r in-plane overlap energy 0 . Setting all band parameters other than 0 equal to zero, (4.11) and (4.12) at 0 yield E p 3 0 a0 ; 2 4:15

74

M. S. Dresselhaus and G. Dresselhaus

which Blinowski used for the dispersion relations for the stage 1 compounds. This linear relation is typical of two-dimensional models for graphite for the energy bands near the zone edge, as seen in gure 36. The dispersion relations for the stage 2 compounds can be obtained from (4.11) and (4.12) by setting all band parameters other than 0 and 1 equal to zero, yielding four dispersion relations at 0 given by
2 2 E 1 1 3 0 a2 2 1=2 ; 0 4 2 2 E 1 1 3 0 a2 2 1=2 : 0 4

4:16 4:17

The Fermi surface for this model is a cylinder so that the volume of the carrier pockets is found from (4.13) neglecting all band parameters other than 0 for the stage 1 compounds and neglecting all band parameters other than 0 and 1 for stage 2 compounds. Dispersion relations in the form of (4.15)(4.17) were then applied by Blinowski et al. (1980) to interpret optical reectivity measurements in several stage 1 and stage 2 acceptor compounds. From analysis of the reectivity data, values for EF , and hence the fractional charge transferred to the graphite bounding layers, were deduced on the basis of values of the band parameters 0 2:4 eV and 1 0:188 eV. These values for 0 and 1 are to be contrasted with their values in pristine graphite (0 3:16 eV, 1 0:39 eV) are given in table 13. The large departure of 0 and 1 , as reported by Blinkowski et al., for the stage 1 and stage 2 intercalation compounds implies large changes in the in-plane (0 ) and c-axis (1 ) interactions. It remains to be seen whether the dispersion relations with these values of 0 and 1 can explain the various transport, Fermi surface and electron spectroscopy studies that have been carried out on low stage acceptor compounds. The application of the model of Blinowski et al. to the interpretation of optical reectivity data is further discussed in section 4.3. 4.1.3. Zone-folding e ects Because of the superlattice structure in the c-direction for well-staged materials and the additional in-plane superlattice in the intercalate layers, zone-folding techniques are used to relate both the electron and phonon dispersion relations of the intercalation compounds to those of the graphite host. Zone folding is applied both to phenomenological and rst principles calculations. The zone-folding technique is commonly used for good metals where the free-electron dispersion relation Ek 2 k2 =2m is folded into a Brillouin zone appropriate to the crystal symmetry h of the metal. The e ect of a periodic potential is then treated in perturbation theory in the context of the zone-folded representation (reduced zone scheme). In the case of intercalation compounds, the high symmetry of pristine graphite is used as an approximate symmetry for intercalated graphite, and the disturbance to the graphite periodicity by the intercalant is treated as a perturbation. Some simple examples of in-plane and c-axis zone folding for intercalation compounds are given below. In-plane zone folding is readily demonstrated in terms of the p2 2R08 inplane superlattice that occurs for the stage 1 alkali metal compounds C8 K, C8 Rb and C8 Cs and illustrated in gure 20. In this case, the planar unit cell for the intercalation compound is four times larger in area than the graphite unit cell, so that the corresponding planar unit cell for the intercalation compound in reciprocal space (dashes) is one-fourth as large as that for graphite, as shown in gure 41. Zone

Intercalation compounds of graphite

75

Figure 41. Section through p graphite Brillouin zone (kz 0) illustrating the zone folding the p associated with a p 3 3R308 superlattice (dotted lines) and a p2 2 superlattice (dashed lines).

folding allows mapping of the larger graphite Brillouin zone on to the smaller zone for the intercalation compound, and therefore an identication of the electronic dispersion relations for graphite is made using the reduced zone scheme of the intercalation compound. More specically for the case of the 2 2R08 superlattice, the mapping of the larger graphite zone brings the graphite M-point into the zone centre of the intercalation compound as shown in gure 41. Since the M- and points in the reduced zone are equivalent, the symmetry-reducin g potential associated with intercalation will produce a relatively strong interaction between M-point and -point states. With regard to lattice optical properties, zone-folding e ects are expected to turn on new Raman and I.R.-active modes, as the M-point (which does not ordinarily contribute to rst-order, allowed processes) maps into the -point; this is discussed in section 5. p p For the p 3 3R308 structure shown in gure 21, the planar Brilloun zone is one-third the size of the graphite planar zone, as indicated by the dots in gure 41. In this case it is seen that the K-point of the graphite zone maps into the -point for the intercalation compound. The physical di erences in the electronic structure of intercalated graphite arise from the changes in periodic potential resulting from intercalation. The dominant changes occur near the Brillouin zone boundaries introduced by the zone-folding procedure. In the absence of a perturbing potential, zone folding gives energy levels identical to graphite but described in a reduced zone scheme. If the perturbing potential is small, then the Fermi surface will not di er signicantly from constant energy surfaces of the SWMcC model even though the volume contained within the Fermi surface may be changed by as much as two orders of magnitude. In addition, zone folding of the K-points into the -points for p p the case of the p 3 3R308 superlattice would result in a carrier pocket about k 0 for intercalated graphite, while for the p2 2R08 superlattice, no such carrier pockets near the -point would result. For most acceptor compounds, the intercalate in-plane structure is not commensurate with the graphite honeycomb structure and no simple mapping of the graphite structure on to the reduced zone scheme of the intercalant can be made. Zone folding is also applied to the superlattice structure along the c-direction. Because of the very strong intraplanar binding and the very weak interplanar binding, the electronic energy levels show much less dispersion along the kz -direction that is observed in the kx ky -plane. Because of this weak dependence, as a rst approximation, the intercalate layer is replaced by a graphite layer and the

76

M. S. Dresselhaus and G. Dresselhaus

Figure 42. Construction of the Fermi surfaces by Tanuma et al. (1978) for a fourth-stage graphiteK compound. (a) Cross-section of the Fermi surface is shown assuming a rigid band model for graphite. (b) Fermi surface is folded back to the new Brillouin zone HKH which is introduced by the superlattice in the c-axis direction. (c) Schematic Fermi surfaces deduced from (b) by introducing energy gaps due to the periodic potential of the superlattice. (d) Fermi surfaces constructed in (c) are shown in the extended zone scheme.

di erences in interlayer spacing and potential energy are treated in perturbation theory. This approach readily provides the correct number of energy bands for a compound of given stage and form the basis for a phenomenological model for the electronic dispersion relations for a compound of arbitrary intercalant and arbitrary stage (Leung and Dresselhaus 1981b), as described in section 4.1.4. Furthermore, caxis zone-folding concepts have been applied to the interpretation of quantum oscillatory phenomena in graphiteK compounds by Tanuma et al. (1978) and Higuchi et al. (1980) , who related a graphitic Fermi surface in the full graphite Brillouin zone to a Fermi surface obtained by introducing new Brillouin zone boundaries according to the c-axis superlattice periodicity kz =Ic . Such zone folding yields new Fermi surface pockets as illustrated by gure 42 for the repeat distance Ic corresponding to a stage 4 graphiteK compound. Staging introduces a superlattice along kz which results in new Brillouin zone boundaries indicated in ~ gure 42(a) by H . Folding of the Fermi surface on to the reduced zone scheme is shown in gure 42(b) on the basis of the rigid band model, discussed in the previous section. The periodic potential for the superlattice opens up gaps at the zone boundaries which modify the Fermi surfaces, as illustrated in gure 42(c) for the reduced zone and in gure 42(d) for the extended zone. 4.1.4. Electronic band models for intercalated graphite Several rst principles band calculations have been carried out for graphite intercalation compounds, but all have so far focused on stage 1 compounds. The

Intercalation compounds of graphite

77

restriction to stage 1 is brought about by the complexity of the problem for higher stage compounds. This complexity can be appeciated from the following example. For the simplest intercalation compound, rst stage C6 Li, there are six carbon atoms and one Li atom per unit cell, assuming AA interlayer graphite (A) and intercalate () stacking order. Thus C6 Li has 39 occupied orbitals out of a total of 70 orbitals, considering all 1s2 2s2 2p6 levels for both the C and Li atoms. The unit cell p p for stage 2 graphiteLi with the same in-plane superlattice p 3 3R308, and assuming CAABBCCA c-axis interlayer stacking order, contains 36 carbon atoms and three lithium atoms, which for the same 1s2 2s2 2p6 levels gives 390 orbitals, 225 of which are occupied. Because of the large number of orbitals that are involved in the higher stage compounds, the rst e orts toward rst principles band calculations have been for stage 1 compounds with small unit cells. The band structure for the simplest graphite intercalation compound, rst stage p p C6 Li with a p 3 3R308 in-plane superlattice, was rst calculated by Holzwarth et al. (1977, 1978a, b), assuming interlayer intercalate stacking. This material is amenable to any of the standard band calculational methods, and excellent results for Ek are expected using current techniques. From the basic calculation for intercalate stacking, it should be possible to calculate Ek) for other stacking sequences (such as or ) using perturbation theory. In their treatment, Holzwarth et al. (1978b) used a modied KKR method based on an ionic potential for C Li and obtained the energy bands along high symmetry directions 6 shown in gure 43, which refer to the hexagonal Brillouin zone of gure 35. It is of interest to note that the bands in gure 43 derived from folding of the graphite -

Figure 43. Band structure of C6 Li by Holzwarth et al. (1978b) based on an ionic crystal potential (C Li ). Carbon -bands are represented by dashed lines; -bands are 6 represented by solid lines. Dispersion relations are shown along MTK (kz 0 plane), A (the kz direction), and LRASH (kz =Ic plane). Note that the Li s-bands lie above the Fermi level.

78

M. S. Dresselhaus and G. Dresselhaus

bands (dashed lines) remain highly graphitic in the intercalation compound. The band labelled 1 in gure 43 is derived from the Li s-band. Since the Li 1 band lies above the Fermi level throughout the Brillouin zone, it is concluded that the Li intercalant is fully ionized (fractional charge transfer f 1) with one electron/Li atom transferred to the graphite bands. It is also of interest to note the very weak dispersion of the graphite -bands along kz , and the greater dispersion for the lithium 1 band along kz . Holzwarth et al. (1978b) used their calculation along high symmetry directions to evaluate coe cients for an LCAO interpolation expansion (Slater and Koster 1954) to obtain energy bands throughout the Brillouin zone. These LCAO-derived energy bands were then used to calculate the Fermi surface (shown in gure 44). E ective masses and de Haasvan Alphen cross-sectional areas for six principal orbits on the Fermi surface thus obtained are given in table 14 for the two bands which give rise to the Fermi surface (see gure 43). Since the Li 1 band lies above the Fermi level and is only weakly hybridized with the graphite -bands, the Fermi surfaces in gure 44 are of primarily graphite -band character. Experimental conrmation of the one-electron energy dispersion relations calculated by Holzwarth et al. has come from angle-resolved photoemission energy distributions measured by Eberhardt et al. (1980) , who obtained -point energies in excellent agreement with the Holzwarth calculation as shown in table 15. In this table the zone-folded bands are indicated by an asterisk. For the graphite bands that are not zone-folded, there is a clear indication of the levels in the intercalation compound with those in the graphite host. A self-consistent band calculation for C8 K with a p2 2R08 superlattice and intercalate interlayer stacking (see gure 20) has been carried out by Ohno et al. (1979). These authors took advantage of the observation that the real space unit cell could be chosen to include only two intercalate layers, thereby simplifying their calculation considerably. The resulting Brillouin zone is the distorted hexagonal zone shown in gure 45. Atomic pseudopotentials were chosen for the C and the K atoms, and an ionic Madelung potential was introduced to account for the charge transfer. The Madelung potential was determined self-consistently, yielding a fractional charge transfer of f 0:6, which is to be contrasted with f 1 for the fully ionized intercalant obtained theoretically by Holzwarth et al. in C6 Li, and conrmed by the angle-resolved photoemission experiment of Eberhardt et al. (1980). The band

Table 14. de Haasvan Alphen frequencies and masses from the band calculation for C6 Li by Holzwarth et al. (1978b). Orbit Band designation 1 2 3 4 Lower Upper Lower Upper Plane kz 0 kz 0 kz =Ic kz =Ic 2 kx 3a0 ky 0 Centre H A L M Carrier type Electron Electron Hole Electron Electron Hole DHVA frequencies (104 T) 0.40 0.21 0.54 0.48 0.20 0.55

Masses m=m0 0.98 0.32 0.64 0.47 0.71 2.13

5 Lower 6 Lower

Intercalation compounds of graphite

79

Figure 44. Perspective drawings of the lower (a) and upper (b) band Fermi surfaces of C6 Li based on the calculation by Holzwarth et al. (1978b) using a C Li potential and 6 LCAO interpolation. The surfaces are shown in half of the Brillouin zone between the kz 0 and kz =Ic planes.

Table 15. -point binding energies for C6 Li and C8 Rb relative to EF (in eV). Asterisk denotes zone-folded bands. Each original graphite band contributes more than one energy level at k 0 in the compound. The level in C6 Li is folded back from the degenerate K1 point graphite level. Graphite Symmetry label 1 2;3 1 1 2;3 ; 2;3 1 1 1 1 Experimenta 0 74.6 0.3 77.2 0.3 78.1 0.3 . . 720.6 0.3 Theoryb 0 74.7 76.6 78.2 720.1 720.7 C6 Li Experimenta 70.5 0.3 75.5 0.5 79.3 0.3 713.0 0.5 . 715.2 0.5 722.5 0.4 Theoryc 71.3 75.9 79.3 713.3 714.9 715.4 721.8 C8 Rb Experimentd 70.4 0.2 73.8 75.9 79.5 0.3 711.7 0.5 716.0 0.2 (Rb4p) 722.5 0.3

References: a From Eberhardt et al. (1980). b From Painter and Ellis (1970). c From Holzwarth et al. (1978b). d From McGovern et al. (1980).

80

M. S. Dresselhaus and G. Dresselhaus

Figure 45. Brillouin zone for the p2 2 superlattice with stacking, and used by Ohno et al. (1979) for the energy band calculations in gure 46. The primitive unit cell has two formula units C8 K.

Figure 46. Electronic energy levels for rst stage C8 K with stacking of the intercalate layers calculated by Ohno et al. (1979). The symmetry points are labelled using the notation of gure 45. The potassium s-bands 1 and 1 are shown to be below the Fermi level of and Z respectively.

Figure 47. Fermi surfaces for C8 K for (a) the upper band and (b) lower band as deduced from the electronic bands of gure 46, and rst given by Inoshita et al. (1977). The open surfaces give rise to two-dimensional transport behaviour and the closed surfaces correspond to three-dimensional behaviour.

Intercalation compounds of graphite

81

structure calculated by Ohno et al. is shown in gure 46 and is very similar to that calculated previously by Inoshita et al. (1977) using a tight binding method. Because of the two C8 K groups that are included in the unit cell, two suitably modied potassium s-bands are identied in gure 46 and are labelled 1 and 1 . Since the 1 band crosses the Fermi level, some of the Fermi surface, especially that near Z, will be of basically potassium s-band character. Near the Brillouin zone boundary, the Fermi surface will be primarily of graphite -band character. Based on the energy bands of gure 46 a Fermi surface was calculated. The results were very similar to the Fermi surface previously calculated by Inoshita et al. (1977) and this Fermi surface is shown in gure 47. Note that the Fermi surface in gure 47 refers to the hexagonal Brillouin zone of gure 35. About the A axis, the Fermi surfaces are closed and three-dimensional in character, while the Fermi surfaces near the zone boundary are generally cylindrical or open, thereby giving rise to twodimensional properties. This Fermi surface has been extensively used to explain both transport and Fermi surface measurements in C8 K and frequent reference to gure 47 is made throughout section 4. The extension of these rst principles calculations to higher stage compounds has not yet been carried out. For the higher stage compounds, phenomenological models have been developed as described below, and application of these models have been made to the interpretation of experimental results. The rst principles calculations for stage 1 compounds discussed above provide strong evidence for the predominantly graphite -band nature of the occupied states. This fact coupled with strong experimental evidence that the structural, electronic and lattice properties of the intercalation compounds are largely graphitic in character suggest that a phenomenological model for the electronic structure should be based on the graphite -bands, with appropriate perturbations introduced to account for the intercalantgraphite bounding layer interaction. Various phenomenological models based on this idea have been proposed to discuss experimental results on graphite intercalation compounds. A phenomenological model which makes direct contact with both the SWMcC model and the rst principles band calculations for stage 1 alkali compounds has been discussed by Dresselhaus and Leung (1980). This phenomenological model which gives Ek throughout the Brillouin zone, makes use of the three-dimensional Fourier expansion of the graphite -bands, previously developed by Johnson and Dresselhaus (1973) to account for the optical properties of pristine graphite. In this model the superlattice periodicity appropriate to the intercalation compound is added to the basic graphite symmetry through suitable zone folding of the Brillouin zone for pristine graphite. The in-plane intercalate ordering gives rise to zone folding of the in-plane wave vectors and in this way is sensitive to the intercalate species. The staging phenomenon introduces a c-axis superlattice structure which is incorporated into the calculation by c-axis zone folding. The zone folding is accomplished by centring a Fourier expanded e ective hamiltonian k on each of the reciprocal lattice points for the superlattice. The zone-folded representation of the SWMcC matrix hamiltonian is then transformed into a layer representation. The e ect of the intercalate layer is treated by substitution of an intercalate layer hamiltonian for one appropriate to a graphite layer. The interaction between graphite bounding layers and the intercalate layer is treated as a perturbation. This perturbation is dependent on band parameters that are related to the overlap between the graphite -band pz orbitals and orbitals on the intercalate layer. These band parameters can be

82

M. S. Dresselhaus and G. Dresselhaus

Figure 48. Electronic energy levels derived by zone folding of the SWMcC dispersion relations for a p2 2 superlattice. Results by Leung and Dresselhaus (1981a) are shown for compounds with stages n 1, 2, 3.

evaluated by comparison with either rst principles calculations or experimental Fermi surface, optical and magneto-optica l data. It should be noted that once these interaction band parameters are evaluated for a specic intercalant, the phenomenological bands for that intercalant are determined for all stages that have the same in-plane ordering. The disperson relations obtained by this model for a p2 2R08 structure are shown in gure 48 for stages n 1, 2 and 3 using recent values for the SWMcC parameters (table 13) and a consistent set of expansion parameters previously given by Johnson and Dresselhaus (1973) for graphite, but introducing no additional parameters. In this form, the bands are broadly applicable to all compounds with the indicated in-plane structure. For stage 1, the results are compared in gure 49 with the rst principles calculation of Ohno et al. (1979) for C8 K and to Holzwarth et al. (1978b) for C6 Li and good agreement is obtained for the bands related to the graphite -bands. To obtain the hybridized metal s-bands by this model for a compound of arbitrary stage, it is necessary to t the phenomenological bands to the rst principles hybridized s-bands to determine the overlap band parameters between the graphite pz orbitals and the intercalate orbitals. One attractive feature of the phenomenological model is that once the band parameters are determined for a compound of any stage, such as a stage 1 compound where the graphiteintercalate interactions are important, the dispersion relations for all stages can then be evaluated without introduction of additional band parameters. The phenomenolo-

Intercalation compounds of graphite

83

Figure 49. Comparison of the SWMcC derived electronic levels (solid) for rst stage C6 Li and C8 K with those from rst principles bands (dashed) shown in gures 43 and 46 respectively (from Dresselhaus and Leung 1981b).

gical model thus provides a useful tool for the calculation of the electronic structure for high stage compounds where rst principles calculations are di cult to make and the phenomenological model is expected to be most convergent. Referring to gure 48, we note that the dispersion of the energy bands along kz is very small. The small magnitude of the dispersion arises from the connement of carriers within the n contiguous graphite layers between two consecutive intercalate layers. Such a carrier connement was also considered by Holzwarth (1980) in connection with a phenomenological model for Ek based on a single set of n contiguous layers bounded by two intercalate layers. Though the original graphite SWMcC bands display a larger dispersion in the kz -direction than do the bands in gure 48 for the intercalation compound, the same number of states is in both cases contained in an energy interval comparable to the band gaps in the intercalation compound. The rst principles calculations for the stage 1 compounds C6 Li and C8 K yield the screening of the intercalate layer by the graphite bounding layers directly by mapping the z-dependence of the charge density from the wavefunctions of the electron orbitals. For the phenomenological models used to calculate the dispersion relations for the higher stage compounds, this screening e ect is treated as part of the intercalategraphite bounding layer interaction. This interaction is explicitly evaluated from comparison with a rst principles calculation for stages n 1 and n 2 (if available), or from a non-linear ThomasFermi screening calculation as described below. It is then argued that for a given intercalant, the intercalantgraphite bounding layer sandwich for the higher stage compounds has approximately the same interactions as for the stage 2 compound. The e ect of screening of the intercalant by the surrounding graphite layers has been explicitly considered by Pietronero et al. (1978, 1979) who applied a Thomas

84

M. S. Dresselhaus and G. Dresselhaus

Table 16. Room temperature in-plane conductivity a for a number of stage 1 donor and accept or compounds.a Intercalant (Pristine graphite) K Rb Cs Li Br2 (stage 2) a 1 cm1 9 1:1 10 > > > > = 5 1:0 10 > > > > 5; 1:0 10 2:5 104
5

References Spain (1973) 8 > Murray and Ubbelohde (1969) > > > Ubbelohde (1972) < Gueard et al:1977a r > > Fischer 1979 > > : McRae et al: 1980a Basu et al: 1979 McRae et al: 1980a 8 < Ubbelohde (1972) Sasa et al:1970 : Pfluger et al:1980 Ubbelohde (1972) Pfluger et al:1980 Ubbelohde (1969, 1972) Vogel et al:1979 Foley et al: 1977 Vogel et al: 1977 8 < Fuzellier et al:(1977a) Thomson et al:1977 : Streifinger et al: 1979 Ubbelohde (1969, 1972) McRae et al: 1980a Fuzellier et al: 1977a Streifinger et al:l1979 Perrachon et al. (1979) Ubbelohde (1972)

2:4 10 5

2:2 105 3:5 105 1:6 105 4:7 105 5:0 105 2:3 105 1:3 105 1:1 105 1:6 105

ICl HNO3 AsF5 SbF5 H2 SO4 SbCl5 (stage 3) FeCl3 AlCl3


a

For HOPG, a 2:5 10 4 1 cm1 (Spain 1973) and for OFHC copper, 5:8 10 5 1 cm1 . Conductivity values are given for stage 1 compounds unless otherwise indicated. It should be noted that the maximum conductivity does not usually occur at stage 1. Representative values for a are listed, with some scatter found among values reported by the various authors.

Fermi model to calculate the screen charge distribution z as a function of distance z from the intercalant layer for a dilute compound. Their results are expressed in a simple analytical form z 3z3 z z0 4 ; 0 4:18

where the characteristic screening length z0 is related to the Fermi velocity F and the transverse dielectric constant "? by c h "? F 3 4:19 z0 ; 2 6e 0 where 0 is the two-dimensional charge density at z 0. Using values for 0 6:4 1014 cm2 and "? 3:4, a characteristic length of z0 4:0 A is obtained

Intercalation compounds of graphite

85

Figure 50. Dependence on reciprocal stage of the in-plane conductivity of several graphite alkali metal compounds, from the work of McRae et al. (1980a).

(Pietronero et al. 1979), which corroborates a number of experimental results that show the intercalate layer to be e ectively screened by a single graphite bounding layer. In principle, a self-consistent band calculation accounts for these screening e ects directly. On the other hand, the ThomasFermi treatment of the screening phenomenon has been applied to modelling the electrical conductivity (Pietronero et al. 1980a, b), the dielectric response (Pietronero and Strassler 1979), and the magnetic susceptibility (Safran and Di Salvo 1979) of intercalated graphite. 4.2. Transport properties 4.2.1. In-plane electrical conductivity The most widely studied property of graphite intercalation compounds is their electrical conductivity, mainly due to two distinguishing characteristics: the high inplane conductivity a and the large anisotropy of the in-plane to c-axis conductivity (a =c ) that can be produced by intercalation (Ubbelohde 1972, 1976). The ability to prepare intercalation compounds with room temperature conductivities comparable to copper but with one-fourth the mass density o ers attractive commercial applications. A detailed review of the transport properties of graphite intercalation compounds has recently been given by Fischer (1979). Electrical conductivity studies have been made on a wide variety of donor and acceptor compounds, and some typical values for stage 1 compounds are given in table 16. The large electrical anisotropy, resulting from the high in-plane conductivity a relative to the low c-axis conductivity c , makes accurate measurement of the conductivity tensor in the intercalate compounds di cult and a variety of special techniques have been developed to make measurements of improved reliability (Ubbelohde 1972, McRae and Herold 1977, Zeller et al. 1977, 1978, 1979). For the donor compounds the electrical anisotropy is not large enough to invalidate the use of the conventonal four-point method, and thus the four-point method is the method commonly used. For the acceptor compounds, the high anisotropy in the conductivity prevents unevenly injected current from becoming uniformly distributed over the sample, and the injected current tends to be concentrated in those layers where the current lead makes good electrical contact, so that the e ective conducting cross-sectional area is less than the geometrical area. For this reason, contactless r.f. measurements have been developed, utilizing eddy currents generated

86

M. S. Dresselhaus and G. Dresselhaus

by the r.f. magnetic elds. Similar arguments can be made in support of the use of r.f. contactless methods for the measurement of other transport properties in acceptors such as magnetoresistanc e and Hall e ect. Additional complications are associated with the tendency of the samples to cleave, exfoliate, form microcracks and desorb. The use of multistage and poorly characterized samples has sometimes made it di cult to compare measurements by di erent workers. In table 16, values for a using the r.f. contactless method are quoted for acceptor compounds, where available. The most widely studied aspects of the conduction behaviour is the large increase in the in-plane conductivity with increasing concentration for both donor and acceptor intercalants. This is illustrated in the plot of the room temperature conductivity a versus reciprocal stage (1=n) in gure 50 for several intercalants, where a for graphite is 40 m cm, and a 1=a . An order of magnitude increase in a is obtained for many intercalants, but the largest increases in a are found in the strong acid acceptor intercalants, especially AsF5 which has been reported by Vogel (1977) and Foley et al. (1977) to have a room temperature value for a 6:2 105 1 cm1 greater than or comparable to that in copper, 5:8 105 1 cm 1 . The large increase in a can be understood physically in the following way. Graphite, the host material, has a high in-plane mobility: room temperature mobility 13 000 cm2 =V s as compared to CU with 35 cm2 /V s and to Si with 1600 cm2 /V s (Spain 1973, Kittel 1952). Despite its high carrier mobility, graphite has a modest conductivity because of its low carrier concentration, with a room temperature value of 2 104 carriers/atom (or 2 1019 cm3 ). The intercalation process provides a mechanism for the injection of carriers (electrons for donors and holes for acceptors) from the intercalate layer, which has a relatively low mobility, to the graphite layers, which have a high mobility. A similar transfer of carriers from a low mobility doped region (GaAlx As1x ) to a high mobility undoped region (GaAs) also occurs in the superlattice semiconductor layer materials grown by molecular beam epitaxy (Dingle 1975, Esaki and Chang 1976). Because of the release of carriers from the intercalate layer, this layer becomes charged, positively for donors and negatively for acceptors. The charge that is released gives rise to a charge density in the graphite layers that falls o rapidly with distance measured from the intercalate layer (see section 4.1). The screening length is approximately the thickness of a single graphite layer, the graphite bounding layer, as has been documented by Raman spectroscopy (section 5.3), the FermiThomas screen calculation (Pietronero et al. 1978, 1979), along with other determinations. We thus assume that the graphite bounding layers (adjacent to the intercalant layers) have a high current density due to their high carrier density and that the current density on the graphite interior layers is much lower and can be approximated by some average value. In modelling the conductivity we thus assume that the graphite bounding layers make the major contribution to the in-plane conductivity. Though smaller in magnitude, the contribution from the graphite interior layers can be signicant in some cases, particularly for the alkali metal compounds. For most intercalants, the conductivity in the intercalate layer is negligibly small for several reasons: (1) the mobile electron and hole carrier density tends to be low (2) the carrier mobility is relatively low and (3) ionic mobilities tend to be very low. Such an interpretation assumes that the Fermi level lies below the intercalate conduction band but above the intercalate valence levels as shown explicitly by the energy band calculation for C6 Li (Holzwarth et al. 1977, 1978a, b). For the case of C8 K, it is however clear from

Intercalation compounds of graphite

87

Figure 51. Additive conductance model for graphite intercalation compounds (see equation (4.20)).

gure 46 that the electrons associated with the 1 potassium s-band will make a signicant contribution to the conductivity. McRae et al. (1980a) have measured the in-plane conductivity for very dilute alkali metal compounds (for example, stage 15 graphiteK), and have concluded on the basis of a threefold to fourfold increase in conductivity relative to graphite that the intercalant is present in an ordered array and not as a statistical distribution even in such dilute compounds. This observation together with X-ray di raction results on dilute compounds support a layered model for graphite intercalation compounds over a wide intercalate concentration range. For a single stage sample of stage n, the in-plane conductivity can be described qualitatively in terms of a simple phenomenological model (Dresselhaus and Leung 1979) based on the observation that the total conductance per unit cell of length Ic is equal to the sum of the conductances of the constituent layers contained within the unit cell. Several versions of this simple model have been given by other authors (Spain and Nagel 1977, Bok et al. 1978, Pietronero et al. 1980a). On the basis of the diagram shown in gure 51, the conductanc e can be written as ) 0 Ic a = 0 di i = 0 2c0 gb =0 n 2c0 gi =0 ; n 2; a a a a 4:20 00 n 1; Ic a = 0 di i = 0 c0 gb =0 ; a a a in which n is the stage of the sample, c0 is the graphite interlayer separation and Ic is the repeat distance. The graphite interior layer thickness is taken to be c0 , di is the 0 intercalate layer thickness, c0 is the thickness of a graphite bounding layer except for 00 stage 1 where it is denoted by c0 . Furthermore, a , 0 , i , gb and gi are respectively a the in-plane conductivities of the intercalation compound, pristine graphite, the intercalate layer, the graphite bounding layer and the graphite interior layer. The separation ds of two graphite bounding layers between which an intercalate layer is 0 00 00 sandwiched is ds di c0 . For the acceptor compounds c0 c0 c0 is a good approximation, while for the donor compounds the model suggested by gure 28 may be more appropriate . According to the model given by equations (4.20), a at low intercalate concentrations is dominated by the fraction of the unit cell occupied by the highly

88

M. S. Dresselhaus and G. Dresselhaus

0 conducting graphite bounding layers 2c0 =Ic , thereby yielding an approximately linear dependence of (a =0 ) on (1=n) in good agreement with much of the published a experimental data (see gure 50). In the low stage limit where the semi-insulating intercalate layer occupies a signicant fraction of the unit cell di =Ic , saturation behaviour in the (a = 0 ) versus a (1=n) relation of (4.20) results, leading eventually to a decrease in (a = 0 ) for the a lowest stage compounds, in agreement with experimental observations. The phenomenological model implies that the saturation and fall-o e ect at low stage is more important for intercalants with large di values (for example, for FeCl3 where di 6:02 A) and less important when di is small (for example, for Li where di 0:36 A), also in agreement with experimental observations. Contributions from the graphite interior layers are signicant when the charge transfer to these layer is appreciable and the carrier mobility on the graphite interior layers is signicantly higher than on the graphite bounding layers. The experimental conductivity data (as for example in the data of McRae et al. (1980a)) show that the maximum conductivity is found in the alkali metal donor compounds with K, Rb and Cs at a higher stage (n 3 to 5) than for the acceptor compounds where the maxima in most cases occur for n 2. These results suggest that the contribution from the graphite interior layers is greater for the graphitealkali metal compounds with K, Rb and Cs than for the acceptor compounds of comparable stage with intercalants such as AsF5 , SbF5 , HNO 3 and Br2 . This conclusion is also in agreement with magnetoreection results on the stage dependence of the Fermi level (see section 4.6) and with the oscillator strengths of infrared modes (see section 5.4). The phenomenological conductivity model suggests that maximum enhancement of the in-plane electrical conductivity of graphite intercalation compounds is achieved when the charge density transferred to the graphite layers from the intercalant is maximized, while minimizing the intercalate thickness di , and maximizing the mobility in the graphite bounding layers. Although the charge transfer to the graphite bounding layers tends to be signicantly higher for the alkali metal donor compounds than for typical acceptor compounds, the mobility on the graphite bounding layers tends to be lower for the alkali metal donor compounds, presumably due to the greater coupling to the intercalant and consequently more e ective carrier scattering (see gure 28). Using a rigid band, two-dimensional graphite model for both the electron and phonon systems, Kamimura et al. (1980) calculated the phonon-limited resistivity for C12n M alkali metal compounds (M K, Rb, Cs and stage n 2). In this calculation it was assumed that the alkali s-band lies above EF , as suggested by Knight shift data (section 4.7.2), specic heat data (section 4.7.4) and transport measurements (section 4.2). The di erence in electron density from one graphite layer to another was neglected. Two electron pockets about the zone edges HKH and H 0 K 0 H 0 are assumed and both intrapocket and interpocket scattering were considered. The phonon modes were approximated by those of graphite, the electronphonon matrix element was found using a rigid ion model and the pseudopotential approximation, and both phonon absorption and emission were considered. The dominant part of the resistivity was shown to arise from intrapocket scattering which results in a linear increase in with increasing temperature. Above 200 K, interpocket scattering becomes important and the in-plane resistivity rises more rapidly than the linear-T dependence that dominates below 200 K. Calculated values for the mobilities are 405

Intercalation compounds of graphite

89

Table 17. c-axis conductivity c and anisotropy ratio a =c for several donor and acceptor compounds. Intercalant HOPG K K Li AlCl3 Br2 HNO3 H2 SO4 AsF5 FeCl3 Stage 1 1 2 1 1 2 1 1 2 1 c 1 cm1 8.3 1:94 103 1:97 102 1:8 104 6.1 1.6 1.8 0.90 0.24 10 a =c 3:0 103 56 8:6 102 14 2:6 104 1:4 105 1:7 105 1:8 105 2:7 106 1:1 104 Reference Tsuzuku (1979) Murray and Ubbelohde (1969) Murray and Ubbelohde (1969) Basu et al. (1979) Ubbelohde (1972) Ubbelohde (1972) Ubbelohde (1972) Ubbelohde (1972) Foley et al. (1977) Vogel (1979)

and 702 cm2 /V s for stage 3 (C36 K) and stage 5 (C60 K) compounds; these values are about a factor of 20 lower than in pristine graphite. A model for the in-plane conductivity of acceptor compounds has been presented by Bok et al. (1978), in which the contribution from the graphite bounding layers was modelled by a two-dimensional carrier density n2d n3d c0 and the scattering time was estimated from the BlochGruneisen relation. Pietronero et al. (1980a) presented a model for the electrical conductivity of acceptor compounds based on a two-dimensional model for the electronic structure. This model was proposed for acceptor compounds with su cient intercalate concentration to drive the Fermi level below the band overlap region. According to their model, electronphonon scattering was considered by both phonons on the graphite layer and scattering from modes of the charged intercalate molecules. In this connection they solved Poissons equation to obtain the screening for the graphite bounding layer and considered the electronplasmon scattering, with the graphite bounding layer plasmon coupled to the intercalate layer molecular modes. Explicit results were obtained for stage n 1 and n 2 graphiteAsF5 , namely a 0:9 105 cm1 and a 2:3 105 cm1 , which are to be compared with the experimental values of 4:7 10 5 cm1 and 6:3 105 cm1 obtained by Foley et al. (1977). This model thus overestimates the scattering by a factor of 3 to 5. The authors attribute this discrepancy to an underestimate of the velocity of sound for the intercalate modes as obtained by their jellium model for the dynamics of charged intercalate molecules. On the basis of the results obtained for the K donor compounds and AsF5 acceptor compounds, an estimate made for stage 3 compounds suggests that the mobility for the acceptor compound is approximatel y ve times greater than for the donor compound, consistent with a much stronger coupling between the graphite bounding layers and the intercalant for donors than for acceptors. In a latter paper, Pietronero et al. (1980b) calculated the electrical conductivity of a charged graphite layer within the tight binding framework using analytic expressions for the electronphonon coupling and considering charge uctuations on the intercalant (Pietronero and Strassler 1979). On the basis of this calculation, the maximum conductivity for intercalated graphite at room temperature was estimated to be approximately twice that of copper. According to this approach, Pietronero and Strassler (1979) suggest that the intercalant and stage

90

M. S. Dresselhaus and G. Dresselhaus

dependence of the conductivity arise from charge uctuations in the intercalant layer, in contrast with the phenomenological approach given in equation (4.20). 4.2.2. c-axis electrical conductivity Whereas the addition of both donor and acceptor intercalants to graphite increases the in-plane conductivity a , the behaviour of the c-axis conductivity c is very di erent, in so far as donor intercalation tends to increase c while acceptor intercalation tends to decrease c (Ubbelohde 1972, Fischer 1979). Typical values for c for a number of donor and acceptor compounds are given in table 17. The experimental di culties associated with measurements of c-axis conductivity in pristine graphite and in the intercalation compounds has limited the amount and accuracy of available data. Values reported for 0 in pristine graphite vary considerably according to the c type of graphite (single crystal, kish graphite and HOPG), and the quality of the sample as measured by the residual resistance ratio dened as a 300 K= a 4:2 K (Spain 1971, Tsang and Dresselhaus 1976, Ono 1976, Tsuzuku 1979). A typical value for 0 in pristine graphite is 10 ( cm)1 . c There has been considerable disagreement with regard to the interpretation of these c-axis measurements. Recent experimental and theoretical work on the temperature dependence of 0 , and on the relation between 0 and the residual c c resistance ratio by Ono (1976) and Tsuzuku (1979) have shown that the dominant mechanism for c-axis conduction in pristine graphite is band conduction and not hopping. These studies have further shown that stacking faults associated with extended basal plane dislocations rather than the electronphonon interaction dominate room temperature scattering processes for c-axis conduction in samples with high in-plane residual resistance ratios a 300 K= a 4:2 K. Consistent with this interpretation is the lack of a signicant temperature dependence of 0 in pristine c graphite. It has also been found that the extended basal plane dislocations that dominate c-axis scattering become more dense through the intercalation process (Saito and Tsuzuku 1973). The tendency for intercalated samples to exfoliate and to form microcracks further complicates c-axis conductivity measurements in intercalated graphite. A simple phenomenological model for c-axis conduction suggested by gure 51 is the addition of the resistances of the various layers of the sample. Thus the c-axis resistivity is found by summing the resistance across a repeat distance Ic X 4:21 ci ti =Ic ; c
i

in which ci and ti are respectively the resistivity and thickness of layer i. For donor compounds the intercalate layer has relatively low c-axis resistivity, while for acceptor compounds, the c-axis resistivity of the intercalate layer is very high. For all donor intercalants that have been studied, the c-axis conductivity c increases upon intercalation relative to 0 for the graphite host. For example, table c 17 gives c =0 as 230 for C8 K, and as 2170 for C6 Li. The large increase in c for c the stage 1 compounds implies a strong overlap of the graphite pz -orbital with the potassium s-orbital and some overlap between consecutive intercalant layers, in addition to the large increase in the carrier density. Furthermore, inspection of the Fermi surface for C8 K, as shown in gure 47, indicates that there will be a signicant contribution to the c-axis conduction from the spherical carrier pocket, and a rather

Intercalation compounds of graphite

91

smaller contribution from the cylindrical carrier pockets which have very large e ective masses in the c-direction (two-dimensional carriers). The Fermi surface for C6 Li (see gure 44) also suggests orbits where c-axis conduction can occur e ectively. It should be noted that the anisotropy ratio for C6 Li is the lowest for any of the intercalation compounds (see table 17). For the higher stage alkali metal compounds, c remains larger than 0 , and this c can be understood in terms of hybridization of the graphite -orbitals on the graphite bounding layers with alkali metal s-orbitals. Also of importance is the introduction of new conduction paths along grain boundaries, defect sites containing intercalant, or possibly DaumasHerold domain walls (see section 2.4). On the other hand, measurements of c at room temperature for acceptor compounds typically show a signicant decrease in c relative to 0 for the graphite c host (Ubbelohde 1972, Foley et al. 1977), the magnitude of c =0 depending strongly c on intercalate species. This decrease in c for acceptor compounds is attributed to the high electrical impedance across the intercalate layer, arising from the low overlap between the graphite pz -orbitals and the molecular orbitals of the intercalant. Because of the order of magnitude increase in a and the order of magnitude decrease in c for typical acceptor compounds upon intercalation, an anisotropy ratio in excess of 10 6 can be achieved (Foley et al. 1977). Values for the anisotropy ratio a = c are given in table 17 for several donor and acceptor compounds. The transport of charged carriers across the intercalate layer may involve a tunnelling process in the case of the acceptor compounds, though the lack of correlation between the decrease in (c = 0 ) and the magnitude of the intercalate c thickness casts some doubt on the tunnelling mechanism for c-axis conduction. Furthermore, the observation that (c =0 ) is approximatel y an order of magnitude c greater at 77 K than at 295 K led Ubbelohde (1972) to conclude that c-axis conduction is driven by a band conduction mechanism in acceptor compounds, and not a hopping mechanism. Studies of spin di usion using the electron spin resonance (E.S.R.) technique (see section 4.7.3) have led Khanna et al. (1978) to

Figure 52. Temperature dependence of the total (electrons plus holes) carrier density for graphite. The curves are for three theoretical models and the points are experimental data (Spain 1973).

92

M. S. Dresselhaus and G. Dresselhaus

Figure 53. Temperature dependence of the mean in-plane carrier mobility for several types of graphite samples. The highest mobilities are achieved for single crystal samples. The low temperature mobility decreases with decreasing crystalline ordering perfection (Spain 1973).

Figure 54. Temperature dependence of the in-plane resistivity of graphite (inset from Spain 1973) and of the acceptor compound second stage graphiteAsF5 (Zeller et al. 1979).

propose a di usion mechanism for c-axis conduction in the intercalation compounds. As discussed in section 4.7.3, the c-axis conductivity for a one-carrier model is related to the di usion constant D by c e2 DNEF ; 4:22

Intercalation compounds of graphite

93

Figure 55. Temperature dependence of the resistivity of rst stage heavy-alkali metal graphite intercalation compounds (Potter et al. 1979), normalized to the resistivity at 300 K. Note the much stronger temperature dependence of donor compounds as compared with acceptors (see gure 54).

in which NEF is the density of states at the Fermi level. In this connection, temperature dependent studies of D in C8 K by Lauginie et al. (1980) using the E.S.R. technique suggest a weak temperature dependence of the di usion constant. Further careful measurements of c-axis conductivity in both donor and acceptor compounds will be particularly helpful in identifying the dominant conduction mechanism. 4.2.3. Temperature dependence of the conductivity The temperature dependence of the conductivity is qualitatively di erent in the intercalation compounds as compared with the graphite host material. For the case of pristine graphite the small magnitude of the Fermi energy results in a decrease in carrier concentration by a factor of 5 on cooling from room temperature to 77 K, with very little decrease upon further cooling to T 4:2 K (Spain 1973). This e ect is shown in gure 52 where the total carrier density N, due to majority electrons nM , majority holes pM and minority holes pm , is plotted as a function of temperature T. At T 0, intrinsic graphite exhibits carrier neutrality with the total electron carrier density nM equal to the total hole density, nM pM pm . A large increase in the inplane mobility is, however, achieved on cooling graphite as shown in gure 53, since the phonon scattering channel is greatly reduced at low T. Sugihara et al. (1978) have recently carried out a detailed study of the temperature dependence of the mobility of graphite. As a result of these temperature dependences of the carrier density and mobility, the in-plane conductivity 0 for graphite (HOPG), shown in a the inset of gure 54, increases slowly (by a factor of 2) on cooling from room

94

M. S. Dresselhaus and G. Dresselhaus

Table 18. The coefcients A, B, C in the temperature dependence of the resistivity A BT CT 2 for a few stage 1 donor compounds. Intercalant Stage K K Rb Cs 1 1 1 1 A (m cm) 1.86 8:36 10 3 9:66 10 3 4:06 10 2 B (m cm=K) 7:48 10 3 3:4 10 3 7:10 10 3 7:26 10 3 C (m cm=K2 ) 1:536 104 9:34 105 7:63 105 6:97 105 Authors Suematsu et al. (1980a) Potter et al. (1979) Potter et al. (1979) Potter et al. (1979)

temperature to 77 K and then increases more rapidly (by a factor of 5) on cooling from 77 to 4.2 K. In contrast to the behaviour of 0 , the temperature dependence of 0 is very c a weak, exhibiting a non-monotonic dependence on T and a marked dependence on sample quality as measured by the residual resistance ratio 300 K=4:2 K. The dominant scattering mechanism for c-axis conduction is defect scattering, in contrast to 0 where electronphonon scattering dominates (Tsuzuku 1979). a Since the carrier density for both donor and acceptor intercalation compounds is dominated by the carriers arising from charge transfer from the intercalate layers, the carrier density for graphite intercalation compounds is expected to be essentially temperature independent. Thus in contrast to the graphite host material, the temperature dependence of the conductivity for intercalated graphite is directly associated with the temperature dependence of the mobility. This argument accounts for the qualitatively di erent temperature-dependen t behaviour in graphite and its intercalation compounds. For the acceptors, the increase in a on cooling from 300 to 77 K is typically a factor of 3 or 4 (Ubbelohde 1972), with smaller increases on cooling to 4.2 K (Zeller et al. 1979) as shown in gure 54. A much larger increase in a is observed on cooling the alkali metal donor compounds (Murray and Ubbelohde 1969, Gue rard et al. 1977a, Tanuma et al. 1978, Potter et al. 1979, Onn et al. 1979). For example, the results of Potter et al. (1979) in rst stage C8 K show a factor of 10 increase in a from 300 to 77 K and a further factor of 20 increase from 77 to 4 K as shown in gure 55. For a signicant range of T, the functional form for the temperature dependence of the resistivity of C8 K is found to be A BT CT 2 (Potter et al. 1979, Suematsu et al. 1980a), though the values for the coe cients A, B and C given in table 18 exhibit considerable variation from one sample to another for the same nominal intercalant and stage. The constant term A is expected to depend on the residual resistance ratio and is observed to increase with the size of the intercalant ion, so that the larger Cs ions scatter the bounding layer electrons more e ectively than the smaller K ions (Potter et al. 1979). The term linear in T is consistent with an electronphonon scattering mechanism while the term quadratic in T has been identied with an electronelectron scattering mechanism because of the large temperature range over which this term contributes signicantly. Kamimura et al. (1980) have considered theoretically the electron phonon scattering process for the alkali metal donor compounds and have found the electronphonon scattering process to greatly reduce the in-plane room temperature mobility of the alkali metal donor compounds, in agreement with experiment. These authors considered both intrapocket scattering which is dominant below 200 K and interpocket scattering which becomes increasingly important at higher temperatures.

Intercalation compounds of graphite

95

The quadratic form A BT CT 2 has also been used to interpret the temperature dependence of the resistivity for acceptors, as for example stage 2 graphiteAsF5 (Zeller et al. 1979), though for the acceptor compounds, the coe cients A and B are smaller, indicative of the weaker temperature dependence. For both donor and acceptor compounds, a is highly sensitive to electron phonon scattering, consistent with typical metallic behaviour and with the temperature-dependent in-plane mobility in graphite. Applying a phonon scattering mechanism to the temperature dependent a data, characteristic values of the Debye temperatures D in the range 150 to 400 K have been proposed (Ubbelohde 1972), in good agreement with values of D between 325 and 425 K obtained by Mizutani et al. (1978) for Cx K and Cx Cs compounds on the basis of specic heat measurements. Strong evidence for metallic in-plane behaviour is also found in the large free carrier contribution to the optical reectivity (section 4.3) and to the large increase in the Fermi surface cross-sectional areas upon intercalation. In contrast, the temperature dependence of the c-axis conductivity of graphite intercalation compounds is qualitatively di erent from that of the graphite host, in so far as c increases with decreasing temperature for both donor and acceptor compounds. Typical acceptor compounds exhibit a one order of magnitude increase in c on cooling from room temperature to 77 K, while the donor compound C8 K exhibits a two orders of magnitude increase in c in this temperature interval (Murray and Ubbelohde 1969), so that the temperature dependence for a is similar to that for c , in the case of both donor and acceptor compounds. Another consequence of this temperature dependence of c is the persistence of the high anisotropy ratio a =c down to low temperatures. The temperature dependence observed for c suggests that electronphonon scattering could be a dominant scattering mechanism for c-axis transport in the intercalation compounds, in contrast to the behaviour in pristine graphite where stacking faults associated with extended basal plane dislocation are dominant. The observation of an increase in c with decreasing T is at variance with a hopping transport mechanism or a tunnelling mechanism for the acceptors. Whether or not a di usion mechanism, as has been suggested by Khanna et al. (1978) or Lauginie et al. (1980) , can account for the observed temperature dependence of c remains to be demonstrated. Further work is needed to investigate the possibility of the introduction of conduction paths with inplane components through the intercalation process; such a mechanism for c-axis conduction has been previously proposed for the pyrolytic graphite host material (Spain 1971). 4.2.4. Hall e ect For a single carrier metal, the carrier density is readily determined by measurement of the Hall coe cient RH , where RH is related to NH by NH 1 RH ec 4:23

and NH is the carrier density as determined from the Hall e ect. A second method for determination of the carrier density is by measurement of the magnetoresistanc e H 0=0 yielding an average Hall mobility hH i given by 2 1=2 hH i c 4:24 ; 0 H 2

96

M. S. Dresselhaus and G. Dresselhaus

which when combined with the zero eld conductivity 0 1=0 yields the carrier density NMR 0 ; ehH i 4:25

where NMR is the carrier density as determined from the relation (4.25). For simple metals that obey a one-carrier model, the carrier densities as determined by the two methods are equal, i.e. NMR N H . There are a number of other methods that are also used to determine the carrier density. These techniques are complementary and must be considered together to obtain a reliable picture for the carrier density of graphite intercalation compounds. A number of these methods are discussed elsewhere in section 4. One very sensitive method for measurement of the carrier density is by direct determination of the Fermi surface cross-sectional areas and the k-space volumes contained in each carrier pocket using quantum oscillatory techniques, such as the de Haasvan Alphen and Shubnikovde Haas techniques, discussed in section 4.5. Other methods that provide information on the carrier density include thermopower measurements, optical reectivity measurements near the plasma edge (section 4.3.2), electron energy loss spectroscopy (section 4.4.2), angle-resolved photoemission studies (section 4.4.3) and from calculated electronic energy dispersion relations Ek (section 4.1). From knowledge of the Ek relations and the Fermi energy EF , the volume in k-space of specic carrier pockets is obtained. Little work has been done with the thermopower technique to determine the carrier density, though Ubbelohde (1968b) did show that thermopower measurements could provide information on the sign of the dominant group of carriers. For the case of pristine graphite, the Hall coe cient is very small because the electron and hole densities and mobilities are approximately equal. The small inequalities give rise to a complicated dependence on temperature and magnetic eld (Spain 1973, Tsuzuku and Sugihara 1975, Ayache 1980). At very low temperatures and magnetic elds, the minority hole carriers make a signicant contribution to the Hall coe cient because of the relatively high mobility of these carriers. Hall measurements have been carried out on low stage alkali metal compounds (Murray and Ubbelohde 1969, Guerard et al. 1977a, Suematsu et al. 1980a, Onn et al. 1979) and in the acceptor compounds graphiteFeCl3 (Dzurus and Henning 1957), graphiteH2 SO 4 (Ubbelohde 1969) and graphiteHNO3 (Ubbelohde 1969, Kawamura et al. 1975). The standard conguration for these Hall measurements is Hkc and Jka. The magnitudes of the Hall coe cient RH in the intercalation compounds tend to be small, and thus there has been some di culty in certain cases to obtain quantitative measurements. Substantial disagreement is found in the literature for RH measured on di erent samples of the same nominal intercalant and stage and at the same temperature, as can be seen from the values of RH given in table 19 for several donor compounds. Nevertheless, a number of important conclusions can be reached from the Hall data that are presently available. Whereas RH for pristine graphite is negative at room temperature, measurements of RH in all acceptor compounds show a positive Hall coe cient, even at low intercalate concentrations where the increase in conductivity is only 30%. The positive sign of the thermopower for intercalated graphiteHNO3 (Ubbelohde 1968a, b) is also consistent with these Hall coe cient

Intercalation compounds of graphite

97

Table 19. Average in-plane Hall mobility hH i, Hall coefcient RH , and carrier densities NMR and NH for various donor and acceptor compounds. T (K) 1.2 1.2 1.2 1.2 4.2 77 77 78 77 300 1.2 4.2 77 77 268 300 1.2 4.2 4.2 77 77 300 1.2 1.2 4.2 77 300 4.2 77 300 77 300 77 300 4.2 77 300 4.2 77 300 300 300 300 300 300 hH i (cm2 /V s) 1 590 5 330 12 880 6 900 1:6 104 1 190 1 000 722 1 050 130 4 980 7 500 1 800 1 780 460 251 3 050 77 400 7 600 1 000 2 500 466 2 780 3 760 7 700 3 700 2 300 10 000 1 190 112 1 800 150 1 070 110 3 770 1 300 268 3 300 1 800 1 380 1 980 3 200 1 750 920 1 630
b

Compound C8 K a C8 K a C8 K a C8 K a C8 K b C8 K c C8 K b C8 K a C8 K a C8 K b C24 Ka C24 Kb C24 Kb C24 Ka C24 Ka C24 Kb C36 Ka C36 Kb C36 Ka C36 Ka C36 Kb C36 Kb C48 Ka C48 Ka C48 Kb C48 Kb C48 Kb C8 Rbc C8 Rbc C8 Rbc C24 Rbc C24 Rbc c C8 Csc C8 Cs C24 Csb C24 Csb C24 Csb C36 Csc C36 Csc C36 Csc C8 AsF5 d C16 AsF5 d C6 HNO3 d C6:6 FeCl3 d C13:2 FeCl3 d

NMR (cm3 ) 5:12 1020 3:3 1021 3:9 1021 6:48 1020 5:98 1020 4:3 1021 1:69 1019 4:76 1019 5:5 1021 3:30 1019 6:5 1021 4:5 1021 2:04 1019 5:49 1019 2 1021 1:5 1021 6:0 1020 2:0 1021 6:2 1021 4:9 1021 5:1 1021 1:8 1021 1:6 1021 5:3 1020 1:26 1021 1:22 1021 1:02 1021 7:1 1020 8:9 1020

RH (cm3 s1 ) 1:2 103 0:32 103 j1:9j 103 0:13 103 4:5 104 8 105 1:1 104 1:2 104 j3:4j 103 1:45 104 6:33 103 1:6 103 1:23 103 4:93 103 2:97 103 8:8 104 2:6 102 2:80 102 1:6 103 2:11 102 1:23 103 8:8 104 j8:03j 103 1:08 101 5:7 103 5:9 103 8:0 103 1:1 104 6 105 4:8 105 1:8 104 1:9 103 1:9 103 8:4 104 3:3 103 3:7 103 5:2 103

NH (cm3 ) 9:86 10 20 3:4 10 21 5:1 10 21 1:27 10 21 2:24 10 21 7:1 10 21 2:32 10 20 2:23 10 20 3:4 10 21 2:96 10 20 5:1 10 21 7:1 10 21 7:77 10 20 5:78 10 19 1:1 10 21 1:1 10 21 7:8 10 20 8:1 10 21 3:3 10 21 3:3 10 21 7:4 10 21 1:8 10 21 1:7 10 21 1:2 10 21

References: a Suematsu et al. (1980a);

Onn et al. (1979); c Guerard et al. (1977a);

Vogel (1979).

98

M. S. Dresselhaus and G. Dresselhaus

Figure 56. Hall coe cients RH of graphiteK intercalation compounds as a function of temperature as measured by Suematsu et al. (1980a). The Hall coe cients for the n 2 compounds are negative for the entire temperature range, while for n 1, RH is negative at low temperatures and changes to positive at higher T. Signicant di erences in behaviour are found for di erent samples (see symbols in gure) of the same nominal stage.

measurements. In contrast, RH for donors is negative, except for the anomalous behaviour of the stage 1 alkali metal compounds to be discussed below. Measurements of RH on both donor and acceptor compounds are consistent with an increase in carrier concentration upon intercalation, with smaller increases found for acceptor compounds than for donor compounds of comparable stage index. This result is in agreement with the smaller fractional ionization of the intercalant for the case of the acceptor compounds (where f is typically 0.3) as compared with the donor compounds (where f can be as large as unity). Thus the high in-plane conductivity of acceptor compounds relative to that for donor compounds is associated with the much larger values found for the in-plane mobilities of the acceptor compounds, discussed further in section 4.2.5. Except for the stage 1 alkali metal compounds, the experimental results for RH have been interpreted in terms of a one-carrier model, with a negative Hall coe cient for donors and a positive Hall coe cient for acceptors. In this connection it should be noted that measurements of the quantum oscillatory phenomena associated with the Fermi surface of intercalated graphite (section 4.5) provide strong evidence for more than one inequivalent carrier pocket in the Brillouin zone. From an analysis of RH for stage 2 alkali metal compounds, Onn et al. (1979) also concluded that more than one carrier type is necessary to interpret their Hall measurements above 77 K. As models for the Fermi surface for various graphite intercalation compounds become available, a more quantitative analysis of the Hall measurements will

Intercalation compounds of graphite

99

become possible. Since detailed models of the Fermi surface are presently unavailable except for C8 K and C6 Li, the analysis of the Hall measurements has been carried out in terms of a one-carrier model and (4.23) to yield NH . The results thus obtained are given in table 19, except for the stage 1 compounds, which are discussed below. For the alkali metal compounds with n 2, the Hall coe cients exhibit no magnetic eld dependence below 2 T (Suematsu et al. 1980a) and display only a weak temperature dependence (Suematsu et al. 1980a, Onn et al. 1979). It should be noted that there is signicant disagreement between the values for RH in table 19 reported by di erent authors. While Hall measurements provide important information on the dominant carrier type and on the carrier density, measurement of RH by itself does not provide a denitive determination of the carrier density. The data in table 19 are, however, consistent with a carrier density NH that (1) has only a weak temperature dependence, and (2) increases with decreasing stage (see gure 56). In contrast, the Hall coe cient RH for the stage 1 alkali metal compounds display an unusual behaviour with regard to the sign of RH and the temperature and magnetic eld dependence of RH. Anomalous positive values of RH have been reported by Guerard et al. (1977a), Potter et al. (1979) and Suematsu et al. (1980a). This observation has led Guerard et al. (1977a) to interpret their Hall measurements on the stage 1 compounds in terms of a two-carrier model, where RH is given by RH 1 p2 n2 p n ec pp nn 2 4:26

and where p and n denote the hole and electron carrier densities, and p and n are the corresponding mobilities. For the two-carrier model, the resistivity and average Hall mobility are given by and 1= nn pp e p 1 n = p 2 n p : p=n n = p 0 H 2 n 4:27 4:28

Since the galvanomagneti c measurements yield three independent quantities , RH and (=0 H 2 ), while equations (4.26)(4.28) depend on four quantities n, p, n and p , an explicit determination of these parameters is not possible. Measurements of the Hall e ect by Guerard et al. (1977a) and Suematsu et al. (1980a) agree that RH for the stage 1 alkali metal compounds exhibits a strong temperature dependence. These authors, however, do not agree on the detailed temperature dependence, which may in practice be sample-dependent. This is suggested by the work of Suematsu et al. which shows a sign change in RH from negative to positive above T 170 K for one sample but not for another (see gure 56). In contrast, Gue rard et al. found RH for C8 Rb to be positive at 4 K and negative at 77 and 300 K. These sign changes in RH as a function of temperature are consistent with a conduction process where both electrons and holes make a signicant contribution. It should be also noted that Potter et al. (1979) reported small positive values for RH in C8 K, C8 Rb and C8 Cs for T 4:2 K with a linear dependence of RH on magnetic eld for these stage 1 compounds. The multicarrier conduction in stage 1 alkali metal compounds is also supported by the electronic energy band calculations of Inoshita et al. (1977) and Ohno et al. (1979). These models yield the multiconnected Fermi surface shown in gure 47 where electron

100

M. S. Dresselhaus and G. Dresselhaus

orbits occur around the spherical pieces of Fermi surface about the - and A-points and the cylindrical pieces about the K-point, though hole orbits occur about the cylindrical pieces near the M-point. It would be of great interest to calculate the Hall coe cient for such a Fermi surface. Since NH cannot be obtained reliably from (4.23) for the case of the stage 1 alkali metal compounds, these results for NH are not included in table 19. 4.2.5. Magnetoresistance As discussed in the previous section, analysis of magnetoresistanc e data yields values for the average Hall mobility, and when combined with zero eld resistivity measurements, the average carrier density NMR can be deduced. The angular dependence of the magnetoresistanc e further provides information on the topology of the Fermi surface, and is the method traditionally used to determine whether the Fermi surface is open or closed. Finally, magnetoresistanc e measurements at high elds and low temperatures exhibit quantum oscillatory behaviour (the Shubnikov de Haas e ect), which determines Fermi surface cross sectional areas, as discussed in section 4.5. The analysis of magnetoresistanc e data to yield hH i and NMR is carried out in the low magnetic eld limit where quantum e ects are unimportan t and where exhibits an H 2 eld dependence, as given by (4.24). The resistivity of pristine graphite exhibits a eld dependence of H 1:85 which is su ciently close to the conventional H 2 dependence that a simple analysis of magnetoresistanc e data can be made to yield an average mobility hH i c2 =0 H 2 1=2 , which agrees within a few per cent with a more quantitative approach that includes the kz -dependence of the masses explicitly (Tsuzuku and Sugihara 1975). Because of the departure of from a strict H 2 dependence in graphite, the values of are reported at 0.3 T, following the convention of Soule (1958), when used for a determination of the average Hall mobility hH i. In the low eld range, hH i for graphite exhibits only a weak dependence on magnetic eld. However, for H > 6 T , a linear magnetic eld dependence of is observed, and this anomalous behaviour has been attributed to impurity scattering in the quantum limit by McClure and Spry (1968). Magnetoresistanc e measurements on pristine graphite yield a value of hH i 1:3 104 cm2 =V s at room temperature, and for HOPG an increase by a factor of 40 to 50 is achieved on lowering the temperature to 4 K. For single crystal graphite an increase by a factor of 150 is achieved in this temperature interval, as shown in gure 53. The highly mobile minority hole carriers contribute signicantly to the Hall mobility and are responsible for anomalous temperature and magnetic eld e ects at low temperatures. At 4 K, typical values for the relaxation time and mean free path for graphite are t 5 1011 s and l 103 cm. Magnetoresistanc e measurements on graphite intercalation compounds also yield relatively high carrier mobilities, though smaller in magnitude by about a factor of 10 to 102 relative to pristine graphite at comparable temperatures. Magnetoresistance measurements made in the quadratic magnetic eld region below 4 T yield average Hall mobilities independent of magnetic eld. A linear H dependence in is, however, found at high elds (Suematsu et al. 1980a), as is also the case for pristine graphite (McClure and Spry 1968). The most extensive magnetoresistanc e measurements have been made on low stage alkali metal donor compounds (Murray and Ubbelohde 1969, Gue rard et al. 1977a, Suematsu et al. 1980a, Onn et al. 1979), though some measurements have also been made on acceptor compounds with

Intercalation compounds of graphite

101

HNO3 (Ubbelohde 1969, Kawamura et al. 1975, Vogel 1979), FeCl3 (Vogel 1979) and AsF5 (Zeller et al. 1979). Systematic magnetoresistanc e measurements as a function of T and H have been carried out for only a few intercalants and stages. Some typical values for hH i that have been obtained are given in table 19. Also included in this table are values for NMR obtained from (4.25). It is of interest to note that NMR and NH are not in quantitative agreement, indicative of the approximate nature of the one-carrier model assumed in the analysis. As discussed in section 4.2.4, the analysis of (4.25) assumes a one-carrier model, which is not valid for the stage 1 alkali metal compounds, and therefore N MR is not given in table 19 for the stage 1 alkali metal compounds. The stage 1 alkali metal compounds, the compounds that have been studied most extensively, exhibit an average Hall mobility hH i 102 cm2 =V s at 300 K, hH i 103 cm 2 =V s at 77 K and hH i 104 cm2 =V s at 4 K. Although the high stage compounds also exhibit signicant increases in mobility as the temperature is lowered, the e ect is much smaller than for the stage 1 compounds, as can be seen in table 19. It is of interest to note that the calculated value of 405 cm2 =V s for C36 K by Kamimura et al. (1980) is in good agreement with the measurement of Onn et al. (1979), though the calculated value of 705 cm 2 =V s for C60 K seems low relative to the stage dependent results of Onn et al., which show signicant increases in hH i with increasing stage n. Despite the scatter in the data in table 19, the decrease in NMR with increasing stage is apparent, as is the increase in hH i with increasing stage.

Figure 57. Angular dependence of the transverse magnetoresistance in graphite and in stage 4, 3 and 1 graphiteK compounds. The data were taken by Suematsu et al. (1980a) at 4.2 and 2.4 K. The oscillatory behaviour near 0 in the stage 4 sample is associated with the Shubnikovde Haas e ect. Of particular interest is the cusp-like behaviour near 908 for the stage 1 and 3 samples.

102

M. S. Dresselhaus and G. Dresselhaus

The magnetoresistanc e and Hall e ect measurements for the acceptor compounds generally yield higher mobilities and lower carrier densities than for donor compounds of the same nominal stage, as shown in table 19. As for the case of the donor compounds, the mobility for acceptor compounds increases as the temperature is decreased. For example, Zeller et al. (1977) obtained for a stage 1 graphite AsF 5 sample, values of hH i 2:0 103 cm2 =V s at 293 K, hH i 1:4 104 cm2 =V s at 77 K and hH i 4:67 10 4 cm2 =V s at 4.2 K. For this acceptor, these authors found (=) to go as H 2 at low elds, but to saturate at high elds, with the saturation e ect becoming more pronounced as T was decreased. An interpretation of the temperature and eld dependence of the mobility provided evidence for two types of hole carriers, each having di erent mobilities. Using the value of NMR 9 10 20 =cm 3 and an intercalate density of 5:9 1021 =cm3 , a value for the fractional charge transfer f 0:15 was obtained, in fair agreement with other estimates of f for rst stage graphiteAsF5 (Weinberger et al. 1978a) (section 4.7.1) who reported f 0:24, and Hanlon et al. (1977) (section 4.3) who reported f 0:05). The only system for which the angular dependence of the transverse mgnetoresistance has been measured to obtain information on the Fermi surface topology is the graphiteK system (Suematsu et al. 1980a). The angular dependence of the magnetoresistanc e for an ellipsoidal Fermi surface has the functional form cos 2 sin2 , where is the angle of H relative to the c-axis. For graphite, is very small since the Fermi surfaces for electrons and holes are well approximated by highly elongated ellipsoids, with cos 2 as shown in gure 57(a). For small angles , the magnetoresistance s for the graphiteK compounds show an angular dependence similar to that of graphite, though comparative values of for graphite and the intercalation compounds have not yet been reported. A more rapid increase (by a factor of 2) was found in the high eld region (H > 2 T) by Suematsu et al. (1980a) for the magnetoresistanc e = 0 with Hkc than for H ? c. For 908, major di erences were found in the intercalation compounds relative to graphite, with a cusp-like behaviour observed near 908 for the stage 1 and stage 3 compounds (gures 57(d) and 57(c)). Some structure near 908 is also observed for the stage 4 compound (gure 57(b)). The cusp-like behaviour for the stage 1 and stage 3 compounds is identied with open orbits for 908, which is consistent with the Fermi surface calculated by Inoshita et al. (1977) . The open orbit for the stage 3 compound was identied with open orbits arising from c-axis zone-folding e ects. The e ect of c-axis zone folding on the Fermi surface is illustrated in gure 42, where it is seen that the c-axis cuts introduced by the superlattice can result in cylindrical pieces of Fermi surface. The structure in gure 57(b) for the stage 4 compound at 908 is tentatively identied with magnetic breakdown e ects, across open orbits as indicated by the dashed lines in gure 42(b) (Suematsu et al. 1980a). The arguments further indicate that with increasing stage, the probability of magnetic breakdown between open orbits increases, so that as n ! 1, the angular dependence observed for pristine graphite (gure 57(a)) should be achieved. Thus the angular dependence of the magnetoresistanc e supports a c-axis zone-folding model, that was previously proposed by Tanuma et al. (1978) to explain the quantum oscillatory e ects observed in alkali metal donor compounds (see section 4.5).

Intercalation compounds of graphite

103

4.2.6. Transport anomalies at structural phase transitions Anomalies in the temperature dependence of the resistivity, Hall e ect and magnetoresistanc e have been used to identify phase transitions in a variety of donor and acceptor compounds. Temperature-dependen t transport measurements have the advantage of simplicity relative to other techniques sensitive to phase transitions, but have a disadvantage in so far as they do not yield information on the structure of the phases involved in the transition. Temperature-dependen t transport measurements have been used to identify structural phase transitions in several low stage alkali metal donor compounds (Onn et al. 1977, 1979, Suematsu et al. 1980a, b) and in several acceptor compounds, including graphiteHNO3 (Nixon et al. 1966, Ubbelohde 1968a, b, 1969, Kawamura et al. 1975), graphiteH2 SO 4 (McRae et al. 1980a), and graphiteAsF5 (Zeller et al. 1979). Phase transitions are expected to give rise to anomalies in the transport properties in so far as these transitions a ect the scattering mechanisms. For example, anomalies observed at the orderdisorder transformation in the graphiteHNO3 compounds are consistent with a signicantly increased scattering of the -electrons in the graphite bounding layers by a disordered intercalant relative to an ordered intercalant. This scattering process further suggests that even in the highly anisotropic acceptor compounds, the scattering mechanisms have some threedimensional character. In this connection, it is of interest to note that the magnitudes of the anomalies in the transport properties tend to be larger for the acceptor compounds than for the donor compounds. Since the temperature-dependen t di raction studies yield both the transition temperature and the identity of the stuctures involved in the structural phase transition, the topic of phase transitions is discussed in detail in the context of structural properties (see section 3.4). 4.3. Optical properties 4.3.1. Optical properties of graphite In general, the frequency-dependent optical properties of a solid are determined by the relative contribution to the complex dielectric tensor e from free carriers (ecarriers ), interband electronic transitions (einterband ), phonon excitations (elattice ), and other processes not treated in detail (ecore ). The transitions contributing to ecore occur at frequencies higher than the frequency range under consideration, and therefore can be treated in terms of an average quantity. The complex dielectric tensor is thus written as e e1 ie2 ecarriers einterband elattice ecore ; 4:29

in which e1 and e2 are the real and imaginary parts of the dielectric tensor. The normalization of the dielectric tensor requires that in a vacuum e and ecore reduce to the unit tensor. The tensor for the complex index of refraction n ij is dened in terms of the complex dielectric tensor n ij2 e1 ie2 : 4:30

In the case of graphite and its intercalation compounds, the hexagonal symmetry requires two independent tensor components "a and "c , corresponding respectively to the optical electric eld parallel and perpendicular to the layer planes. Because of the very di erent magnitudes of "a and "c , graphite and its intercalation compounds exhibit highly anisotropic optical properties.

104

M. S. Dresselhaus and G. Dresselhaus

In the case of pristine graphite, ecarriers , elattice and einterband all contribute signicantly at infrared frequencies. Large changes in the relative magnitudes of these contributions occur upon intercalation, providing important information on both free carrier e ects and on the electronic energy of the graphitic and intercalate levels for each intercalation compound. Although the electronic structure of graphite intercalation compounds relates closely to the structures of the graphite host and of the intercalant, a number of important di erences are found in the optical properties. These di erences are described below and classied under the categories of free carrier e ects, other optical structures and optical anisotropy where the behaviour for the polarizations Ekc and E ? c are contrasted. Optical measurements on intercalated graphite are normally carried out on samples prepared from highly oriented pyrolytic graphite (HOPG) because this host material yields samples of suitable size and good optical surfaces. Since optical measurements on HOPG and single crystal graphite samples yield the same optical spectra (Ergun 1968), it is inferred that optical studies on HOPG and on intercalated HOPG samples provide information on the electronic and lattice mode structure of intrinsic graphite and intercalated graphite respectively. For pure graphite, the dielectric tensor e is dominated by the einterband term in the infrared and visible regions of the spectrum, since the valence and conduction bands of graphite are essentially degenerate along the edges of the Brillouin zone HKH and H 0 K 0 H 0 (see gure 37). Along the edges of the Brillouin zone, the E3 bands are split only by the very small spinorbit interaction energy of 2:2 104 eV (McClure and Yafet 1962, Dresselhaus et al. 1966). Interband transitions thus occur starting at the very low spinorbit energy for the kz value where the Fermi level crosses the HKH axis (Dresselhaus and Dresselhaus 1965) and continuing to higher energies for di erent kz values. For pristine graphite, the Fermi level is located at approximately the mid-point of the small band overlap region which is 0.04 eV wide. The onset of interband transitions occurs at a very low frequency and therefore contributes strongly to einterband . These low frequency interband transitions e ectively obscure the characteristic structure in the optical reectivity normally identied with the plasma frequency. For pristine graphite, the plasma frequency for Eka is estimated to be at 0.5 eV (McClure 1964), corresponding to a room temperature carrier density for holes and electrons of 2 10 4 /atom (as determined from transport measurements (Spain 1973) and an in-plane dielectric constant of "a 2:8 (Ergun 1968)). At low frequencies, structure associated with the E1u I.R.-active lattice mode is observed by Brillson et al. (1971) at 1588 cm1 (0.197 eV), and the A2u mode by Nemanich et al. (1977b) at 868 cm1 (0.108 eV) (see gure 75). The in-plane and caxis lattice modes contribute respectively to the dielectric tensor components "a and "c . At higher frequencies, structure in the optical reectivity of pristine graphite is identied with specic electronic interband transitions (Taft and Philipp 1965, Philipp 1977). In the vicinity of 0.8 eV, K-point transitions from occupied states in the E2 valence band to empty states in the E3 conduction band above the Fermi level (0.706 eV) and from occupied states in the valence E3 band to empty conduction states in the E1 band (0.889 eV) are observed using modulated reectivity techniques (Bellodi et al. 1975, Nakao 1979, Misu et al. 1979). Because of trigonal warping e ects, there is a strong admixture of the E3 bands (Dresselhaus 1974) so that both E3 bands couple to the E1 and E2 bands and contribute to the lineshape of each component of the doublet structure (Nakao 1979). The dominant structure in the optical reectivity of graphite is, however, observed at 5 eV, and identied with

Intercalation compounds of graphite

105

Figure 58. Reectivity measurements by Guerard et al. (1977a) for stage 1, 2 and 3 graphite Rb donor compounds. The data exhibit a metallic-type plasma edge.

interband transitions between -bands around the M-point in the Brillouin zone (Taft and Philipp 1965, Balzarotti and Grandolfo 1968, Greenaway et al. 1969, Johnson and Dresselhaus 1973). Interband transitions involving -bands give rise to optical structure in the 13 eV range. A large anisotropy in the optical properties of graphite is observed primarily because ecarriers is much larger for the polarization Eka than for Ekc due to the large anisotropy in the e ective mass tensor. Furthermore einterband is signicantly di erent for the two polarizations due to di erences in selection rules and magnitudes of matrix elements (Ergun et al. 1967, Greenaway et al. 1969, Anderegg et al. 1971, Johnson and Dresselhaus 1973). On the other hand, many of the interband transitions occur away from high symmetry points in the Brillouin zone, and therefore exhibit optical structures closely associated with maxima in the joint density of states. These maxima tend to be considerably less anisotropic than the other factors dominating the optical properties. Thus einterband for pristine graphite is less anisotropic than ecarriers . 4.3.2. Free carrier e ects in donor and acceptor compounds Intercalation with either donor or acceptor intercalants results in a large increase in ecarriers and a large decrease in einterband at low frequencies, so that the optical transmission exhibits a pronounced transmission window associated predominantl y with free carrier e ects. The large increase in ecarriers arises from the large increase in carrier density (by as much as a factor of 800 for the case of C6 Li) and is consistent with measurements of the electrical conductivity, and Hall coe cient discussed in section 4.2. At the same time einterband at low frequencies is much smaller in typical graphite intercalation compounds because the Fermi level is generally upshifted above the H-point E3 band extremum in donor compounds and downshifted below the K-point E3 band extremum in acceptor compounds so that the strong interband transitions across the nearly degenerate graphite E3 bands are suppressed. Therefore, for almost all of the graphite intercalation compounds that have been studied, the optical reectivity exhibits the characteristic Drude edge at the plasma frequency !p and the optical transmission is characterized by a transmission window near !p , typical of metallic free carrier absorption. As is illustrated in gure 58 for the alkali metal compounds with Rb (Guerard et al. 1977a), the reectivity is high for ! < !p ,

106

M. S. Dresselhaus and G. Dresselhaus

Table 20. Reectivity minima and transmission maxima associated with free carrier effects in various donor and acceptor compounds. Frequency (in eV) of reectivity minima (m) and transmission maxima (M)a 2.65(m) 1.8(m) 2.62(m) 1.72(m) 1.44(m) 2.64(M) 1.77(M) 1.27(M) 2.00(m) 1.25(m) 1.11(m) 2.8(m) 1.68(m) >3.4(m) 1.77{(m) 1.40{(m) 1.33(m) 1.22(m) 1.8(m) 1.4(m) 1.18(m) 1.13(m) 1.12(m) 1.27(m) 1.16(m) 1.14(m) 1.61(m) 1.29(m) 1.10(m) 1.06(M) 1.02(m) 1.10(m) 1.09(M) 1.22(m) 1.85(M) 1.29(M) 1.51(M) 1.10(M)

Intercalant K Rb

Stage 1 2 1 2 3 1 2 4 1 2 3 1 2 1 1 2 3 4 1 2 3 4 2 2 3 4 1 2 3 2 2 2 1 1 1 2 1 2

Authors Zamini and Fischer (1977) Guerard et al. (1977a)

Cs

Hennig (1965)

Puger et al. (1980)

Li Ba AsF5

Basu et al. (1979) Fischer (1977) Hanlon et al. (1977a) { Also Blinowski et al. (1980)

SbF5

Thompson et al. (1977)

SbCl5

Blinowski et al. (1980) Eklund et al. (1981)

HNO3

Fischer et al. (1976)

Br2 ICl FeCl3 H2 SO4 AlCl3


a

Hennig (1965) Blinowski et al. (1980) Blinowski et al. (1980) Hennig (1965) Perrachon (1978) Hennig (1965) Hennig (1965)

The reectivity minima and transmission maxima correspond approximately to the screened plasma frequency !p .

Intercalation compounds of graphite

107

Figure 59. Reectivity measurements by Eklund et al. (1981) for graphiteSbCl5 acceptor compounds of various stages. With decreasing stage index, the plasma edge becomes more pronounced, but does not shift much in frequency. Additional stage-dependent structure is found below the plasma edge and identied with intervalence band transitions. A simple model of Lorentzian oscillators (dots) is used to t the experimental data (lines).

and a sharp edge and a low minimum reectivity is found for ! !p . In some cases einterband makes an important contribution to the dielectric constant and shifts the plasma edge toward the interband frequency. Optical studies of free carrier phenomena have the advantage of requiring no electrical contacts and can be performed on samples in encapsulated ampoules with suitably chosen optical windows. To obtain reproducible optical measurements, especially on low stage reactive compounds, the samples are prepared in the same ampoules (in situ) as are used for the optical measurements (Hanlon et al. 1977a, b). Reliable results have also been obtained using special optical ampoules into which the samples are transferred after cleavage in an inert atmosphere inside a dry box. Pioneering work on the optical properties of graphite intercalation compounds was carried out by Hennig (1965) who examined the transmission window associated with the free carrier absorption process. Hennig noted that intercalation signicantly increased ecarriers , raised EF for donor compounds and lowered EF for acceptor compounds. Results obtained by Hennig for the photon energy corresponding to maximum transmission are given in table 20 for a number of donor and acceptor compounds. More recently, the free carrier contribution to e has been studied by the reectivity technique and results obtained by various workers for the reectivity minima for a number of donor and acceptor compounds are given in table 20. Except for the case of low stage compounds, the behaviour of the reectivity and transmission measurements can be modelled approximately by a single carrier Drude model where the plasma frequency for the polarization Eka is given by

108

M. S. Dresselhaus and G. Dresselhaus !2 p;a 4Ne2 4opt;a ; mopt;a "core;a "core;a 4:31

where m opt;a and opt;a are respectively the in-plane optical e ective mass and the real part of the complex optical conductivity, is the relaxation time, N is the carrier density and "core;a is the in-plane core dielectric constant. A similar expression can be written for the plasma frequency for the polarization Ekc by transcription of all subscripts a ! c. The quantities !p , "core and are found from a t to the lineshape of the plasma edge structure. If interband e ects can be neglected, the reectivity minimum and the transmission maximum occur near !p . In accordance with (4.31), the plasma frequency is upshifted with increasing intercalate concentration or decreasing stage as illustrated in gure 58 for graphiteRb for 1 n 3 (Gue rard et al. 1977a) and in gure 59 for graphiteSbCl5 for stages 2 n 5 (Eklund et al. 1981). As shown in table 20, for acceptor compounds with the intercalants AsF5 , SbF 5 and HNO 3 , similar !p values are achieved for compounds of the same stage index. Likewise for the donors K, Rb and Cs, similar reectivity curves are obtained for compounds of the same stage index. We further note that for both donor and acceptor compounds the absolute magnitude for the reectivity in the region ! !p increases with decreasing stage, while the magnitude of the reectivity at the reectivity minimum decreases with decreasing stage. For some compounds (for example, C6 Li), the low frequency reectivity approaches 100%. A comparison of the reectivity data for donors and acceptors shows that !p for donor compounds occurs at higher frequencies than for acceptor compounds of the same stage. For example, if !p for donors is a factor of 2 greater than for acceptors, this would suggest that the carrier density for donors is greater by a factor of 4. This observation is consistent with a signicantly higher carrier generation per intercalant ( f ) for donors than for acceptors. The relatively weak dependence of the reectivity minima on stage for certain acceptor compounds (such as graphiteSbCl5 in gure 59) suggests that interband e ects may contribute signicantly to the dielectric constant and may result in shifts of the plasma edge. Other acceptor compounds, however, such as graphiteAsF5 , do show signicant stage dependence of the plasma edge (Hanlon et al. 1977a). Because of the multiple pieces of Fermi surface which are present in graphite intercalation compounds, as suggested by studies of quantum oscillatory phenomena (section 4.5), a quantitative model for the optical reectivity should include free carrier contributions from multiple carrier pockets, as well as contributions from interband transitions. To obtain consistency between opt and dc , it has been necessary to introduce more than one carrier type, at least for the stage 1 and 2 alkali metal donor compounds (Fischer 1979). Recently, Blinowski et al. (1980) have interpreted optical reectivity measurements in terms of a model which explicitly contains einterband contributions (see section 4.1), leading to a more quantitative t than was previously available, to the plasma edge measurements for a number of acceptor compounds (see table 20). As optical measurements are extended to cover a wider frequency range, it will become increasingly important to include the einterband contribution when interpreting optical measurements. A simplied method for including interband contributions involves use of an oscillator t to reectivity data, and the solid curves in gure 59 represent such a t to the experimental data. As discussed in section 4.3.4, the plasma frequency !p;c for the polarization Ekc is much lower than !p;a , consistent with the much lower conductivity, opt;c opt;a .

Intercalation compounds of graphite

109

Figure 60. Progression of d ln R=dE spectra taken by Shieh et al. (1979) during HNO3 intercalation as the sample passes from stage 1 (i.e. graphite) to stage 2. The transitions labelled A and B are characteristic of pure graphite and are sensitive to EF .

Information on the free carrier plasma edge is also obtained using electron loss spectroscopy (section 4.2.2), and the results obtained with this technique are consistent with the optical measurements. The electron loss spectra also provide information on plasma e ects associated with all the graphite -electrons, and this plasma structure is found in the vicinity of 7.5 eV. In principle, such information is also available using optical techniques, but such optical studies have not been pursued in intercalated graphite. 4.3.3. Other optical structures In dilute donor and acceptor compounds (corresponding to stages n 4), infrared modulated reectivity studies exhibit a doublet structure in the vicinity of 0.8 eV, as shown in gure 60 (Shieh et al. 1979) for the graphiteHNO3 system. The similarity of this structure to the doublet structure observed in pristine graphite in the same photon energy range (Guizetti et al. 1973, Bellodi et al. 1975) provides further strong evidence that the electronic structure of intercalated graphite is graphitic in nature. A lineshape analysis of such data, following the work of Nakao (1979) and Misu et al. (1979) on pristine graphite, could yield valuable information on the weak stage dependence of the transition energies, or of the SWMcC band parameter 1 , denoting the nearest-neighbou r interlayer overlap energy. It is also signicant that the doublet structure is more pronounced for intercalated graphite than for pristine graphite, and this has been attributed to a large decrease in "2;a , the imaginary part of the in-plane dielectric constant with intercalation, even for as

110

M. S. Dresselhaus and G. Dresselhaus

dilute a compound as stage 9 graphiteHNO2 (Shieh et al. 1979). Such a large decrease in the interband contributions to "2;a would result from a downshift of the Fermi level below the K-point extremum for acceptors and above the H-point extremum for donors. When the Fermi level is within the band overlap region, the threshold for interband transitions is the spinorbit energy of 2:2 104 eV, so that very low frequency transitions dominate "2;a because of the (1=! 3 ) dependence of "2 . As the Fermi level passes out of the band overlap region, the threshold for interband transitions increases and the contribution of transitions between E3 bands to "2;a decreases corespondingly. The absence of the doublet structure for n 3 acceptors suggests that the observed phenomenon is associated with the overlap energy between atoms on two adjacent graphite interior layers. This interpretation is, however, at variance with the observation of the doublet structure in stage 3 donor compounds (Shieh et al. 1979). In addition, structure is observed in the vicinity of 0.55 eV (see gure 59) in seveal acceptor compounds for stages n 3 (Fischer et al. 1976, Hanlon et al. 1977a, b, Wachnik 1977, Shieh et al. 1979, Blinowski et al. 1980, Eklund et al. 1981). Since this structure is not observed in pristine graphite and is only weakly dependent on intercalant or stage (n 3), it could be associated with K-point transitions between graphite -bands one of which is signicantly perturbed by the intercalate graphite bounding layer interaction. A detailed analysis of this structure awaits the availability of band models for high stage compounds. Additional optical structure has also been reported at yet lower frequencies in SbF5 , SbCl5 and other acceptor compounds (Hanlon et al. 1977b, Blinowski et al. 1980, Eklund et al. 1981). Such structures are identied with intervalance band transitions, as has also been suggested by Blinowski et al. In this connection it should be noted that the analysis of Shubnikovde Haas data in the donor compounds with K and Rb suggests a 0.3 eV splitting of the E3 levels associated with the graphite bounding layer intercalate layer interaction (Dresselhaus and Leung 1980). Also, the two-dimensional model of Blinowski et al. (1980) predicts a band gap of 0.37 eV for n 2 compounds. No detailed calculations of the optical properties of graphite intercalation compounds have yet been carried out. One-electron energy band calculations have been carried out only for C8 K (Inoshita et al. 1977, Ohno et al. 1979) and C6 Li (Holzwarth et al. 1978b), but the optical properties implied by these band calculations have not yet been computed. A model for the interband contribution to the dielectric constant for acceptor compounds has, however, been proposed by Blinowski et al. (1980), based on a calculation of the electronic levels for a single sheet of graphite. The overlap with neighbouring planes was estimated from SWMcC band parameters and the complex in-plane dielectric function "! due to both interband and free carrier contributions was calculated and compared with experiment for stage 1 and 2 acceptor compounds (see also section 4.1.2). This model gives an improved t to the reectivity data relative to a simple Drude t and has been applied to several compounds as indicated in table 20. Few optical studies of intercalated graphite have been carried out for photon energies above the plasma edge where the dominant interband transitions are observed in pristine graphite. Avogadr o et al. (1977) measured the reectivity and thermoreectivity of a stage 3 graphiteHNO3 compound. They showed the dominant M-point -band transition, which exhibits a reectivity peak at 4.90 eV in pristine graphite, to be almost una ected by intercalation, having a peak

Intercalation compounds of graphite

111

reectivity at 5.15 eV and structure in the thermoreectivity peaking at 4.85 eV in stage 3 graphiteHNO 3 . Avogadro et al. also reported a broad semi-transparent behaviour in the frequency range between 2 and 4 eV for this compound. Optical reectivity measurements in the visible and near U.V. have also been reported by Puger et al. (1980). Their measurements were made on stages n 1, 2, 3 graphiteCs compounds, and their reectivity spectra exhibit peaks associated with the dominant M-point -band transitions in graphite, but downshifted by 0.2 eV in C36 Cs, and by 0.3 eV in C24 Cs. These authors also reported the absence of such a transition in the stage 1 compound, which is surprising in comparison with the energy loss results for C8 K by Hwang et al. (1980) which show an energy loss peak at 3.8 eV, as is discussed in section 4.4.2. However, in the absence of a KramersKronig analysis of the optical reectivity data, it would be premature to conclude that the behaviour of the dominant graphite -band transition in C8 Cs is signicantly di erent from that in C8 K. Electron energy loss spectroscopy studies (section 4.4.2) have been carried out on several intercalation compounds and show in agreement with the optical measurements described above that the interband transitions in the intercalation compounds are closely identied with transitions also occurring in pristine graphite, such as the -band transitions near 5 eV, the - transitions near 13 and 25 eV. Some information on transitions between graphitic and intercalate levels has also been provided by Eklund et al. (1978) for the graphitebromine system through the resonant Raman enhancement of intercalate vibrational modes for laser excitation energies close to the electronic transition energies, as discussed in section 5.3. Transitions between FeCl3 levels have been identied in the electron loss spectra from rst stage graphiteFeCl3 (Ritsko and Mele 1980a, b). These authors have also identied a transition from an iron 3p core level to an empty 3d state at the Fermi level in the same compound, again using electron energy loss spectroscopy (sections 4.4.2 and 4.4.4). 4.3.4. Optical anisotropy The optical properties for the polarization Ekc axis in both pristine graphite and intercalated graphite are extremely di cult to measure quantitatively because of the experimental problem of preparing good optical quality `a-faces, containing both aand c-axes. In contrast, high quality optical surfaces are easily prepared by cleavage on `c-faces, containing only a-axes. The most reliable `a-face samples that have been prepared to date are mechanically polished and then sputter-etched with argon ions (Zanini et al. 1978b). The mechanical polish produces improved optical surfaces but degrades the crystal ordering, while the argon ion sputtering removes much of the crystal damage but introduces a pitted surface with lower reectivity. It is further shown that the measured value of the dielectric tensor component "c for Ekc, is very sensitive to surface treatment. Values of "c 3:4 0:4 have been obtained by reectivity measurements on sputter-etched `a-faces by Zanini et al. (1978b) and "c 5:8 was obtained by Nemanich et al. (1977b) from mechanically polished surfaces. These values are to be compared with "c 5:4 reported by Venghaus (1977) obtained by energy loss experiments through thin `c-face samples. Optical reectivity studies have been carried out on `a-faces for both graphite (Ergun et al. 1967, Greenaway et al. 1969, Nemanish et al. 1977b) and for intercalated graphite (Zanini and Fischer 1977). The optical anisotropy can also

112

M. S. Dresselhaus and G. Dresselhaus

Figure 61. Polarized reectance spectra of rst stage C8 K and C8 Cs, taken by Zanini and Fischer (1977) with light incident on a sputter-etched `a-face prepared from HOPG. The reectance data show little dependence on alkali metal species (K or Cs), but a strong dependence on polarization geometry, with !p;a !p;c .

be probed by the electron energy loss technique (section 4.4.2) which measures either Im 1="a q or Im 1="c q. Theoretical considerations suggest that the free carrier contribution to the dielectric tensor ecarriers for Ekc is smaller by more than one order of magnitude than for Eka, primarily because the e ective mass component mzz is about two orders of magnitude greater than mxx in graphite. Thus !p;c , the plasma frequency for Ekc, is expected to be much smaller than !p;a for Eka !2 4Ne 2 =m "c !2 : p;a zz p;c 4:32

An anisotropy in the interband contribution to the dielectric tensor is also expected, since einterband depends on the momentum matrix elements jhjpz jij2 for Ekc and jhjpx jij2 for Eka. On the basis of k p perturbation theory, these matrix elements are related to the e ective mass components by the relation jhjpx jij2 jhjpz jij
2

mzz : mxx

4:33

The calculation of the optical anisotropy of graphite by Johnson and Dresselhaus (1973) based on a three-dimensonal electronic energy band structure shows that the dominant structures in "c and "a occur at approximatel y the same photon energies, primarily because interband transitions occur over large volumes of the Brillouin zone. Thus for graphite, interband contributions to both "a and "c are primarily sensitive to the joint density of states. However, the structure for the Ekc polarization tends to be about one order of magnitude smaller relative to that for the Eka polarization, reecting the anisotropy in the matrix elements and polarization-dependent selection rules. These aspects of the JohnsonDresselhaus calculation are in good agreement with both optical reectivity (Greenaway et al. 1969) and electron energy loss (Buchner 1977) measurements on graphite. Two-dimensional models such as those by Bassani and Parravicini (1967) and Painter and Ellis (1970), have frequently been used to discuss optical properties of graphite, though twodimensional models are not as useful for treating the optical anisotropy as are threedimensional models.

Intercalation compounds of graphite

113

The e ect of intercalation on the optical anisotropy has not yet been fully established. The above theoretical considerations suggest that for donor compounds a decrease in the optical anisotropy of the e ective mass components should result in a decrease in the optical anisotropy and likewise a decrease in the electrical anisotropy. On the other hand, in the case of acceptor intercalation compounds where intercalation increases the electrical anisotropy and the materials become more two dimensional, the optical anisotropy should also increase. Highly oriented pyrolytic graphite (HOPG) shows a high optical anisotropy with high transmission for Ekc, but high reection for Eka. Far-infrared polarizers based on this high anisotropy have in fact been built (Rupprecht et al. 1962). An increase in optical anisotropy through intercalation could perhaps result in practical device applications. Optical reectivity experiments have been performed by Zanini and Fischer (1977) on `a-face samples of C8 K and C8 Cs that were mechanically polished and sputter-etched prior to intercalation (Zanini et al. 1978b). Some preliminary measurements were also made on stage 2 samples of graphiteK. The results on the stage 1 donor compounds (see gure 61) exhibit Drude-like metallic reectivity for both polarizations Eka and Ekc. The reectivity curve for the Eka polarization on the `aface sample has a Drudge edge similar to that for the `c-face sample with regard to the reectivity minimum, but a lower reectivity than the `c-face sample below 2 eV. The Drude edge for the Ekc polarization is, however, shifted to lower energies, consistent with the lower electrical conductivity for Ekc and the larger e ective mass component mzz relative to mxx . The results for the Ekc spectra for C8 K and C8 Cs are similar, but small di erences are found, in so far as the reectivity edge for C8 Cs is at slightly lower energy than for C8 K. The lineshape analyses of the reectivity curves for both polarizations were treated in terms of a Drude model. A single carrier type was assumed for the c-axis properties, and two carrier types for the in-plane properties, one carrier type corresponding to a very high mass anisotropy ratio (denoted as 2D carriers) and the other carriers corresponding to a lower mass anisotropy ratio (denoted as 3D carriers). Such a delineation of carrier types is suggested by the Fermi surface model for C8 K by Inoshita et al. (1977) where the three-dimensional carriers are the ones associated with the carrier pockets about the - and A-points of gure 47. The results of the analysis by Zanini and Fischer yield an e ective mass anisotropy for their 3D carriers of mzz =mxx 8 for C8 K, which is to be compared with the mass anisotropy ratio of 102 for pristine graphite. From their analysis Zanini and Fischer conclude that the highly anisotropic 2D carriers located about the HKH axis make the dominant contributions to the in-plane electrical conductivity, while the more spherical 3D carriers dominate the c-axis conductivity. Their analysis also gave rough agreement between the anisotropy ratio measured in the d.c. conductivity, and inferred for the optical conductivity from the reectivity data. Reectivity measurements with Ekc for a stage 2 graphiteK sample did not reveal a metallic plasma edge above 0.5 eV, suggesting a larger mass anisotropy and a much higher electrical anisotropy in C24 K relative to C8 K (Zanini and Fischer 1977). On the basis of their optical anisotropy measurements, these authors also concluded that the concentration of 3D carriers is signicantly reduced in C24 K relative to C8 K. Further work using both the optical reectivity and the electron energy loss technique is needed to elucidate the optical anisotropy of graphite intercalation compounds and to relate such data to c-axis conductivity measurements on the same materials.

114

M. S. Dresselhaus and G. Dresselhaus

Figure 62. The electron energy loss function Im 1="!; q of C8 K derived from the electron energy loss spectra with momentum transfers q between 0.13 and 0.66 A1 . The arrows indicate the positions of the -electron plasma calculated from the quadratic momentum dispersion relation of (4.34) with 0:43. (From the work of Hwang et al. 1980.)

4.4. Electron spectroscopy studies 4.4.1. Introductor y comments The use of electron spectroscopy to study graphite intercalation compounds has covered a wide range of experimental techniques and has yielded a great deal of information about the electronic structure of these materials. In addition, high energy elastic electron scattering experiments, as carried out in a transmission electron microscope yield important information on the in-plane structure, as discussed in sections 3.2 and 3.4. In this section, we focus on electron spectroscopy used as a tool to study the electronic structure of the intercalation compounds, often yielding information otherwise unavailable. With electron spectroscopy techniques it is possible to probe the electronic structure over a wide energy range and also in great detail for the energy bands near the Fermi level. These studies provide strong conrmation that the energy levels in the intercalation compounds are closely related to the energy levels in the parent graphite and intercalant materials. These electron spectroscopy studies further determine the shifts in energy levels and Fermi energy associated with intercalation. 4.4.2. Electron energy loss spectroscopy Low energy electron energy loss spectroscopy has provided an important technique for studying the electronic structure of graphite (Venghaus 1975, Buchner 1977). This technique measures the dielectric response function e!; q for a range of q vectors which can span the Brillouin zone, while optical measurements yield the dielectric response function only for q 0, because of the small magnitude of the wave vector of light relative to Brillouin zone dimensions. Electron energy loss spectroscopy has been applied successfully to measure the dielectric response function "a !; q for intercalated graphite, including the stage 1 donor compound C8 K (Hwang et al. 1979, 1980) and for the stage 1 and 2 graphiteFeCl3 acceptor

Intercalation compounds of graphite

115

compounds (Ritsko and Rice 1979, Ritsko and Mele 1979, 1980a, b). Preliminary results have also been reported for second stage C24 K by Hwang et al. (1980b). Electron loss structure associated with both interband transitions and plasma resonances have been identied. Results for the electron energy loss function Im 1="a !; q obtained by Hwang et al. (1980) for C8 K are shown in gure 62 for a range of momentum transfers 0:13 < q < 0:66 A1 . In C8 K the energy loss peak at 2.5 eV associated with the free carrier plasma oscillations is in good agreement with the energy of the plasma edge in the optical reectivity, as discussed in section 4.3. Ritsko and Mele (1980b) also reported structure at 1:2 0:2 eV in the stage 1 FeCl3 compound and attributed this structure to interband transitions between graphite -bands. However, as discussed in section 4.3.2, the plasma edge for stage 1 acceptor compounds is generally observed in the optical reectivity in the vicinity of 1.2 eV. It is of course quite possible that interband transitions may also occur in this photon energy range. The peak in the dielectric function associated with the -electron plasma resonance is found in pristine graphite at 7.5 eV, but downshifted in C8 K to 6.3 eV and in the graphiteFeCl3 system to 5.9 eV for stage 1 and 6.25 eV for stage 2. However the plasma dispersion relation in the intercalation compound C8 K, Epq Ep 0 2 q2 =m; h 4:34

is the same as in graphite ( 0:42) to within experimental error (see gure 62). From a t to their measured plasmon dispersion curve, Ritsko and Mele (1980b) obtained a charge transfer of 0.10 charge per intercalant, for both stage 1 and stage 2 compounds, based on a chemical formula C6:6 FeCl3 and C12:5 FeCl3 for these compounds. The dominant interband M-point -band transition at 4.9 eV in graphite is downshifted to 3.8 eV in C8 K and to 4.3 eV for the rst stage FeCl3 compound. This experimental value of 3.8 eV for C8 K has been compared with the calculated values for the -band M-point transition of 4.8 eV by Ohno et al. (1979). In the case of the experiments, the contribution to the peak is expected to occur for a signicant region of the Brillouin zone, so that an identication of such peaks with specic points in the Brillouin zone is only approximate. The interband transition between graphite -bands near 13 eV is unshifted in C8 K, though somewhat broadened, and in the stage 1 FeCl3 compound extends from 13 to 15 eV. In addition, a doublet structure at 22 and 28 eV is found in C8 K near the 25 eV transition observed in graphite. This doublet structure has been identied with zone-folding e ects which map the M-point into the zone centre. For the case of the graphiteFeCl3 compound, electron loss peaks at 2.75, 3.9, 8.0 and 9.0 eV have been identied with transitions between FeCl3 levels, since similar peaks occur in pristine FeCl3 at exactly these energies. The main conclusions reached by these workers is that the graphite and bands in the intercalation compounds are well described by a rigid band model. By studying transitions with signicant electron momentum transfer, Fermi level shifts of 0.9, 0.7 and 1 eV relative to the graphite Fermi level were reported for the stage 1 and 2 graphiteFeCl3 and for C8 K respectively. The measurement of "c ! by electron loss transmission experiments through thin samples has not yet been exploited in graphite intercalation compounds. Venghaus (1975) and others (Daniels et al. 1970) have shown that the most accurate measurements of "c ! in highly anisotropic materials such as pristine graphite are obtained by electron loss transmission experiments.

116

M. S. Dresselhaus and G. Dresselhaus

4.4.3. Photoemission studies Angle-resolved and angle-integrate d photoemission studies have both been reported in graphite intercalation compounds. The angle-resolved technique which provides a highly sensitive determination of the energy dispersion relations Ek has been applied to C6 Li by Eberhardt et al. (1980) and an accurate determination of the energies of the -point valence bands has been made and these results are given in table 15. Some preliminary results for Ek in C8 Rb have also been reported by McGovern et al. (1980) and their reported -point energies are also included in table 15. The results for C6 Li are in excellent agreement with the rst principles calculation of Holzwarth et al. (1977, 1978b) and conrm directly the zone-folding approach used in band calculations for intercalated graphite (see section 4.1). The absence of any structure in the photoemission spectra near EF is consistent with the ndings of the calculations that the Li 2s-band is unoccupied. Measurement of the energy of the Li 1s core level yields 56.8 eV as compared with 54.86 eV in Li metal and 57.6 eV in LiF, thereby providing strong support to the interpretation that the Li in the intercalation compound is closer to a free Li ion conguration than to a metallic Li ion embedded in a free electron gas plasma. The departure from the free ion 1s Li core level in LiF could be due to some overlap of the Li ions in the intercalant layer with the electron plasma on the graphite bounding layers (see gure 28(b)). Of signicance to the interpretation of experiments relevant to interband transitions and Fermi level shifts are the unequal band shifts of the various graphite and levels in C6 Li relative to their values in graphite (see table 15). It should also be noted that in C6 Li, the K-point degenerate 1 level which is at EF in pristine graphite, lies 0.5 eV below the Fermi level in C6 Li. This result is in disagreement with the band calculation of Holzwarth et al. (1977, 1978b) which places the 1 level 1.3 eV below EF . This disagreement may imply that the calculated Fermi surface shown in gure 44 could di er signicantly from the actual Fermi surface. In this connection it is of interest to note that McGovern et al. (1980) found with the angle-resolved photoemission technique that the zone-folded 1 band in C8 Rb at the point is 0:4 0:2 eV below the Fermi level, very close to the value in C6 Li. The density of states for rst stage C8 K, C8 Rb and C8 Cs as well as higher stage Cs compounds has been probed by the angle-integrate d photoemission studies of Oelhafen et al. (1980) , following a number of previous studies on pristine graphite (McFeeley et al. 1974, Willis et al. 1974). The two lowest energy peaks observed in the photoemission spectrum of pristine graphite a 3.0 and 4.7 (0.1) eV have also been observed by Oelhafen et al. in the stage 1 alkali metal compounds. The density of states spectra NE are very similar for the three compounds C8 K, C8 Rb and C8 Cs, including a high density of states close to the Fermi level which is identied with the alkali metal s-band. Shifts in Fermi level in the intercalation compounds relative to the Fermi level in graphite are reported, and the values given for the Fermi level shifts for C8 Cs, C8 K and C8 Rb are respectively 1.0, 1.4 and 1.8 eV. Oelhafen et al. (1980) also report that the dependence of EF on stage index is very weak for the graphiteCs compounds (1 eV for stage n 1 and 0.7 eV for n 7), but this result is inconsistent with other determinations of the stage dependence of the shift in Fermi level (see sections 4.5 and 4.6). The photoemission experiments of Oelhafen et al. (1980) also measure the energy of the minimum of the 1 alkali metal band relative to the Fermi level and values of 1.15, 1.2 and 1.25 eV are reported for C8 Cs, C8 Rb and C8 K respectively. The value of 1.25 eV is to be compared with the values of 1.23 and 0.9 eV calculated by Inoshita et al. (1977) and Ohno et al. (1979) respectively, for

Intercalation compounds of graphite

117

the case of C8 K. It is of interest to note that in the pristine alkali metal these band minima lie 1.5, 1.8 and 2.0 eV below EF respectively, indicating that the alkali metal s-band is only partially occupied for each of these intercalation compounds. If the 3=2 carrier density associated with the alkali metal s-band is assumed to have an EF dependence, the fractional intercalant ionization for C8 Cs, C8 Rb and C8 K is found to be 0.33, 0.46 and 0.50 respectively. Values are also reported for the bandwidths of the alkali metal s-bands for various Cs compounds, 1.75, 0.95, 0.73 and 0.45 eV respectively for stage 1, 2, 3 and 4 compounds. Except for the stage 1 compound, very good agreement is obtained by the authors with estimates made on the basis of a free electron model for the alkali metal s-band in the intercalation compound. From analysis of their density of states spectra and assuming a charge transfer of 0.3 electrons/alkali metal atom, Oelhafen et al. obtained a value of 0.19 states/(eV C atom) for the density of states at the Fermi level for C8 Cs and 0.17 states/(eV C atom) for C8 K, which they compared respectively to 0.23 and 0.24 states/(eV C atom) as deduced from low temperature specic heat measurements for these compounds (Mizutani et al. 1978). These density of states values correspond to an enhancement in the density of states at the Fermi level by a factor of 30 relative to that in pristine graphite. These authors further obtained an enhancement by a factor of 5 for the density of states of the graphite -bands at the Fermi level in these intercalation compounds relative to that in graphite. Also using the photoemission technique, Bach et al. (1971) obtained a density of states spectrum for graphiteBr2 acceptor compounds, showing a narrowing of the graphite density of states, the narrowing e ect becoming more pronounced with increasing Br2 concentration. This narrowing e ect was interpreted in terms of a transfer of charge out of the graphite layers into the bromine intercalant layers. This interpretation was supported by the appearance of a peak in the density of states above the graphite peak identied with the bromine 4p level and a shoulder below the graphite peak identied with the 4s bromine level. The intensity of the structures associated with the bromine levels was found to increase with increasing bromine concentration, consistent with this interpretation. No estimates of the amount of charge transfer of the bromine to the graphite layers were made. 4.4.4. Spectroscopy of core levels Core level shifts give valuable information on the electronic environment of atoms in solids and can be probed by such techniques as ESCA (Electron Spectroscopy for Chemical Analysis), electron energy loss spectroscopy and Auger spectroscopy. The most detailed information is obtained in cases where the experimental measurements of shifts are made with respect to reference materials (for example, measurement of the 1s Li core level in C6 Li as compared with LiF, or for a given intercalant the core level in a stage n and in a stage n 1 compound). This type of accurate comparison has been attempted by Bach (1971) on the graphiteK donor compounds using the ESCA technique. On the basis of these measurements he conrmed that electrons are transferred from the potassium to the graphite layers. The ESCA studies on acceptors by Schlogl and Boehm (1979) have focused on the determination of the ionic complex that is present in the intercalate layer for the intercalants MoCl5 , NiCl2 , CuCl2 , SbF 5 and FeCl3 . A doublet structure with about 2 eV separation was characteristically found in the photoelectron spectra of the core levels for these materials. This doublet structure was interpreted in a preliminary

118

M. S. Dresselhaus and G. Dresselhaus

report in terms of the existence of two di erent types of carbon sites, one type that is bonded to the intercalant and the other type that is not bonded. The energy loss spectra of pure graphite show a broad structure near 287 eV associated with transitions from the 1s carbon core states to the empty carbon 2p state at the Fermi level. Experiments carried out in stage 1 graphiteFeCl3 by Mele and Ritsko (1979) show this electron loss peak to shift because of the lowering of the Fermi level due to intercalation. Their lineshape analysis of the energy loss spectra for the intercalation compound yielded a shift in Fermi energy of 0.9 eV relative to graphite, in agreement with their electron loss measurements below 40 eV. Ritsko and Mele (1980a) also identied in stage 1 graphiteFeCl3 a small peak at 56.7 eV in the energy loss spectra with transitions from a Fe 3p core state to empty 3d states at the Fermi level in the intercalation compound. A study of the Auger spectrum corresponding to the ejection of an electron from a 1s carbon level, the lling of this level by a valence state and the simultaneous emission from a valence state has been carried out in rst stage C8 Rb and C8 Cs by Oelhafen et al. (1979). The Auger spectra for the intercalation compounds show an additional sharp structure not present in the graphite spectra. The sharp structure is identied with the alkali metal 1 band (see gure 46), consistent with a linewidth that is twice the width of the density of states for the 1 band. This twofold increase in the Auger linewidth relative to the density of states width occurs because the electron falling down to the core level and the ejected electron can each originate from anywhere in the occupied 1 band. Some additional shoulders on the Auger structure associated with graphite bands could be due to new graphite -band valence states formed by the upshift of the Fermi level through intercalation. The application of electron spectroscopy to the study of the electronic structure points out the importance of using diverse experimental techniques for the determination of a consistent model for the electronic structure of graphite intercalation compounds.

4.5. Fermi surface studies 4.5.1. Quantum oscillatory phenomena Since quantum oscillatory phenomena yield Fermi surface cross-sectional areas directly, this technique has been especially valuable in providing the density of carriers of a given type, the number of equivalent carrier pockets in the Brillouin zone, and in some cases also the location of the carrier pockets within the Brillouin zone. Quantum oscillatory studies yield values for the extremal Fermi surface crosssectional areas perpendicular to the magnetic eld. To map out the Fermi surface, the direction of the magnetic eld is varied with respect to the crystallographi c axis. Because of the threefold symmetry of the graphite Fermi surface with respect to the c-axis, anisotropy studies are carried out as a function of , the angle between the magnetic eld H and the c-axis, thereby yielding the geometrical shape and volume of each carrier pocket. Quantum oscillatory phenomena can be observed in the magnetic susceptibility (de Haasvan Alphen or DHVA e ect), the electrical conductivity (Shubnikovde Haas or SdH e ect), the temperature of an adiabatically isolated sample (magnetotherma l oscillations), the Hall e ect, the ultrasonic attenuation and other phenomena. For the case of the Shubnikovde Haas e ect, the oscillatory term in the longitudinal magnetoconductivit y in the acoustical phonon scattering regime is given by (Kahn and Frederikse 1959):

Intercalation compounds of graphite

119

Figure 63. Fermi surface model for graphite illustrating the three types of extremal constant energy orbits: majority electrons, majority holes and minority holes.

X kT 2!i k H h 2 h 0 !i EF i

1=2 X 1
r1

1r r1=2 cos

EF 2r i 2r h !i exp ; sinh 22 rkT =!i h !i 4:35

where the sum is over the rth harmonic of the ith carrier, EF is the Fermi energy, !i eH=m c is the cyclotron frequency at the Fermi surface extremum for the ith i carrier, and i is a phase factor for the ith carrier. The oscillatory cosine term gives rise to a phase change of 2 whenever the Fermi level crosses a magnetic energy subband extremum. Equivalently, the periodicity of the cosine term is expressed in terms of a characteristic de Haasvan Alphen (DHVA) frequency equal to ch=eAi , where Ai is the extremal Fermi surface cross-sectional area. The e ect of collisions results in a reduction of the oscillatory conductivity through the nite time between collisions h which is expressed in terms of the Dingle temperature TD 2=k (Kahn and Frederikse 1959). In addition, the e ective mass associated with each extremal Fermi surface cross-sectional area can be found through measurement of the temperature dependence of the amplitude of the quantum oscillations. Thus quantum oscillatory phenomena also permit measurement of the angular dependence of the e ective mass tensor. 4.5.2. The Fermi surface for graphite Graphite is one of the classical semimetals that has been a prototype material for Fermi surface studies. A large variety of quantum oscillatory phenomena has been observed in graphite and the results are summarized in the review articles by McClure (1971) and Spain (1973) . Graphite is also a material for which the oneelectron energy band structure is well understood (section 4.1). Measurements that have been of particular importance to the determination of the one-electron energy band model are (DHVA) frequencies for majority electron, majority and minority hole carrier pockets, the magnetoreection K- and H-point interband transitions and the modulated infrared reectivity spectrum.

120

M. S. Dresselhaus and G. Dresselhaus

Figure 64. Shubnikovde Haas oscillations in a fourth stage graphiteK compound as a function of magnetic eld for several values of temperature (from Suematsu et al. 1980a).

The Fermi surface of pure graphite consists of two sets of three elongated, trigonally warped hole and electron carrier pockets (see gure 63). These carrier pockets have threefold symmetry with respect to the vertical edges HKH and H 0 K 0 H 0 of the hexagonal graphite Brillouin zone. The Fermi surfaces are highly anisotropic, with a length in the c-direction about 13 times the width perpendicular to that direction. Thus to within experimental error, the dependence of the experimentally determined cross-sectional areas cannot be distinguished easily from the cos dependence, characteristic of a cylindrical Fermi surface and a twodimensional solid, except for measurements in the range 908 808. The main Fermi surface results for graphite are expressed in terms of the de Haasvan Alpen frequencies for Hkc axis, which are 6.5, 4.9 and 0.33 T for the majority electron, majority hole and minority hole surfaces respectively (Soule et al. 1964, Soule 1964, Williamson et al. 1965, Woollam 1970, Cooper et al. 1970, Toy et al. 1977). The minority hole frequency is associated with the intersection of the hole Fermi surface with the Brillouin zone boundary at =2c0 . Several interpretations of the quantum oscillatory phenomena in graphite intercalation compounds have been made in terms of graphitic constant energy surfaces. 4.5.3. Quantum oscillatory studies of graphite intercalation compounds The rst quantum oscillatory phenomena to be reported on graphite intercalation compounds were on the graphitebromine system using the Shubnikovde Haas e ect in the transverse magnetoresistanc e (Bender and Young 1971, 1972). This work demonstrate d that graphite intercalation compounds could be prepared with su cient perfection to satisfy the requirement for the observation of quantum oscillatory phenomena, !c 1, which states that an electron orbit in a magnetic eld must be completed before scattering. This work further showed that di erent de Haasvan Alphen (DHVA) frequencies occur in intercalated graphite relative to pristine graphite. The new DHVA frequencies were typically found to be more than an order of magnitude higher than graphite DHVA frequencies, and to show

Intercalation compounds of graphite


Table 21. DHVA frequencies for various stage graphiteK compounds for Hkc-axis. Oberved Extremal crossDHVA frequency sectional area (tesla) (cm2 ) 0.33 4.88 6.49 2870 133 149 282 306 439 146 260 121 144 238 264 338 24 135 152 191 243 267 290 453 0:31 10 12 4:67 10 12 6:21 10 12 1:26 10 1:42 10 14 2:69 10 14 2:92 10 14 4:19 10 14 2:76 10 15
14

121

Stage Graphite (1) 1 2

Authors/comments Minority hole Majority hole Majority electron Tanuma et al. (1978) and Higuchi et al. (1980) Dresselhaus et al. (1980)

3 4

0:23 10 14 1:28 10 14 1:45 10 14 1:82 10 14 2:32 10 14 2:55 10 14 2:77 10 14 4:32 10 14

1:16 10 14 1:38 10 14 2:28 10 14 2:53 10 14 3:24 10 14

1:40 10 14 2:50 10 14

Tanuma et al. (1978) and Higuchi et al. (1980) Tanuma et al. (1978) and Higuchi et al. (1980)

Dresselhaus et al. (1980)

Table 22. DHVA frequencies (in tesla) observed in dilute graphiteBr 2 compounds where no stage dependence was found. Results are given for magnetothermal oscillations (MTO) by Rosenman et al. (1979), Shubnikovde Haas (SDH) by Bender and Young (1972) and de Haasvan Alphen (DHVA) by Tanuma et al. (1978). The orbit identication is given by Rosenman et al. (1979). Orbit identication 2 2 3 2 " MTO (tesla) 14.5 29.3 61.5 120.5 183.5 78.4 159.5 100.5 414.5 480.5 545.5 SDH (tesla) 19.1 61.5 79.6 DHVA (tesla) 23 50 66 135 85 121 470.5

122

M. S. Dresselhaus and G. Dresselhaus

approximately a cos angular dependence, consistent with the high anisotropy of the majority electron and hole Fermi surfaces of graphite. This initial work was carried out on dilute compounds that were not well characterized for stage. Nevertheless, this work provided an important stimulus to high magnetic eld studies of graphite intercalation compounds. With improved sample preparation techniques, it has become possible to prepare a large number of graphite intercalation compounds satisfying the condition !c 1, so that a number of donor and acceptor compounds have been studied by various quantum oscillatory techniques, including de Haasvan Alphen, Shubnikovde Haas and magnetothermal oscillations. Of the various quantum oscillatory techniques, the Shubnikovde Haas requires the simplest experimental setup. An example of results obtained by this technique is given in gure 64 for a stage 4 graphiteK sample at several temperatures (Suematsu et al. 1980a). The magnetotherma l and de Haasvan Alphen techniques have the advantage over the Shubnikovde Haas technique in requiring no electrical leads. To obtain reproducible results on the same sample from run to run and on di erent samples of the same nominal stage and intercalant, it appears to be necessary to make measurements on as-grown, encapsulated samples. A number of conclusions can be reached from the results of the various quantum oscillatory phenomena that have been studied to date, and these are discussed in some detail below. These conclusions relate to the number of DHVA frequencies that are observed, the dependence of this number on stage, the magnitudes of the DHVA frequencies and the relation of these magnitudes to observed transport properties, the angular dependence of the DHVA frequencies and the shapes of the Fermi surfaces, the e ective masses of specic carrier pockets and carrier mobilities. The experimental results that have been reported have emphasized either stage-dependent DHVA frequencies or stage-independent frequencies, which have quite di erent implications on the Fermi surfaces for graphite intercalation compounds. Two types of interpretations have thus been developed to explain the DHVA results, one of which is motivated by the stage-dependent DHVA frequencies, the other by stageindependent DHVA frequencies. These two types of interpretation s are discussed below. A listing of DHVA frequencies for several graphiteK compounds reported by Tanuma et al. (1978) and Higuchi et al. (1980) is given in table 21 as an illustration of stage-dependent DHVA frequencies. These results obtained by the DHVA technique are in good agreement with SdH data (see table 21) by Dresselhaus et al. (1980) on graphiteK compounds. Stage-dependen t DHVA frequencies have also been reported for the donor intercalant Rb by Dresselhaus et al. (1980), and for the acceptor intercalants AsF5 by Iye et al. (1980) and by Markiewicz et al. (1980a), and HNO3 by Batallan et al. (1980). In contrast, table 22 gives DHVA frequencies for the intercalant Br2 for which stage-independen t DHVA frequencies have been reported by several workers. In all cases the measurements were carried out on dilute Br2 compounds that were not well characterized for stage, though di erent quantum oscillatory techniques were used: Shubnikovde Haas measurements by Bender and Young (1972), de Haasvan Alphen measurements by Tanuma et al. (1978) and magnetotherma l measurements by Rosenman et al. (1979). Other intercalants for which stage-independen t DHVA frequencies were reported include SbCl5 by Batallan et al. (1978) and FeCl3 by Woollam et al. (1979). From the various measurements that have been reported, a number of statements can nevertheless be made.

Intercalation compounds of graphite

123

The number of observed DHVA frequencies for a given sample is in most cases considerably greater than for graphite. For the workers who nd stage-dependen t Fermi surface e ects, the low stage compounds (n 1; 2) tend to exhibit a small number of DHVA frequencies, with increasing numbers of frequencies observed as the stage index is increased. A large number of DHVA frequencies is also found by workers reporting stage-independen t DHVA frequencies, though in this case, many of the reported frequencies are identied as harmonics. Many of the observed DHVA frequencies in both donor and acceptor compounds are greater by several orders of magnitude than those in pristine graphite. This observation is interpreted as a large increase in the Fermi surface cross-sectional areas resulting from the transfer of charge from the intercalant to the graphite layers. The large increase in carrier density thus obtained is consistent with the large increase observed in the low frequency electrical conductivity (section 4.2.1) and the observation of metallic plasma edges in the optical reectivity (section 4.3.2). The angular dependence of the DHVA frequencies has been measured for a number of donor and acceptor compounds, including the following intercalants: K (Tanuma et al. 1978, Higuchi et al. 1980, Dresselhaus et al. 1980), Rb (Dresselhaus et al. 1980), Br2 (Rosenman et al. 1979), FeCl3 and PdCl2 (Woollam et al. 1979), AsF 5 (Markiewicz et al. 1980, Iye et al. 1980). For the range of that was explored, the results in almost all cases are consistent with a cos behaviour, characteristic of a cylindrical Fermi surface. Since accurate measurements have not in most cases been reported out to large angles ( 908), the anisotropy data do not distinguish between an open cylindrical two-dimensional Fermi surface, or the highly elongated ellipsoid-like closed Fermi surfaces that occur in pristine graphite. From the work of Dresselhaus et al. (1980) and Shayegan (1981) where some of the lower DHVA frequencies were followed out to large angles, there is strong evidence that some of the DHVA frequencies correspond to closed Fermi surfaces. There is also evidence that other DHVA frequencies correspond to open Fermi surfaces. For example, the anisotropy of the magnetoresistanc e has been measured out to 908 in stage 1, 3 and 4 graphiteK compounds by Suematsu et al. (1980a) as discussed in section 4.2.5. In this work, the cusp-like behaviour observed near 908 for the stage 1 and 3 compounds has been interpreted in terms of open orbits associated with c-axis zonefolding e ects. The measurements for stage 4 graphiteK, however, show no clear cusp-like behaviour, which is not understood in relation to the observations on the lower stage compounds. By measuring the temperature dependence of the amplitude of the quantum oscillations, e ective masses associated with a few of the DHVA frequencies have been determined. For some of the DHVA frequencies very small cyclotron e ective masses have been observed, comparable to e ective masses in pristine graphite, while for other electron and hole orbits, large e ective masses have been reported. Cases where small e ective masses have been found include m =m0 0:075 for the orbit in C48 K by Suematsu et al. (1980b). Woolam et al. (1979) have also reported some small e ective masses in the graphiteFeCl3 system. Thus, small masses have been found for both donor and acceptor compounds. In contrast, higher e ective masses in the range 0:1 < m =m0 < 1:2 have been reported for graphiteHNO3 compounds by Batallan et al. (1980). Because of the wide range of e ective mass values that have been reported, systematic measurements of the masses associated with each of the DHVA frequencies are needed for application to the transport properties.

124

M. S. Dresselhaus and G. Dresselhaus

K Figure 65. Fermi energy EF relative to band extremum at the K-point versus reciprocal stage for graphiteK compounds. The open triangles are from the Shubnikovde Haas measurements of Dresselhaus et al. (1980), the closed triangles from the de Haasvan Alphen measurements of Higuchi et al. (1980), and the open triangle at stage 9 from the magnetoreection measurements of Mendez et al. (1980a). The 2=3 K curves are a t of the data to the functional form EF 1=n .

The existence of carriers with long relaxation times and high mobilities is well documented in the study of quantum oscillatory phenomena. For example, quantum oscillations have been observed for temperatures as high as 51 K in stage 3 graphite PdCl2 by Woollam et al. (1979). This high carrier mobility is also conrmed by low temperature transport measurements of the transverse magnetoresistanc e in both donor and acceptor compounds (see section 4.2.5). There is presently a large amount of experimental data being accumulated for quantum oscillatory phenomena in graphite intercalation compounds. To explain the observed quantum oscillatory e ects three types of theoretical models have been used. (1) Fermi surface models deduced directly from band calculations. Such band calculations are available only for C6 Li (Holzwarth et al. 1977, 1978b) and C8 K (Inoshita et al. 1977 and Ohno et al. 1979). (2) Tight binding models based on the graphite host structure and c-axis zone folding e ects have been proposed by Tanuma et al. (1978) and Dresselhaus et al. (1980). (3) Nearly free electron models based on a Harrison-type construction and in-plane zone folding have been proposed for several acceptor compounds by Batallan et al. (1978) and Bok (1978). A discussion of these models with regard to the experimental observations is presented below. The application of the rst principles calculations is possible only for the measurements on C8 K, where as shown below, good agreement is obtained between experiment and the theoretical model. For C8 K a single DHVA frequency of 2875 tesla has been identied experimentally. On the basis of the Fermi surface for C8 K calculated by Inoshita et al. (1977) and shown in gure 47, this high DHVA frequency is identied with the large cylindrical pieces of Fermi surfaces located about the edges HKH of the Brillouin zone, which have a calculated DHVA

Intercalation compounds of graphite

125

frequency of 3030 tesla, in good agreement with the experimental value. Although the Fermi surface calculation explains this high DHVA frequency quantitatively, no DHVA frequencies have been observed corresponding to the more spherical (threedimensional) pieces of Fermi surface given by the calculation about the - and Apoints in the Brillouin zone. A model that has been proposed to account for the DHVA frequencies observed for the higher stage graphiteK compounds and also applied to the acceptor graphiteAsF5 compounds emphasizes the stage dependence of the DHVA frequencies reported for these intercalants. This model, rst introduced by Tanuma et al. (1978), assumes that the charge released from the intercalate layer is uniformly distributed in the graphite layers. Using a rigid band model, the Fermi surfaces for the -electrons in the intercalation compounds are related to constant energy surfaces for -bands in graphite. Zone folding along the c-axis is introduced to account for the larger real space unit cell for the intercalation compounds. A specic application of this model has been made to the case of stage 4 graphiteK as shown in gure 42. In this case the model shown predicts ve DHVA frequencies, and ve main frequencies have been identied experimentally. However, only qualitative agreement was obtained between the measured and calculated cross-sectional Fermi surface areas. For the case of the stage 3 compound the two observed DHVA frequencies were explained by identifying them with the second and third largest cross-sections of the model, and assuming that the largest cross-section was not observed. From this identication of the DHVA frequencies, the Fermi energy relative to the band extremum associated with a given carrier pocket can be found. Results for the Fermi energy thus obtained are shown in gure 65 for several well-staged graphiteK compounds. The Fermi energy is presented as a function of reciprocal K stage (1=n) and in terms of EF measured relative to the K-point extremum for the carrier pocket. The points are the measurements of Higuchi et al. (1980) and K Shayegan et al. (1980) and the curves are a t of the data to EF 1=n2=3 . Each of the two curves in the gure corresponds to a distinct carrier pocket. The energy bands dening the smaller carrier pocket are the ones for which interband Landau level transitions are observed by the magnetoreection technique (section K 4.6) and a value for EF as measured by the magnetoreection technique of Mendez et al. (1980a) for a stage 9 sample is also included in the gure. Since the model yields the cross-sectional Fermi surface area, the volume enclosed by the Fermi surface and the carrier density can also be calculated. The results obtained for the stage 3 and stage 4 compounds, expressed in terms of the fractional charge transfer of the intercalant, are respectively f 0:17 for C36 K and f 0:21 for C48 K, which are considerably lower than indicated by other experiments. Such low values for f can be explained by assuming that the DHVA experiment has not measured all of the Fermi surfaces. In this connection, it is of interest to note that application of the same model to DHVA experiments by Iye et al. (1980) on graphiteAsF5 compounds yield values for f 0:22, 0.22 and 0.26 respectively, for stage n 2; 3; 4 compounds, in reasonable agreement with results by other workers. Although the Tanuma model accounts for many of the observations in the graphiteK and graphiteAsF5 systems, certain physical modications to the model are necessary to provide an improved t to the experimental results. These modications involve introduction of an interaction between the intercalant and graphite layers, or expressed in another way, the screening e ect of the intercalant by

126

M. S. Dresselhaus and G. Dresselhaus

the graphite bounding layer must be considered explicitly. Such a modication of the graphite -bands is possible through use of the phenomenological model for Ek by Dresselhaus and Leung (1980, 1981) which explicitly includes the intercalate graphite layer interaction. A totally di erent approach to the interpretation of DHVA frequencies in graphite intercalation compounds has been proposed by Batallan et al. (1978) and by Bok (1978) in an attempt to explain the apparent stage-independenc e of DHVA frequencies reported for certain acceptor intercalants such as SbCl5 (Batallan et al. 1978), Br2 (Rosenman et al. 1979) and FeCl3 (Woollam et al. 1979). The insensitivity to intercalate concentration (stage) implies a high degree of carrier localization in the bounding layer planes, consistent with the interpretation of measurements of the electrical conductivity, and Raman and I.R. lattice mode spectra. This high degree of localization of carriers in the graphite bounding layers (Cb ) led Batallan et al. and Bok to propose a Cb ICb sandwich model about the intercalate (I) layer to provide the dominant contribution to the quantum oscillations (Bok 1978). In this model all the charge released by the intercalant is assumed to reside in the graphite bounding layers. This metallic sandwich also plays a dominant role in the conductivity and metallic Drude edge behaviour in the optical reectivity. Using a nearly free electron model and in-plane zone folding, e orts have been made to account in detail for the observed DHVA frequencies for compounds with SbF5 by Batallan et al. (1978), and with Br2 by Rosenman et al. (1979). This has not been an easy task from either an experimental or theoretical point of view. As indicated in table 22, some of the DHVA frequencies are identied as fundamental frequencies, others as harmonics or combination modes. Charge density wave q vectors have also been invoked to account for some of the observed frequencies. For the acceptor compounds a large number of DHVA frequencies are observed, and it remains to be veried that all the reported DHVA frequencies are intrinsic to single-staged, well-characterized and encapsulated samples. For the case of graphiteBr2 the DHVA frequencies shown in table 22 indicate some discrepancies in the values measured by di erent groups. Although the stoichiometry is approximatel y known for the various acceptor compounds, the in-plane ordering has not been established in detail. Hence it remains to be demonstrate d that the in-plane zone folding used to interpret the DHVA frequencies corresponds to the actual structural ordering. Furthermore for some acceptor compounds, such as graphiteFeCl3 , for which an insensitivity to stage has been reported for the DHVA frequencies for stage n 2, the ordering in the intercalate layer is incommensurate to that of the graphite layer so that zone folding of the in-plane graphite bands cannot be carried out. Another challenge to a nearly free electron type model is the explanation of the small e ective masses (close in magnitude to masses found in pristine graphite), observed for some of the DHVA frequencies. Application of the metallic sandwich model to the intercalants Br2 and SbCl5 yields fractional charge transfer values of f 0:5 and 1.0 respectively. It may be di cult to reconcile these large values of f with transport data. The model implies an equal concentration of electrons and holes which may be di cult to reconcile with galvanomagneti c measurements and reectivity data near the plasma edge. The observation of stage-dependent DHVA frequencies by Batallan et al. (1980) in the graphiteHNO3 system has led these authors to consider a modied interpretation of their quantum oscillatory phenomena based on the two-dimensional graphitic electronic structure proposed by Blinowski et al. (1980) which includes screening of the intercalant by the graphite bounding layer (see section 4.1.2).

Intercalation compounds of graphite

127

The explanation of the experimentally determined DHVA frequencies, their dependence on stage, angular dependence and e ective mass values in terms of theoretical models remains an active challenge. 4.6. Magneto-optica l phenomena 4.6.1. General considerations The observation of intraband (cyclotron resonance) and interband Landau level transitions have provided powerful techniques for the study of the electronic structure of graphite and its intercalation compounds. These techniques are powerful because they are based on resonant spectroscopy and therefore permit precise measurements to be made of Landau level separations. These techniques furthermore emphasize specic high symmetry points in the Brillouin zone where the magnetic energy levels are extremal. As in the case of the quantum oscillatory phenomena discussed in the previous section, the condition for the observation of resonant Landau level transitions is !c 1, which is readily satised in graphite and its intercalation compounds at low temperature, high magnetic elds and highly ordered samples.

Figure 66. Landau level contours for graphite with Hkc-axis from the K-point ( 0) to the H-point ( 1=2). The magnetic energy levels are computed for a eld of 5 T using 8, the band parameters of table 13. The Fermi level for graphite, indicated by EF is lowered upon intercalation by acceptors and raised for donors. Landau level transitions occur at the K- and H-points where the density of states for the magnetic sub-bands are extremal. Since Landau level transitions occur from occupied valence levels to unoccupied conduction levels, the lowering or raising of the Fermi level results in the cut o of specic Landau level transitions.

128

M. S. Dresselhaus and G. Dresselhaus

4.6.2. Cyclotron resonance studies Intraband cyclotron resonance transitions have been studied extensively in graphite. The initial experiments were carried out in the high quantum limit where the Landau level spacing is independent of the quantum number of the initial state. Using circularly polarized microwave radiation, Galt et al. (1956) obtained an accurate value for the K-point conduction band e ective mass, m 0:061. The complexity of the observed cyclotron resonance spectra led to theoretical calculations of the magnetic energy level structure (McClure 1960, Inoue 1962, Dresselhaus and Dresselhaus 1965), including selection rules and matrix elements for Landau level transitions. The predicted selection rules for cyclotron resonance transitions of n 1 3m (m integer) were conrmed by microwave cyclotron resonance experiments using linearly polarized radiation (Williamson et al. 1966) and were later used by Schroeder et al. (1968) to explain in detail the spectra previously observed by Galt. These theoretical models were also used to explain the cyclotron resonance spectra observed in the far infrared by Robinson (1974). Cyclotron resonance transitions have also been observed in graphite intercalation compounds, though the only compounds that have been studied are dilute graphite Br2 samples (Tanuma et al. 1978). For these samples a cyclotron resonance spectrum di erent from pristine graphite was observed. The e ective masses found for the Hkc axis are 0.10, 0.08 and 0.06, all for the hole sense of circular polarization. It is signicant that these magnitudes are comparable to K-point electron and hole masses in pristine graphite. 4.6.3. Interband Landau level transitions Interband Landau level transitions have been extensively studied in the graphite host material (Dresselhaus and Mavroides 1964, Schroeder et al. 1971, Doezema et al. 1979). Interband Landau level transitions occur resonantly about the H- and Kpoints in the Brillouin zone, where the magnetic energy levels are extremal (see gure 66). The resonances in the magnetoreection spectra associated with these interband Landau level transitions have been interpreted quantitatively in terms of the SlonczewskiWeissMcClure (SWMcC) band model (Slonczewski and Weiss 1958, McClure 1957), and have yielded accurate determinations of selected band parameters of this model (see table 13). The spectra associated with the H-point transitions (Toy et al. 1977) depend only on the SWMcC band parameters 0 and while the spectra associated with the K-point primarily yield the parameter 2 combination (0 =1 ) from the measured K-point Landau level separations. The Kpoint spectra also provide a determination of the band parameter combination 4 =0 22 25 =81 which is sensitive to the di erence in Landau level spacings in the valence and conduction bands (Schroeder et al. 1971, Mendez et al. 1980b). From analysis of `forbidden interband transitions (n 1 3m with m 6 0), the magnitude of the trigonal warping parameter 3 can be determined (Schroeder et al. 1971, Doezema et al. 1979). This parameter can also be determined from measurements of `forbidden intraband or cyclotron resonance transitions (Suematsu and Tanuma 1972 and Ushio et al. 1972). Thus the study of interband Landau level transitions in graphite has focused on use of this resonant spectroscopy to establish the form of the highly non-paraboli c dispersion relations for the entire region of the Brillouin zone where the Fermi surface is located (Dresselhaus and Mavroides 1964).

Intercalation compounds of graphite

129

Figure 67. Magnetoreection spectrum using (+) circular polarization for an acceptor compound (FeCl3 stage 7) at a photon energy ! 0:295 eV and for donor h compound (Rb stage 6) at 0:335 eV. For comparison, traces for graphite are h shown at comparable photon energies. The resonances are specied by the quantum numbers for the initial and nal states. (From the work of Mendez et al. 1980a).

The observation of interband Landau level transitions in graphite intercalation compounds has yielded information on both the form of the electronic energy dispersion relations, and the location of the Fermi level relative to the K-point band edge. The observed breakdown of selection rules has further been identied with c-axis zone-folding e ects. The rst observations of magnetoreection resonances in dilute halogen acceptor compounds yielded H- and K-point spectra almost identical to those observed in pristine graphite with regard to resonant magnetic elds and photon energies (Chung and Dresselhaus 1976, 1977). Since these data could be explained quantitatively in terms of the SWMcC band model, it was inferred that the electronic structure associated with most of the graphite layers in dilute compounds were almost unchanged from pristine graphite, and therefore could be described by a rigid band model (Dresselhaus et al. 1977a). Unlike the situation in graphite, magnetoreection resonances are not observed in intercalated graphite below a certain photon energy, the cut-o energy Ex , dependent on both stage and intercalate species. The cut-o of K-point interband transitions below Ex is associated with the introduction of carriers into graphite bands by the intercalant, resulting in a shift in Fermi level EF . Because of the Pauli exclusion principle, interband Landau level transitions are made from occupied valence to unoccupied conduction states. The lowering of EF caused by the introduction of holes in acceptor-type compounds leads to a cut-o of Landau level

130

M. S. Dresselhaus and G. Dresselhaus

Figure 68. Summary of resonant magnetic elds `fan chart for photon energies in the range 0:110 < ! < 0:320 eV for a FeCl3 stage 7 sample using () circularly polarized h radiation. To emphasize the similarities but measurable di erences, a comparison of the results for this compound and for pristine graphite is presented in the same photon energy range. (From the work of Mendez et al. 1980a). Table 23. Effective mass parameters for various intercalants and stages as measured by the magnetoreection technique.a AlCl3 Pure graphite m m c m v
2 0 =1

FeCl3 Stage 7 0.064 0.002 0.055 0.005 0.079 0.008 26.1 0.7 Stage 5 0.063 0.003 0.054 0.006 0.075 0.009 26.6 1.3

Rb Stage 6 0.053 0.001 0.045 0.003 0.065 0.004 31.7 0.8 Stage 8 0.055 0.002 0.046 0.004 0.070 0.007 30.6 1.1 K Stage 9 0.055 0.001 0.045 0.003 0.072 0.006 30.3 0.8

Stage 8 0.065 0.002 0.056 0.003 0.076 0.005 26.0 0.9

Stage 6 0.063 0.001 0.054 0.003 0.076 0.005 26.6 0.3

0.067 0.001 0.056 0.003 0.084 0.005 25.1 0.5

From the work of Mendez et al. (1980a). In the table the reduced e ective mass m is related to the 2 conduction and valence band masses by m 1 m 1 m 1 and 0 =1 is the pertinent SWMcC c v band parameter combination.

transitions as EF drops below the extremum of the magnetic sub-band for the initial stage (Platts et al. 1977). With increasing intercalate concentration, EF moves to lower energies relative to the K-point extremum, therefore increasing the magnitude of Ex . In donor type compounds the introduction of electrons causes EF to rise, leading to the cut-o of interband transitions as EF rises above the extremum of the magnetic sub-band for the nal state. Interband Landau level transitions in graphite intercalation compounds are further complicated by the non-uniform charge distribution in the graphite layers, resulting from the introduction of charge into the -bands by the intercalate layers (Pietronero et al. 1978, 1979). From the point of view of a ThomasFermi screening

Intercalation compounds of graphite

131

Figure 69. Dependence on reciprocal stage (1=n) of the Fermi level measured relative to the K-point band edge for several intercalants. The sign for (EF E8;K ) applies to the 3 donor compounds and the sign to the acceptors. For all intercalants, the graphite value is obtained in the limit 1=n ! 0 and is denoted by open squares. The open triangles ! are the Shubnikovde Haas results of Dresselhaus et al. (1980) and the closed triangle ! is from magnetic susceptibility results of Di Salvo et al. (1979).

model, the high charge density in the graphite bounding layer leads to a large band bending of the magnetic energy levels associated with the graphite bounding layer, and a correspondingly decreased band bending with increasing distance z from the intercalate layer (Mendez et al. 1980a). Because of this large band bending for the graphite bounding layers, the initial and nal states for typical Landau level transitions are normally either both occupied or both empty, thereby preventing the occurrence of Landau level transitions due to the Pauli principle. From this point of view, the observed magnetoreection spectra emphasize contributions from graphite interior layers. Since the resonances in the intercalation compounds exhibit approximately the same linewidths as for pristine graphite (see gure 67), it is inferred that the electronic energy band structure is essentially the same for all graphite layers contributing to the magnetoreection resonances. For well-staged dilute samples (stages n 4) the observed magnetoreection spectra are in most cases qualitatively similar to those of pristine graphite, but show di erences with regard to resonant magnetic elds and in some cases also di erences in resonant lineshapes and selection rules. Illustrative spectra are shown in gure 67 (Mendez et al. 1980a) for an acceptor compound (FeCl3 stage 7) and a donor compound (Rb stage 6). For comparison, spectra are also included for pure graphite at similar photon energies. Each structure in gure 67 is identied with a K-point Landau level interband transition, specied by the quantum numbers for the initial and nal states. Intercalation downshifts the resonant magnetic eld for both () and () senses of circular polarization, corresponding to a decrease in 1 (or equivalently an increase in Landau level separation) for both donor and acceptor compounds with

132

M. S. Dresselhaus and G. Dresselhaus

increasing intercalate concentration. A summary of the observed magnetic elds is given in gure 68 for a number of photon energies for a stage 7 graphiteFeCl3 sample and for graphite. The close similarity of the magnetoreection resonances for graphite and for intercalated graphite shown in gure 68, and observed for other intercalants with stages n 4, implies that the electronic band structure associated with these spectra remains essentially graphitic upon intercalation. From analysis of the magnetoreection spectra it is concluded that within 0.2 eV of the K-point band edge E8;K the electronic structure for energy bands contributing to the spectra 3 can be described by the form of the graphite SWMcC band model, with a weak dependence of the SWMcC band parameters on stage and intercalate species. This dependence of selected band parameters relevant to the E3 bands near the Fermi level (see table 13) is quantitatively deduced from the magnetoreection spectra. Results obtained for several intercalants and stages are given in table 23. It is signicant that the results in table 23 show only a small variation of the K-point conduction and valence band masses ( 925%). The magnetoreection experiment is also sensitive to the introduction of carriers into the graphite interior layers which causes the Fermi level to downshift for acceptors and upshift for donors with respect to the K-point band extremum. Such shifts in Fermi level give rise to the disappearance of interband transitions as the Fermi level either drops below the initial state in the case of donors or above the nal stage in the case of acceptors. This disappearance of an interband transition from the magnetoreection spectra is called the Fermi cut-o phenomena (Platts et al. 1977), and the photon energy Ex where interband transitions disappear determines the Fermi level EF relative to the K-point band edge E8;K . Results for (EF E8;K ) versus 3 3 reciprocal stage are given in gure 69 for several intercalants. Also included in the gure is a value for the Fermi level shift for bands associated with the graphite interior layers of a stage 4 graphiteK compound as determined by magnetic susceptibility measurements (Di Salvo et al. 1979). Results for EF E8;K obtained 3 from Shubnikovde Haas measurements of the smallest belly orbit (Dresselhaus et al. 1980) are also included in the gure. Resonant Landau level transitions are thus identied with magnetic energy levels in highly graphitic E3 bands which give rise to a Fermi surface that is readily measured by the de Haasvan Alphen e ect (see gure 65). The observation of a signicantly larger shift in Fermi level for the donor Rb compound than for the acceptor compounds of comparable stage shows that a larger amount of charge is transferred to the graphite interior layers for the Rb compound. This interpretation is consistent with results obtained on I.R.-active lattice mode spectra (see gure 87) and in-plane electrical conductivity studies (see section 4.2.1). Because of the superlattice c-axis periodicity associated with the staging phenomenon, new Brillouin zone boundaries can form in reciprocal space, thereby giving rise to new magnetic energy sub-band extrema, and hence new series of resonant Landau level transitions. Since the Landau level separations near the K-point depend only weakly on kz , the dominant e ect of the formation of new Brillouin zone boundaries is the admixture of states separated by a reciprocal lattice vector of the smaller zone, thereby giving rise to a breakdown in selection rules. A breakdown in the selection rules for allowed transitions in graphite (McClure 1960, 1971) has been reported (Mendez et al. 1980a) for acceptor compounds with AlCl3 . The observed breakdown in selection rules cannot be explained by trigonal warping e ects, but is consistent with c-axis zone-folding e ects.

Intercalation compounds of graphite

133

Table 24. Values for room temperature magnetic susceptibilitya for several graphite intercalation compounds. Compound Graphite C8 K C8 K C8 K C24 K C36 K C48 K C8 Rb C8 Rb C8 Rb C24 Rb C36 Rb C48 Rb C8 Cs C6 Li C16 Br3 C16 Br2 C16 Br2
a b

Stage 1 1 1 1 2 3 4 1 1 1 2 3 4 1 1 2 2 2

k b 721.1 1.02

? b 70.43 0.28

av b 77.32 0.53

powder

0 c 70.40

p d 0.016 0.64 0.64 0.64 0.25 0.21 0.34 0.56

1.50 1.75 1.87 0.66

0.05 70.02 70.06 0.05

0.53 0.57 0.57 0.25

0.62e 1.04 f 0.83 f 0.72e 0.67e 0.39e 0.23e 0.36 f 0.62 f 0.34e 0.39e 0.49 f

70.38 70.38 70.38 70.39 70.39 70.40 70.33 70.33 70.33 70.36 70.37 70.38 70.34 70.37 70.42 70.42 70.42

0.48 1.17

0.00 0.38

0.16 0.64

0.31 f 0.64 g 70.30 h 70.46i;j 70.63 f

All values of the susceptibility are given in units of 10 6 e.m.u./g. k , ? and av denote respectively the susceptibility for Hkc, H ? c and av k 2? =3. All values in these columns are from Di Salvo et al. (1979). c Standard values for the core susceptibility were used (Delhaes 1977). d Values for the Pauli spin susceptibility were obtained from specic heat data if available (see table 25 and text). e Rudor and Schulze (1954). f Delhaes (1977). g Delhaes et al. (1976). h Juza et al. (1949, 1955). i McDonnell et al. (1951). j Goldsmith et al. (1950).

4.7. Magnetic susceptibility, magnetic resonance and specic heat studies 4.7.1. Magnetic susceptibility The magnitude of the magnetic susceptibility of the graphite host materials is large, highly anisotropic (k 50? ) and diamagnetic over a wide temperature range 1 < T < 1000 K (Shoenberg 1952, Poquet et al. 1960, Sharma et al. 1974), where k and ? respectively denote the susceptibility for Hkc and H ? c. In the intercalation compounds, several changes occur in the behaviour of : (1) the magnitude of k is reduced by at least an order of magnitude for typical intercalation compounds: (2) tends to be positive (paramagnetic ) for donors, and negative (diamagnetic) for acceptors; and (3) the magnitude of is stage-dependent . In the limit of zero intercalate concentration, it is expected that the values of k and ? approach those of pristine graphite. A summary of the results for k and ? in graphite and in several donor and acceptor compounds is shown in table 24. In this table all susceptibility values are given in 106 e.m.u./g. Contributions to from the

134

M. S. Dresselhaus and G. Dresselhaus

Table 25. Low temperature specic heat coefcients in alkali metal donor compounds.a Molar weightb (g) (mJ/molK2 ) 12.011 12.011 11.287 11.287 15.0211 13.0946 12.7432 25.4436 16.8468 15.2784 14.4782 0.0138 0.03 0.500 0.430 0.697 0.241 0.189 0.63 0.25 0.19 0.16

Sample Graphite Graphite C6 Li C6 Li C8 K C24 K C36 K C8 Cs C24 Cs C36 Cs C48 Cs


a

(mJ/mol K4 ) 0.0277 0.026 0.0054 0.0093 0.1502 0.0375 0.0339 0.049 0.085 0.072 0.060

D (K)

Authors Van der Hoeven and Keesom (1963) Mizutani et al. (1978) Ayache et al. (1980) Delhaes et al. (1976)

413 421 710 590 9 234.8 > > 372.9 > > > 358.2 = 341 > 284 > > > 300 > ; 319

Mizutani et al. (1978)

The specic heat coe cients (electronic) and (lattice) are obtained by analysis of the temperature dependence of specic heat data C T T 3 . b The molar weight m is obtained from m xAC AM =1 x where AC and AM are the atomic weights of the carbon and metal atoms and x is the number of carbon atoms per intercalant.

core diamagnetism 0 are also listed. For those compounds where data for the electronic contribution to the specic heat are available (see table 25), the Pauli spin contribution due to free electrons was determined using the relations p 2 NEF ; B 4:36

where B is the Bohr magneton, the density of states NEF was determined from the electronic specic heat coe cient 2 2 k NEF ; 3 4:37

in which k is Boltzmanns constant and the results for p thus obtained are given in table 24. The observed susceptibility results from several contributions, orb 0 p; 4:38

where 0 is the diamagnetic contribution from the core electrons, p is the paramagneti c spin or Pauli contribution from the free carriers, and orb is an orbital contribution from the valence electrons. Normally orb is diamagnetic, and contains both intraband and interband orbital contributions. It is this orbital contribution from the valence electrons which has been shown to be responsible for the large magnitude of the diamagnetic susceptibility in graphite because of the simultaneous occurrence of (1) large interband contributions due to the very small band gaps, and (2) high density of states due to the very weak dispersion of the electronic energy levels along kz . Referring to table 24 we note that the susceptibility tends to be positive for the alkali metal donors and negative for halogen acceptors. For the case of H ? c, the orbital contribution is small, yielding small values for the total susceptibility. It is of interest to note that for the donor compounds, ? is

Intercalation compounds of graphite

135

diamagnetic for some stages (for example, for K and n 3), while k is paramagnetic for the same sample. Application of the LandauPeierls theory of the magnetic susceptibility was rst made by Hove (1955) to graphite and to graphiteBr2 compounds in an attempt to explain the susceptibility measurements in these materials by Hennig and McClelland (1955). This calculation was based on a two-dimensional model for the electronic structure and neglected interband contributions to orb . A decrease in orb by a factor of 5 was obtained for a shift in Fermi level by 0.3 eV on the basis of a rigid band model. On the other hand, Hove was not able to explain the large magnitude of orb in graphite on the basis of the LandauPeierls intraband contribution. A qualitative explanation for the large magnitude of orb in graphite was presented shortly thereafter by McClure (1956) who considered explicitly the interband contribution to orb on the basis of a simple two-dimensional model for the electronic structure of graphite. Haering and Wallace (1957) also attributed the large diamagnetism to interband contributions in orb , but emphasized the twodimensional aspects of the graphite electronic structure in their analysis. Though the two-dimensional models provided a qualitative explanation of the magnitude and anisotropy of the observed , the three-dimensional model later developed by McClure (1960) on the basis of the SlonczewskiWeissMcClure band model provided the rst quantitative t to the measured k in graphite over a wide temperature range 50 < t < 1000 K. This calculation showed explicitly that the major contribution to orb arises through departures from the LandauPeierls formula in the low quantum number limit (n 0, 1 levels). McClure further showed that the source of the anomalous diamagnetism is spread over a width comparable to the -bands along the Brillouin zone edge, with the largest contribution arising from the E3 band overlap region. This calculation was also signicant in yielding the rst determination of 0 , the nearest-neighbou r in-plane CC overlap integral (see table 13). To provide an explanation of the temperature dependence of k that is simultaneously consistent with a large number of other experiments (de Haasvan Alphen e ect, magnetoreection, infrared reectivity), a renement of the basic McClure calculation was made by Sharma et al. (1974) , treating explicitly the trigonal warping of the Fermi surface and using the Fukuyama (1971) formalism for the calculation of . Sharma et al. also calculated the dependence of orb on Fermi level EF . This calculation showed that orb is large and negative when EF is in the band overlap region with singularities at the saddle point E3 band extrema at H- and K-points in the Brillouin zone. A much reduced contribution was found for EF outside the band overlap region, with the magnitude of orb decreasing with increasing EF and nally changing sign at EF 0:4 eV. For EF 0 0:4 eV, orb was found to be paramagnetic and increasing in magnitude with increasing EF . The various calculations agree that the interband contribution of orb to ? is very small. Since the spinorbit interaction for graphite is very small and the g-factor is essentially equal to 2, both the core contribution 0 and the free electron spin contribution p are taken to be isotropic. Thus the large anisotropy in is attributed to the large diamagnetic contribution from interband processes to orb for k , and to the relative unimportance of this e ect for ? . On the basis of these susceptibility calculations, a number of qualitative conclusions can be reached. The relatively small magnitude of the measured susceptibilities of the intercalation compounds (see table

136

M. S. Dresselhaus and G. Dresselhaus

24) can be attributed to the passage of the Fermi level outside the band overlap region. For the donor compounds, the Sharma calculation predicted that orb becomes paramagnetic for a su ciently large shift in Fermi level. Calculations of the orbital contribution to the magnetic susceptibility for Hkc have been made by Safran and Di Salvo (1979) for the specic case of donor alkali metal compounds, using a simplied two-dimensional tight binding model for the graphite -bands, but including graphite bounding layerintercalant interaction through a ThomasFermi charge screening term in the free energy. In this formulation, a local pseudo-Fermi level associated with each layer is introduced to model the ThomasFermi screening and the Fukuyama formula for orb is used to calculate orb as a function of T and EF . In this connection it should be noted that a full band calculation, as has been done for C6 Li and C8 K, already contains the charge distribution as a function of z, so that no ThomasFermi screening terms need be added to a susceptibility calculation based on a full band model. In the limit of high T and low EF , the calculations by Sharma et al. and by Safran and Di Salvo become equivalent as they should. The Safran and Di Salvo (1979) calculation predicts a decrease in orb with increasing EF in the paramagnetic regime, in agreement with experiment. According to their model, each of the graphite layers is identied with a local electrochemical potential i which is used to model the ThomasFermi screening, where i Vi and Vi is the potential at layer i due to the screened intercalate layer. Safran and Di Salvo thus proposed use of this model to estimate the Fermi level relative to the band extremum for the bands closely identied with the graphite interior layers. Analysis of their susceptibility data yielded a value of EF 0:15 eV for fourth stage C48 K which is compared in gure 69 with the magnetoreection results for Fermi level shifts in the graphitealkali metal compounds. To obtain insight into the phase transitions occurring in second stage C24 K, temperature dependent studies of the magnetic susceptibility were carried out by Di Salvo and Fischer (1978) with particular relevance to the behaviour of in the vicinity of the transition temperatures Tl and Tu . The absence of measurable anomalies in at these temperatures was interpreted to indicate that the transitions involve primarily changes in ordering of the alkali metal intercalant layers, in agreement with di raction results. These phase transitions are discussed in section 3.4. Magnetic susceptibility measurements have been carried out on acceptor compounds in an e ort to look for local moments associated with the intercalant. The presence of local moments gives rise to a Curie temperature dependence Ng2 2 JJ 1=3kT ; B 4:39

where N is the density of such local moments, and g and J are the e ective g-factor and J the angular momentum quantum number for the local moments. Evidence for such local moments has been found in the case of graphiteFeCl3 and graphite FeCl2 compounds by Karimov et al. (1971) and by Ohhashi et al. (1974a, b) by observing a CurieWeiss temperature dependence of characteristic of magnetic materials above their ordering temperature (see section 3.4). For these intercalants the local moments were found to be large and associated with the free Fe2 and Fe3 ions and having the same g and J values as in the parent solids. In contrast, study of the temperature dependence of in the non-magnetic acceptor compound graphite AsF 5 by Weinberger et al. (1978b) has shown no evidence for a Curie behaviour, as

Intercalation compounds of graphite

137

could arise from moments due to unpaired spins. Their measurements in the temperature range 77 < T < 295 K show to be small, negative and temperatureindependent, with 3:5 10 7 e.m.u./g. On the basis of this result, Weinberger et al. concluded that there are no more than 0:1B per AsF 5 molecule or equivalently no more than 1% of the AsF5 intercalant is present in the form of a spin 1 species. 2 Though not reporting specic values for the susceptibility for graphiteBr2 compounds, Suematsu et al. (1980b) used the anisotropy of the susceptibility (k ? ) to estimate the fractional volume of unintercalated graphite in a graphiteBr2 sample, exploiting the large magnitude of (k ? ) for graphite relative to that in the intercalation compounds. 4.7.2. Nuclear magnetic resonance The nuclear magnetic resonance (NMR) technique provides information on the electronic charge distribution through measurement of the Knight shift and structural information through measurement of the temperature dependence of T1 , the spinlattice relaxation time, and of T2 , the spinspin relaxation time. The NMR technique permits study of the local environment of a nucleus on a graphite layer (through the 13 C line) or on an intercalant layer (through a chemical species present in the intercalant) thereby providing information often unavailabl e by other techniques. We review here some of the major ndings provided by the NMR technique. Knight shift studies have been carried out primarily on graphitealkali metal donor compounds. The rst work in this area was done by Carver (1970) who measured the Knight shift of the 133 Cs line in C8 Cs and C24 Cs and compared the shift in the intercalation compound to that in Cs metal. The Knight shift (H=H0 ) is proportional to the electron density at the nucleus, j0j2 , and to the electron spin susceptibility p ,

Figure 70. NMR resonance for 13 C in graphite and graphiteK compounds for stages n 1; 2; 3; 9. Note the upshift of the resonant eld with intercalation and the highly asymmetric lineshapes for the intercalation compounds. The solid vertical line is a calibration point for the 13 C NMR resonance in C6 H6 . The dashed vertical line corresponds to the 13 C resonance in pure graphite, from which (H=H0 ) is measured. (From the work of Conard et al. 1980.)

138

M. S. Dresselhaus and G. Dresselhaus


Table 26. Knight shift of the 13 C NMR line for various intercalation compounds relative to graphite. Values for the charge transfer coefcient f deduced from (H=H0 ) are given. (From the work of Conard et al. 1980.) Knight shift H=H0 (p.p.m.) 0 106 75 27 90 33 38 75 34 75 38 35 14 <10 Charge transfer coe cient f 0.00 0.74 1.06 0.38 0.84 1.00 1.07 0.70 0.96 0.70 1.07 0.98 0.12

Material Graphite C6 Li C12 Li C18 Li C8 K C24 K C36 K C8 Rb C36 Rb C8 Cs a C24 Cs C4 8Cs C7 SO3 F Other acceptorsb
a

From Knight shift measurements on the 133 Cs line, Carver (1970) found f 0:55 and f 1:0 for C8 Cs and C24 Cs respectively. b HNO3 (I, II, III), SO3 F (II, III), SO3 (I, II, III) and Br2 (I, II), where I, II, III denote stages n 1; 2; 3 respectively. Because the Knight shift is so small, it is not possible to obtain reliable values for f .

H=H0

8 j0j2 p ; 3

4:40

where H0 is the resonant eld for the nucleus in the same chemical environment as in the metal, but with zero electron charge density at the position of the nucleus. In his analysis, Carver assumed (H=H0 ) to depend only on j0j2 so that the admixture of s-function in 0 could be found from H=H0 1=2 . On this basis he concluded that for second stage C24 Cs the alkali metal intercalant is fully ionized, but only 55% ionized for rst stage C8 Cs (Carver 1970). Consistent results for the intercalate ionization for C8 Cs and C24 Cs were obtained through Mossbauer experiments on the 133 Cs nucleus by Campbell et al. (1977). The isomer shift of the absorption line in the intercalation compounds relative to fully ionic and metallic local environments yielded an estimate of the fractional intercalate ionization of 0:5 0:2 for C8 Cs and 1:0 0:2 for C24 Cs, in good agreement with the NMR results obtained on the same 133 Cs nucleus. Knight shift studies have recently been carried out by Conard et al. (1980) on a number of low stage compounds with the intercalants Li, K, Rb and Cs, but based on the 13 C nuclear resonance line. The advantage of this approach is that the same NMR line can be used to study a large number of intercalation compounds. On the other hand, The Knight shift results obtained from the 13 C line are less conclusive for several reasons. From an experimental point of view, precise measurements of the Knight shift of the 13 C NMR line are di cult because of the low abundanc e of the

Intercalation compounds of graphite


13 C

139

isotope, the very long spinlattice relaxation time T1 , and the high anisotropy of the local elds arising from the anisotropy in discussed in section 4.7.1. Because of the small penetration depth of the r.f. signals (100 mm at 9 MHz), the experiments are normally carried out on powdered specimens. In conventional NMR, the saturation e ect due to long T1 contributes to the broadening of the NMR lines, but in the Fourier transform pulsed NMR (where observation is made at low modulation elds H 1 0), the saturation e ect is due to too short a trigger time relative to T1 and leads to a decrease in signal. This e ect is observed in both pristine graphite and in most graphite intercalation compounds, as can be seen in the NMR lineshapes shown in gure 70. Saturation e ects are however greatly reduced in NMR measurements on the 13 C line for stage 1 and 2 alkali metal compounds because of the relatively shorter T1 for these compounds. From a theoretical point of view, the introduction of carriers into the graphite -bands by charge transfer from the intercalant does not in general lead to increased charge density with s-function character at the 13 C nuclear sites. The Knight shift is mainly sensitive to the sorbitals and not to the p-orbitals that are also present. Asymmetries in lineshape are expected to arise because the 13 C sites in the graphite bounding layers see a higher charge density than 13 C sites in the graphite interior layers. Nevertheless, several conclusions have been reached by Conard et al. (1980) on the basis of Knight shift studies on the 13 C lines. For all the donor compounds that have been studied, the resonant eld is upshifted to higher elds relative to the graphite reference line, as indicated in table 26. For all donor compounds in the table, signicant Knight shifts are found, and these shifts have been interpreted in terms of charge transfer coe cients f from the intercalant to the graphite layers. Estimates for f based on the Knight shift results are also included in the table. However, a quantitative analysis of the Knight shift data awaits a model for the NMR lineshape and an independent determination of p . A denitive separation between the Knight shift and the chemical shift must also be carried out in order to obtain a quantitative determination of f . From the results of table 26 it is noted that the stage 1 donor compounds tend to have incomplete intercalate ionization while the higher stage (n 2) compounds appear to be fully ionized. It should be noted that the Knight shift result for C6 Li in table 26 is not consistent with the band calculations of Holzwarth et al. (1977, 1978b) which show complete intercalant ionization. On the other hand, the results on C8 K are in qualitative agreement with the band calculations of Inoshita et al. (1977) and Ohno et al. (1979) . We also note that the experimental results on C8 Cs based on the 13 C NMR line (Conard 1980) are in qualitative agreement with those reported previously by Carver (1970) for the 133 Cs NMR line. Though Knight shift data provide important information on the charge transfer coe cient f , the present analysis is not yet denitive, nor are estimates of f from Knight shift data fully consistent with determinations of f by other techniques. Conard et al. (1980) also show that the 13 C line for the acceptor compounds with Br2 and HNO3 is very sharp in comparison with pure graphite and with the donor compounds. From the very small magnitude of the 13 C Knight shift for these acceptor intercalants, it was concluded by Conard et al. (1980) that f is very small and that the density of states is small at the Fermi level EF . Table 26 also gives Knight shift results for some acceptor intercalation compounds. The fractional charge transfer coe cients per intercalate unit are found to be much smaller for the

140

M. S. Dresselhaus and G. Dresselhaus

acceptors than for donors, in agreement with results obtained by a number of other techniques. Using spinecho techniques, Carver (1970) also measured the spinlattice relaxation time T1 of both the 13 C and 133 Cs lines for graphiteCs compounds (stages 1 n 5) and showed that the dependence of T1 on intercalate concentration could be explained in terms of either the density of states variation with intercalate concentration given by the SlonczewskiWeissMcClure band model, or by a model where the charge is localized on the graphite bounding layers (Salzano and Aronson 1966a). Since the initial work by Carver on the graphiteCs compounds, several studies of T1 in graphite intercalation compounds have been carried out. Temperaturedependent studies, such as on the linewidth of the 7 Li resonance in C6 Li, yield an abrupt decrease in T1 at a temperature corresponding to the onset of motional narrowing. For C6 Li this motional narrowing is attributed to a phase transition to a liquid-like intercalant at 280 K (Conard and Estrade 1977, Estrade et al. 1980), but this phase transition has not been conrmed by other experiments. Avogadro and Villa (1977) studied both the temperature dependence of T1 and T2 (the spinspin relaxation time) of the proton line in graphiteHNO3 and observed large discontinuities in T1 and T2 at the orderdisorder transition (250 K), consistent with motional narrowing in a liquid-like intercalant. A large anisotropy in T1 and T2 was observed, depending on the angle of the magnetic eld with the crystallographic c-axis. These authors also measured the proton self-di usion coe cient at room temperature by applying a magnetic eld gradient parallel to the applied magnetic eld, and measuring the spinecho height for a known value of the eld gradient. The results obtained by Avogadro and Villa (1977) show that the self-di usion coe cient for graphiteHNO3 is highly anisotropic, with much larger self-di usion parallel to the layer planes. Weinberger et al. (1978a) also carried out temperature dependent T1 and T2 measurements on the 19 F resonance in graphiteAsF5 and found a maximum in T1 1 near 175 K, while T2 1 was observed to change in a monotonic fashion by over two orders of magnitude, but over a broad range of temperatures 136 < T < 295 K. The large T1 values at high temperatures are attributed by Weinberger et al. (1978a) to the rapid reorientation of the intercalant. The single 19 F NMR line was identied by these authors as due to the AsF5 species in the intercalation compound. This is to be contrasted with the results of Resing et al. (1979) on graphiteSbF5 who identied two NMR lines with the 19 F nucleus. The broad and dominant line was attributed to a neutral SbF5 (or SbF 3 ) species, while a weaker sharp line was identied with a charged SbF species. This system was also 6 studied by Ebert and Selig (1977) who reported a single 19 F line. The absence of unpaired spins in the susceptibility study of the graphiteAsF5 system by Weinberger et al. (1978a) was interpreted as an inconsistency with the presence of the AsF 6 species in graphiteAsF5 , since the AsF would contribute a local moment to . On 6 the other hand, the experimental techniques used by Ebert and Selig (1977) and by Weinberger et al. (1978a) may not have been su ciently sensitive for the observation of a second 19 F NMR line, so that the presence of AsF cannot be ruled out. Further 6 work in the use of NMR to identify the intercalate species in the intercalate layer of acceptor compounds is now under way. The temperature dependence of the linewidths has also been measured for a number of systems to determine the onset of motional narrowing and the direction of molecular rotation. For example, NMR proton studies have been made by Facchini

Intercalation compounds of graphite


Table 27. Density of states at the Fermi level NEF as obtained from low temperature specic heat () and Pauli spin susceptibility (p ) measurements and from band calculations (theory). NEF in units 10 20 /(eV cm3 ) From Graphite C8 K C24 K C36 K C8 Cs C24 Cs C36 Cs C48 Cs C6 Li C8 AsF5 C16 AsF5 Cu Ag
c f

141

From p 28e 37e

From theory

6.6a 14.3b 233b 92b 78b 191b 90b 74b 66b 216 c 251 d 250 160

193 f

244 g

References: a Van der Hoeven and Keesom (1963); b Mizutani et al. (1978); Delhaes et al. (1976); d Ayache et al. (1980); e Weinberger et al. (1978b); Inoshita et al. (1977); g Holzwarth et al. (1978b).

et al. (1980) in ternary lamellar derivatives such as C24K (THF) where THF denotes tetrahydrofuran . These results have been interpreted to give a molecular axis of rotation parallel to c for the large THF group. Chemical shift studies have also been used (Ebert et al. 1976, Ebert and Selig 1977) to identify the chemical state of the intercalant in the intercalation compounds. These chemical shift studies utilize the di erent local elds at the position of the nucleus arising from each chemical species at nearest-neighbou r and next-nearest neighbour positions, etc. in the intercalate layer. For example Ebert and Selig (1977) concluded from their study of the chemical shifts that the single 19 F NMR line, which they found for the SbF5 and AsF 5 intercalants, could be identied with either XF5 or XF but not with the XF3 species, where X Sb or As, in contrast with the 6 work of Resing et al. (1979) who concluded that both a charged and a neutral intercalate species were present. The nuclear resonance technique in conjunction with electron spin resonance (ESR) on the same samples was used by Weinberger et al. (1978b) to provide a direct determination of the paramagneti c spin susceptibility for stage n 1 and n 2 graphiteAsF5 compounds. This determination of the paramagnetic spin susceptibility is based on the relation between the electronic and the nuclear susceptibilities e and n e e =n Ie =In n ; 4:41

142

M. S. Dresselhaus and G. Dresselhaus

Table 28. Room temperature g-shifts and linewidths for several crystalline and powder graphite intercalation compounds from electron spin resonance measurements.
Powder measurements g H Authors Wagoner (1960) Lauginie et al. (1980) Lauginie et al. (1980) Lauginie et al. (1980) Muller and Kleiner (1962) +0.0004 +0.0005 70.0012 +0.0030 70.0003 70.0005 70.0006 70.0002 70.0004 70.0083 11.6 28.7 0.52 0.50 1.08 +0.0001 +0.0040 +0.0010 +0.0003 +0.0003 +0.0002 0.58 0.54 1.05 +0.0001 0.40 +0.0004 0.35 1.00 +0.0004 0.05 5.40 70.0006 5.80 not observed ! not observed ! not observed ! 0.48 22.4 13.7 +0.0067 190 2.76 5.27

H0 kC Material Graphite C6 Li C12 Li C8 K C24 K C24 K C36 K C108 K C8 Rb C24 Rb C36 Rb C8 AsF5 C16 AsF5 C24 AsF5 C6 HNO3 C12 HNO3 C16 HNO3 C24 HNO3 C30 HNO3 C16 Br2 C16 ICl C5 SO3 C10 SO3 C15 SO3 gk +0.0473 70.0003 70.0001 70.0007 +0.0001 Hk 2.36 0.70 11.4 3.9

H0 ? Cu g? +0.0003 +0.0006 +0.0006 +0.0007 +0.0009 H? 2.64 0.81 12.6

Lauginie et al. (1980) Lauginie et al. (1980) Muller and Kleiner (1962) Lauginie et al. (1980)

9 = ; o

Khanna et al. (1978) Lauginie et al. (1979) Khanna et al. (1978)

Khanna et al. (1978)

+0.0002 +0.0002 +0.0003

0.54 0.58 0.68

Lauginie et al. (1980)

where e and n are respectively the electronic and nuclear gyromagnetic ratios, and Ie and In are respectively the integrated intensities of the ESR and NMR lines. The nuclear susceptibility for the 19 F NMR line in the graphiteAsF5 compounds is thus determined directly from the Curie law h n NJJ 1n 2 =3kT ; 4:42

where the gyromagnetic ratio n is known for the 19 F nucleus and e is related to the h electron g-factor and to the Bohr magneton B by e gB =. From their analysis, Weinberger et al. (1980b) found values of e 0:59 0:10 107 e.m.u./g for stage 1 and e 0:80 0:14 10 7 e.m.u./g for stage 2. Their measurements showed that e has no measurable anisotropy. It should be noted that these values of e are considerably smaller than values for p for typical donor compounds (see table 24). Identication of e with Pauli spin paramagnetism p 2 NEF allows a B determination to be made of the density of states at the Fermi level yielding a value of 2:8 1021 and 3:7 1021 states/(eV cm3 ) respectively for the n 1 and n 2 AsF 5 compounds. It was noted that these values for NEF are approximately one order of magnitude smaller than found in metals such as Cu and Ag with comparable electrical conductivities (see table 27). Similar conclusions concerning the low density of states at the Fermi level for acceptor compounds are obtained from ESR measurements by Khanna et al. (1978) on AsF5 and SbF 5 compounds and by

Intercalation compounds of graphite

143

Figure 71. A single line is observed in the conduction electron spin resonance of graphite intercalation compounds. The above trace is for C8 AsF5 (stage 1) at room temperature and the parameters A, B and the linewidth H are dened in the gure. The lineshape is Dysonian and typical of metallic conductors. (From Khanna et al. 1978.)

Lauginie et al. (1980) on SO3 and HNO3 compounds. Furthermore, the values for NEF deduced from p are probably overestimated because of correlation e ects. A rigid band model for the density of states was used by Weinberger et al. (1980b) to estimate the Fermi level and the charge transfer coe cient f from the AsF 5 intercalant to the graphite layers. A value of f 0:3 was obtained for the stage 1 compound C8 AsF5 . 4.7.3. Electron spin resonance Electron spin resonance (ESR) has provided an important tool for the characterization of graphite materials because of the sensitivity of this resonance to the concentration of unpaired spins. Because of the very small magnitude of the spinorbit interaction in carbon (0.000 22 eV), the g-factor for graphite is very close to the free electron value of g 2:0023 (McClure and Yafet 1962). Nevertheless, shifts in the g-factor g from the free electron value are measurable and found to be highly anisotropic (Wagoner 1960), with gk gk 2:0023 4:73 10 2 for Hkc and g? g? 2:0023 3 104 for H ? c. Studies on graphite have focused on the anisotropy of the g-factor, on the lineshape of the resonance and on the temperature dependence of the parameters describing the resonance. McClure and Yafet (1962) have shown that the SlonczewskiWeissMcClure band model accounts quite well for the anisotropy and temperature dependence of the ESR measurements. The initial ESR studies on graphite intercalation compounds were carried out by Hennig et al. (1954) in acceptor compounds with intercalants H 2 SO4 , Br2 and UCl4 and by Smaller et al. (1955) in donor compounds with the intercalants Li, Na, K, Ca, Ba. Very narrow linewidths, dependent strongly on intercalate concentration, were found for the graphiteH 2 SO4 compounds, with linewidths ranging from H 4g to a minimum linewidth of H 0:04g. No resonances were identied for the acceptor compounds with Br2 and UCl4 . For all the donor compounds that were

144

M. S. Dresselhaus and G. Dresselhaus

examined, ESR lines were found with linewidths and lineshapes characteristic of metallic samples. Studies by Muller and Kleiner (1962) on donor compounds showed that the large anisotropy of the g-shift observed by Wagoner (1960) for graphite was absent in the donor compounds. Mu ller and Kleiner further showed that the g-shift and linewidths generally increased with increasing intercalate concentration though the behaviour for the lowest stage compounds often did not follow this trend (see table 28). The linewidths generally increase with increasing intercalate atomic number or spinorbit interaction. The results for the donor compounds imply a signicant overlap between the conduction electron wavefunctions and the alkali metal sites. More recent work has focused on accurate measurements of the g-shifts for Hkc and H ? c, characteristics of the observed lineshapes for donors and acceptors, measurement of the dependence of the linewidth on temperature and stage, and determination of the spinspin relaxation time T2 and the c-axis spin di usion time. The interpretation of the ESR measurements has stressed the relation between the c-axis spin di usion and the characteristic behaviour of the c-axis conductivity. Electron spin resonance studies by Khanna et al. (1978) and Lauginie et al. (1980) conrm that both donor and acceptor compounds exhibit a single Dysonian ESR line as shown in gure 71, characteristic of high purity metallic samples. In agreement with the work of Muller and Kleiner (1962), the g-shifts found for both donor and acceptor compounds by Khanna et al. and Lauginie et al. are much smaller and much more isotropic than for graphite, as shown in table 28. The absence of a Curie-law temperature dependence of the susceptibility for rst stage C8 AsF 5 places an upper limit of 10 2 B per AsF 5 molecule. Weinberger et al. (1978a) have used this result to argue against the observed resonance being due to the magnetic resonance of a paramagnetic species, and to support the identication of the resonance with conduction electrons or holes. The g-shifts gk tend to be negative for both acceptors (holes) and donors (electrons) except for some low stage donor compounds where positive gk values are found (see table 28). The magnitude of the g-shifts is stage dependent, also supporting the identication of the resonance with conduction electrons. From the ESR linewidths H, the spinspin relaxation time T2 is determined from the relation h 1=T2 gB H=: 4:43

For acceptor compounds, the linewidths decrease and T2 increases as the temperature is raised. In general, the temperature dependence of the linewidths for the acceptor compounds is gradual, but for the case of the graphiteHNO3 (stage 2 and 3) compounds, Khanna et al. (1978) observed large and abrupt changes in linewidth at the orderdisorder transition temperature at 250 K. This observation on the ESR linewidth is consistent with the behaviour reported by Avogadr o and Villa (1977) for T1 and T2 based on NMR measurements near the orderdisorder transformatio n temperatures. Anomalies in the temperature dependence of H associated with orderdisorder transitions have also been reported for AsF5 compounds (n 1; 2; 3) by Khanna et al. (1978) and for rst stage HNO3 by Lauginie et al. (1979). For the rst stage donor compound C8 K, a characteristic metallic temperature dependence is observed for H in the temperature range 100 < T < 300 K (Lauginie et al. 1980). Several characteristics of the linewidth behaviour of the donor compounds suggests that the spinorbit interaction mechanism of Elliott (1954) is signicant

Intercalation compounds of graphite

145

in the spinspin relaxation process. For the donor compounds, Lauginie et al. (1980) found the linewidth to increase with increasing atomic number, consistent with a signicant admixture of alkali metal wavefunction into the -band conduction electron wavefunction. When combined with the spinorbit interaction, this admixture of wavefunction results in a mixing of the spin states of the conduction electrons. In the absence of local moments, T2 can be related to the resistivity scattering time and to the measured g-shift g by the Elliott relation (1954) T2 =g2 ; 4:44

where is a numerical coe cient characteristic of the material, but probably quite insensitive to the stage index. If T2 is dominated by the spinorbit interaction in graphite intercalation compounds, then would be expected to be less anisotropic for intercalated than for pristine graphite. The Elliott relation can perhaps be used to obtain information on the stage dependence of the c-axis scattering time . From analysis of the parameters of the Dysonian ESR lineshape according to the procedures developed by Dyson (1955) and Feher and Kip (1955) the spin di usion time TD across the skin depth can be determined. Typical values for TD in graphite intercalation compounds are 108 s, tending to be longer for acceptors than for donors. The temperature dependence of TD has been measured experimentally by Lauginie et al. (1980) for rst stage C8 K and TD is found to exhibit a T 2 dependence, characteristic of metallic behaviour. The di usion constant D can be obtained from the di usion time TD D 2 =TD 4:45

and related to c-axis transport phenomena is sensitive to the di usion of electrons in and out of the skin depth . Using the relation between the di usion constant D and the c-axis mean free path c D 1 c 3 and the relation for the c-axis conductivity c Ne2 c =m , we obtain Identifying the kinetic energy of the carriers at the Fermi level with EF 1 m r2 , and F 2 the density of states for a metal at the Fermi level NEF with NEF 3N=2EF c e2 DN EF : we obtain the relation for the c-axis conductivity (Lauginie 1981) 4:49 c 3Ne2 D=m 2 : 4:47 4:46

4:48

Estimates of the c-axis transport properties of pristine graphite indicate that the c-axis mean free path c is comparable with c-axis interlayer separations. From measurement of the di usion constant D from ESR lineshapes it should be possible through (4.46) to obtain c and hence determine whether the dominant conduction process is due to band conduction c Ic or to hopping (c 9 Ic ) in graphite intercalation compounds. Temperature dependent measurements of D also provide important information on the c-axis conduction mechanism. Temperature-dependen t measurements of the ESR lineshape in C8 K by Lauginie et al. (1980) yield a weak temperature dependence for D, which increases somewhat upon lowering the temperature from 300 to 100 K,

146

M. S. Dresselhaus and G. Dresselhaus

in agreement with the sign of the temperature dependence of c (Ubbelohde 1972). Detailed studies are needed to determine whether the magnitude of the observed temperature dependence of c can be explained by (4.49). Estimates for NEF from analysis of ESR lineshapes yield NEF values comparable to those for good metals (10 22 /eV cm3 ) in the case of donor compounds. However, NEF values, lower by approximately two orders of magnitude, are estimated for acceptors from analysis of the ESR lineshapes (Khanna et al. 1978 and Lauginie et al. 1980). These conclusions about the density of states in donor and acceptor compounds (see table 27) are roughly consistent with results obtained for NEF by specic heat measurements (section 4.7.4). The estimate for NEF for acceptor compounds from ESR measurements is low relative to the determination by Weinberger et al. (1978b) based on measurement of the spin susceptibility. 4.7.4. Specic heat Low temperature specic heat measurements yield valuable information about both the electronic structure and the lattice. In general, the specic heat exhibits a temperature dependence C T T 3 ; 4:50

in which the cubic term is the lattice contribution and the linear term is the electronic contribution. Specic heat measurements have been made on the K, Rb and Cs alkali metal compound by Mizutani et al. (1978) and on C6 Li by Delhaes et al. (1976) and by Ayache et al. (1980). No specic heat measurements have so far been reported for acceptor compounds. Results for the coe cients and are summarized in table 25, together with values for the Debye temperature deduced from the coe cients . In all cases (except for rst stage C6 Li) the Debye temperatures D are signicantly lower than in graphite. For the graphiteK compounds, the D values are found to exhibit an approximately linear dependence on reciprocal stage (1=n). The coe cients listed in table 25 are given in terms of a molar unit dened by m xAC AM ; 1x 4:51

where AC and AM are the atomic weights of the carbon and alkali metal and x is the number of carbons per intercalant. Inspection of the values in table 25 shows that for the donor intercalation compounds the measured values are close to those obtained for the pristine alkali metals and more than one order of magnitude greater than in graphite. The electronic specic heat coe cient can be directly related to the density of states at the Fermi level by 2 2 k NEF : 3 4:52

If the specic heat coe cient is expressed in mJ/mol K2 , then and NEF are related by 2:357N EF , where NEF is expressed in units of (eV)1 per carbon atom. Values for the density of states at EF obtained from the measured electronic specic heat coe cients are given in table 27. From these density of states values the Pauli paramagneti c spin susceptibility is calculated and results for p are given in table 24.

Intercalation compounds of graphite

147

In their analysis of the electronic specic heat for the intercalation compounds with stage n 2, Mizutani et al. (1978) used the SlonczewskiWeissMcClure (SWMcC) band model to calculate the density of states at the Fermi level assuming complete intercalate ionization and a uniform charge distribution over the graphite layers. The SWMcC expression for NEF obtained from (4.14) is p 2 NEF 2EF 2 = 30 ; 4:53 assuming 3 4 5 0. The Fermi level EF is also found from (4.14) in terms of , the number of electrons added per carbon atom by charge transfer from the intercalant. The resulting relation between and EF is then p 2 5:54a EF EF 2 = 30 ; so that NEF
2 0

p 1 2 p 2 2 4 30 1=2 : 3

4:54b

Using values for NEF obtained from the SWMcC model, Mizutani et al. obtained good agreement for the functional dependence of on as varied from 1/24 for a stage 2 compound to 1/48 for a stage 4 compound. On the other hand, quantitative agreement with experiment was not achieved with commonly accepted values of the graphite band parameters, perhaps because of signicant shifts in EF due to intercalation, thereby introducing contributions from the E1 band, not included in the above expressions. From the analysis, Mizutani et al. were, however, able to conclude that the ionized metal sheets do not contribute signicantly to the electronic specic heat, at least for the n 2 compounds. In the analysis of their specic heat data, Mizutani et al. (1978) also considered the e ect of the electron phonon enhancement of the specic heat for the case of C8 K where the density of states has been calculated by Inoshita et al. (1977). Using the denition for the electronphonon enhancement coe cient , NEF expt NEF calc a value of 0:21 was obtained.
Table 29. Superconducting parameters for graphite intercalation compounds. Compound C8 K a C8 K C8 Rb C8 Cs C8 K C8 K b C8 K
a b

1 ;

4:55

Tc K 0.55 = 0.39 0.0200.151 ; 0.0200.135 0.1280.198 0.125 <0.3

Reference

Hannay et al. (1965) Koike et al. (1978, 1980a, b) Kobayashi and Tsujikawa (1979) Poitrenaud (1970)

Sample had excess potassium intercalant. Graphite host material was grafoil.

148

M. S. Dresselhaus and G. Dresselhaus

Figure 72. The a.c. magnetic susceptibility of a C8 K sample as a function of temperature at nearly zero d.c. eld. The abrupt change in the magnetic susceptibility near 134 mK is identied with the superconducting transition. (From the work of Koike et al. 1980a).

For the case of the graphiteCs compounds a Schottky-typ e anomaly in the specic heat was found by Mizutani et al. (1978) and identied with a low frequency optical mode associated with the motion of the Cs atoms on the intercalate layer relative to the C atoms on the graphite bounding layers (Kondow et al. 1977). Einstein frequencies of !E 33 and 24 cm1 were obtained for C8 Cs and for the higher stage compounds respectively, using the relation !E k E , where E are the h corresponding Schottky temperatures of 48 and 34 K respectively. It is inferred from the specic heat results that the Einstein frequencies for the intercalant K lie considerably higher than for the Cs compounds, consistent with the higher mass of Cs relative to K.

4.8. Superconductivity The occurrence of superconductivity in graphite intercalation compounds was rst observed by Hannay et al. (1965) who identied superconducting transitions in the rst stage compounds with the intercalants K, Rb and Cs, after an initial unsuccessful search for superconductivity by Hennig and Meyer (1952). Superconductivity in C8 K has been further conrmed by the magnetic susceptibility measurements of Koike et al. (1978, 1980a, b) and of Kobayashi and Tsujikawa (1979) and by resistivity measurements of Koike et al. (1980a, b). The values obtained for the transition temperature Tc by these studies are given in table 29. These more recent determinations of Tc also account for the failure of Poitrenaud (1970) to nd a superconducting transition in C8 K down to 0.3 K, using electron spin resonance techniques. It is of interest to note that Hannay et al. (1965) found that a rst stage graphiteK sample with excess potassium had a higher transition temperature (0.55 K) than stoichiometric samples (0.39 K), though the discrepancy in Tc values between the results of Hannay et al. (1965) and those of more recent workers is not understood, but may be due to impurities. Nevertheless, the occurrence of superconductivity in a compound where neither component is itself

Intercalation compounds of graphite

149

Figure 73. Measurement of a.c. magnetic susceptibility by Koike et al. (1980a) at 90 mK as a function of d.c. magnetic eld H for a rst stage C8 K sample. (a) Hkc-axis. From the magnetic eld dependence of , type I superconducting behaviour is inferred. (b) H ? c-axis. From the magnetic eld dependence of , type II superconducting behaviour is inferred.

superconducting is well substantiated by the susceptibility versus temperature curve shown in gure 72 for C8 K. The anistropic electronic properties of intercalated graphite also give rise to an anisotropy in the superconducting behaviour upon application of a magnetic eld and the anisotropic properties are conventionally described in terms of the angle between an applied magnetic eld H and the c-axis. Through measurement of the magnetic susceptibility as a function of H for various magnetic eld orientations, Koike et al. (1980a, b) found that 0 9 658, C8 K behaves as a type I superconductor. This is shown in gure 73(a) for 08, where full magnetic ux exclusion is found for most of the magnetic eld range, with Hc denoting the thermodynamic critical eld and Hsc a supercooled critical eld. In contrast, for 658 9 908, C8 K behaves as a type II superconductor, showing no complete Meissner e ect, as illustrated in gure 73(b) for 908. The upper critical eld Hc2 (Hc2 10 G) is indicated on the gure. This is the rst material that has been found to exhibit both type I and type II superconductivity . The upper critical eld Hc2 exhibits strong anisotropy which is t to the experimental data using the theoretical relation by Morris et al. (1972)

150

M. S. Dresselhaus and G. Dresselhaus

Figure 74. The angular dependence of Hc2 at 94 mK for a C8 K sample for which Tc 145 mK. Measurements by Koike et al. (1980a) of the thermodynamic critical eld Hc in the type I region show negligible angular dependence, while measurements of the supercooled critical eld Hsc Hc3 show an angular dependence (see text).

Hc2 =Hc2k cos 2 "2 sin2 1=2 :

4:56

In applying (4.56) to the angular range 0 9 658 where type I behaviour is observed, Koike et al. (1980a) made use of the relation Hc2 Hsc =1:695 so that Hc2k was determined by H2ck H sc 0=1:695. The parameter " which enters (4.56) is sensitive to anisotropic behaviour and is given by
2 2 "2 m? =mk Hc2k =Hc2? k =? 2 ;

4:57

where is the coherence length and m is the e ective mass. By this procedure Koike et al. (1980a) were able to plot the angular dependence of Hc2 =Hc2k and the results are given in gure 74 at a temperature T 94 mK for a C8 K sample for which Tc 145 mK. Figure 74 shows the thermodynamic critical eld to be independent of for most of the angular range where type I superconductivity is observed. On the other hand, the supercooled critical eld Hsc plotted in gure 74 as Hsc Hc3 exhibits an angular dependence which is given by (4.56) with the scale factor 1.695 which relates Hsc to Hc2 . The t obtained in gure 74 yields "2 0:026 for this C8 K sample, which from (4.57) results in the anisotropy ratios for the e ective mass and coherence length mk =m? 39 and ? = k 6:2 respectively. Although some variation in these parameters is found from one sample to another, the authors point out that the mass anisotropy is comparable to that found in transition metal layered compounds such as 2HTaS 2 and 2HNbSe (Morris and Coleman 1973), but smaller than the mass anisotropy of 170 found for pristine graphite (Spain 1973). Because of the high anisotropy exhibited by H c2 , Koike et al. (1978) concluded that the electrons associated with the cylindrical parts of the Fermi surface contribute signicantly to the superconducting behaviour. A similar conclu-

Intercalation compounds of graphite

151

sion was reached by Kamimura et al. (1980) on the basis of their calculation of the electronphonon coupling constant ( 0:25). According to the BCS theory of superconductivity , the transition temperature Tc and the Debye temperature D are related by Tc 0:85
D

exp 1=N0V;

4:58

where D 234:8 K from specic heat measurements (table 25) of Mizutani et al. (1978) and the interaction strength N0V 0:14 which is comparable to the interaction strength in the weak coupling superconductor iridium where N0V 0:12, Tc 112:5 mK, and Hc T 0 16g. Application of McMillans formula 1:041 D Tc 4:59 exp 1:45 1 0:62 yields a value of 0:32 0:01 for the electronphonon coupling constant, using 0:1 for the e ective Coulomb repulsion parameter, as suggested by McMillan (1968). This value of is of comparable magnitude to 0:21 obtained by Inoshita et al. (1977) from the electronic density of states, deduced from specic heat measurements (Mizutani et al. 1978), and the value of 0:25 calculated directly by Kamimura et al. (1980). On the basis of this analysis, Koike et al. (1980a, b) concluded that C8 K could be treated as a weak-coupling, highly anisotropic, threedimensional superconductor within the framework of the BCS theory and they evaluated a number of the superconducting parameters. The observation that C8 K can be a type I or a type II superconductor, depending on the magnetic eld direction, is of particular interest. Except for this unusual feature, the superconducting properties of C8 K are qualitatively similar to those observed in other layered compounds, such as 2HNbS2 (Toyota et al. 1976), and alkali and alkaline earth metal intercalates of MoS2 , a semiconductor that can be made superconducting by intercalation (Woollam and Somoano 1976). Although higher stage graphitealkali metals have been investigated for possible superconducting transitions, no such transitions have so far been reported. A preliminary report has however been given by Alexander et al. (1980) of superconducting behaviour in a ternary intercalation compound C8 KHg based on anomalous specic heat behaviour near 1.9 K. Since pristine Hg is superconducting, this family of ternary compounds may exhibit superconductivity for higher stage compounds. If superconductivity could be found in higher stage alkali metal compounds, these systems could perhaps provide interesting prototype materials for the study of two-dimensional superconductivity . 5. Lattice mode spectra 5.1. Methods for studying lattice modes in intercalated graphite The study of the lattice mode spectra of graphite intercalation compounds has been found to be a fruitful research area because of the wealth of microscopic information provided by these studies. The interpretation of these experiments is simplied by the layer structure of these compounds, exhibiting very strong intralayer binding and relatively weak interlayer binding, so that as a rst approximation the lattice modes in the graphite layers are related to the graphite host and in the intercalate layers to the bulk intercalant. This information on the coupling between the graphite and intercalant is provided by studies in the

152

M. S. Dresselhaus and G. Dresselhaus

intercalation compounds of departures from the lattice mode structure of the unperturbed constituent layers. The layer structure of these materials allows the separation of the lattice modes into in-plane and c-axis modes. Experimental studies have largely focused on the in-plane lattice modes because the cleavage plane is a cface, and good optical surfaces can be prepared by cleavage. Three complementary experimental techniques are especially important for the study of lattice mode structure:inelastic neutron, Raman and infrared spectroscopy. All three techniques have been successfully applied to graphite and its intercalation compounds. Infrared spectroscopy probes infrared-active modes of very small wave vector because the wave vector of the incident photons is very small compared with Brillouin zone dimensions. Because of the high optical absorption of graphite and of intercalated graphite, the infrared spectra are taken by measurements of the reectivity. To obtain the lattice mode frequencies from the measured spectra, an analysis of the spectral lineshape is carried out, taking into account contributions to the dielectric constant from lattice mode oscillators and background terms arising from free carrier absorption and interband transitions. Since inelastic neutron and Raman spectroscopy are both scattering phenomena, the change in frequency between the incident and scattered neutron or photon is measured to determine the frequency of the absorbed or emitted phonon. Because of the small wave vector of the incident and scattered photons, rst-order Raman spectroscopy is limited to the observation of lattice modes close to the Brillouin zone centre, in contrast to the case of inelastic neutron scattering where by using a thermal neutron beam, the whole Brillouin zone can be explored. Although rst-order Raman spectroscopy only provides information on q 0 phonons, information on the phonon dispersion relations for other points in the Brillouin zone can be obtained from second-order spectra, where contributions are made by pairs of phonons with wave vectors q and q. Second-order Raman lines are generally broad, because many phonon pairs with di erent energies contribute to the line, and the resulting line empasizes those regions in the Brillouin zone having high densities of states. For the case of graphite intercalation compounds, the high symmetry of the graphite lattice is an approximate symmetry for the larger real space unit cell for the intercalation compounds. This approximate symmetry gives rise to zone-folding phenomena through which certain non-zone-centre q-vectors for graphite are mapped into the zone centre for the intercalation compound, thereby turning on new Raman-active modes. This zone-folding phenomenon is yet another reason why the basic graphite mode structure is of great importance to the understanding of the intercalation compounds. Because of the high optical absorption of graphite and its intercalation compounds, Raman scattering experiments on in-plane modes are conveniently performed on cleaved c-faces using a Brewster angle back-scattering geometry. Incident photons are typically provided by a c.w. argon-ion laser, though other excitation sources are preferred in certain cases, such as resonant enhancement studies. The scattered radiation is analysed by a conventional double grating monochromator . To prevent sample instabilities and intercalate desorption, Raman measurements are conveniently made using low laser power (<50 mW) and low temperature (for example, 77 K). Encapsulation of the Raman and IR samples in ampoules tted with suitable optical windows is usually necessary for maintaining stage delity of the samples during the spectroscopic measurements. To excite c-axis

Intercalation compounds of graphite

153

Figure 75. Zone-centre optical lattice modes in pristine graphite. For the in-plane modes (E1u ; E2g1 ; E2g2 ), only one of the degenerate pair of modes is shown. The c-axis modes (A2u ; B1g1 ; B1g2 ) are non-degenerate. Infrared and Raman activity are indicated. The zero frequency acoustic modes (E1u ; A2u ) corresponding to pure translations are not shown.

modes, light with the E-vector along the c-axis is introduced on to an optical a-face, which can be prepared by sputtering with argon ions (Zanini and Fischer 1977, Zanini et al. 1978c). Inelastic neutron scattering has the advantage of probing the dispersion relations for the entire Brillouin zone. On the other hand, neutron spectroscopy requires large scattering volumes, and for this reason, this technique is limited to studies on intercalation compounds based on a pyrolytic graphite host material. In pyrolytic graphite there is almost random orientation of the a-axes within the scattering volume, but there is c-axis alignment to better than 18 (Moore 1973), and for this reason, inelastic neutron scattering focuses primarily on the dispersion relations for c-axis polarized modes. 5.2. Lattice modes in graphite Because of the strong connection between the lattice mode structure in the intercalation compounds and the graphite host material, a summary of the lattice mode structure of pristine graphite and of isolated graphite layers is presented here. Pristine graphite crystallizes according to the D 4 space group and has twelve 6h vibrational modes at q 0. These modes are shown in gure 75 and have been classied by Brillson et al. (1971) as three acoustic modes (A2u E1u ), three infrared active modes (A2u E1u ), four Raman active modes (2E2g ) and two silent modes (2B1g ).

154

M. S. Dresselhaus and G. Dresselhaus

Figure 76. Graphite phonon dispersion curves along several high symmetry axes from Maeda et al. (1979). The -point symmetries for the graphite structure are indicated.

The frequencies of the in-plane Raman-active E2g mode and the infra-red active E1u mode are well established at !E2g 1582 1 cm1 (Tuinstra and Koenig 1970, Brillson et al. 1971) and at !E1u 1588 2 cm 1 (Brillson et al. 1971, Nemanich et al. 1977b). Because of the strong intralayer force constants relative to the interlayer force constants, the vibrational frequencies of these modes are nearly the same, and are almost entirely determined by the simple optical displacements of the two inequivalent carbon atoms in a single layer plane. The small frequency di erence of 6 cm1 between !E2g2 and !E1u ) is associated with interlayer force constants arising from di erences with respect to interplanar displacements (see gure 75). This small frequency di erence thus provides a measure of the magnitude of these interlayer force constants. Using infrared spectroscopy, Nemanich et al. (1977b) measured the out-of-plane A2u mode at 868 1 cm1 , while inelastic neutron (Nicklow et al. 1972) and Raman (Nemanich et al. 1975) scattering results have identied the low frequency E2g 1 mode at 48 and 42 cm1 respectively. The silent out-of-plane B1g 1 mode at 128 cm1 has been identied by Nicklow et al. (1972) on the basis of inelastic neutron scattering measurements on the low frequency (9470 cm1 ) phonon branches. Analysis of the inelastic neutron scattering data for graphite by Nicklow et al. (1972) has provided phonon dispersion curves for the two lowest frequency phonon branches along M (see the Brillouin zone in gure 35). Near the M-point, the two lowest frequency modes are nearly degenerate at 466 cm1 , and correspond to zaxis displacements. Using the two lowest frequency phonon dispersion curves obtained from these neutron di raction measurements and the zone centre A1u , E1u and E2g 2 mode frequencies measured by IR and Raman spectroscopy, a Born von Karman force constant model has been used to yield the phonon dispersion relations for the higher lying phonon modes. The results obtained on this basis by Maeda et al. (1979) are shown in gure 76 and provide a framework for the discussion of the lattice mode structure for intercalated graphite. Other models for the phonon dispersion relations have also been proposed (Young and Koppel 1965, Nicklow et al. 1972, Nicholson and Bacon 1977, Nemanich and Solin 1979) but none of these models are consistent with the identication of the A2u zone centre mode frequency at 868 cm 1 . Since the only available experimental information above 500 cm 1 is the -point mode frequencies for the E2g2 , E1u and A2u modes, the

Intercalation compounds of graphite

155

Figure 77. Unpolarized room temperature Raman spectra taken by Underhill et al. (1979) in the backscattering geometry (E ? c) for stage n 1; 2; 3; 4; 6 and 11 graphiteferric chloride compounds and for pristine graphite (HOPG). Laser excitation at 4880 A and a power level <50 mW were used to excite in-plane Raman-active E2g2 modes ^ (see inset). The upper frequency component (E2g2 ) is identied with the graphite bounding layer mode and the lower component (E8g2 ) with the graphite interior layer 2 mode.

calculated dispersion relations above 500 cm1 are more tentative than at the lower frequencies where inelastic neutron scattering information along several symmetry directions in the Brillouin zone has been used in obtaining the dispersion relations. Second-order Raman spectra for pure graphite have also been reported (Nemanich and Solin 1977, 1979), showing a strong feature near 2710 cm 1 and two weaker features at 2450 and 3250 cm 1 . Of particular signicance is the observation of second-order frequencies higher than 2!E1u . This result is consistent with the phonon dispersion relations of gure 76 (and with other proposed models) which show that the maximum phonon frequency does not occur at q 0. Since the second-order spectra are associated with phonons of wave vectors q and q, these measurements are sensitive to the phonon density of states. From gure 76 we note that the highest frequency branch along M shows a large density of states near the maximum phonon frequency, thus accounting for the frequency of the 3250 cm1 second-order line reported by Nemanich and Solin (1977, 1979). 5.3. Raman scattering results on graphite intercalation compounds The Raman spectra for graphite intercalation compounds with stage n > 2 characteristically exhibit a doublet structure at frequencies close to the singlet E2g2 peak found in pristine graphite (Dresselhaus et al. 1977b, Song et al. 1976,

156

M. S. Dresselhaus and G. Dresselhaus

Figure 78. Reciprocal stage (1=n) dependence of the Raman frequencies associated with the ^ graphite interior layers (E8g2 ) and the graphite bounding layers (E2g2 ) for various 2 ^ acceptor intercalants. An upshift is found for both !(E8g2 ) and !(E2g2 ) versus (1=n) 2 and the behaviour is similar for all acceptor intercalants (see text for references).

Nemanich et al. 1977c, Solin 1977). This doublet structure for the intercalation compounds is illustrated in gure 77 for the graphiteFeCl3 system, along with the spectrum for pristine graphite (Underhill et al. 1979). A similar doublet structure with a frequency separation of 20 cm1 is found for n > 2 for all intercalants studied, including the acceptors Br2 , IBr and ICl (Song et al. 1976), FeCl3 (Underhill et al. 1979), AlCl3 (Gualberto et al. 1980), SbCl5 (Eklund et al. 1980b), and the donors K, Rb and Cs (Nemanich et al. 1977c, Eklund et al. 1977, Solin 1977, 1980). The lower frequency component is attributed to the E2g 2 carbon atom vibrations in interior graphite layer planes and the lattice mode associated with these layers is denoted by E8g 2 . This identication is supported by the proximity of the E8g 2 mode to 2 2 the E2g 2 mode of pristine graphite, by the decrease in intensity of the E8g2 with 2 decreasing stage (increasing intercalate concentation) and by the vanishing of the E8g 2 line in stage 1 and stage 2 compounds, where there are no graphite interior 2 layers (Nemanich et al. 1977c, Dresselhaus et al. 1977b). It is also signicant that for n 2 the lineshapes are Lorentzian and the linewidth of the E8g 2 component is not 2 sensitive to intercalate concentration. This result suggests that all interior graphite layers have approximately the same set of in-plane force constants. A delocalized charge density introduced by the intercalant into the interior graphitic layers is consistent with the Raman result for E8g 2 . 2 The upper frequency component of the doublet structure (see gure 77) is identied with an E2g2 -type graphitic mode occurring in a bounding graphite layer, ^ and this mode is denoted by E2g 2 . Support for this identication comes from the

Intercalation compounds of graphite

157

Figure 79. Reciprocal stage (1=n) dependence of the Raman frequencies associated with the ^ graphite interior layers (E8g2 ) and the graphite bounding layers (E2g2 ) for various 2 ^ donor intercalants. A downshift is found for both !8g2 ) and !(E2g2 ) versus (1=n) (see 2 text for references).

^ absence of the E2g2 line in pristine graphite, the increase of its intensity with ^ increasing intercalate concentration and the occurrence of a single E2g 2 line in stage 1 and stage 2 compounds where all graphite layers are bounding layers. The upshift ^ in frequency of the E2g2 mode relative to the E8g2 mode is due to the di erence in 2 force constants arising from the di erent environment of the carbon atoms in the graphite bounding and interior layers. The dependence on reciprocal stage of the ^ relative intensities of the E8g2 and E2g2 modes indicates that a single bounding layer 2 on either side of the intercalate layer e ectively screens the intercalate layer from the graphite interior layers. This behaviour is also supported by the z-dependence of the charge distribution calculated on the basis of a ThomasFermi model, yielding a high charge density in the graphite bounding layers and a rapid decrease in charge density with distance from the intercalate layer (Pietronero et al. 1978, 1979). ^ The frequencies for the E8g 2 and E2g2 modes also exhibit a distinctive dependence 2 on reciprocal stage. For the acceptor compounds such as with intercalants FeCl3 , ^ AlCl3 , Br2 , AsF 5 , SbCl5 and HNO3 , both mode frequencies !(E8g2 ) and !E2g 2 2 exhibit approximately the same frequency upshift as a function of reciprocal stage (1=n) as shown in gure 78. A similar stage-dependent upshift of the two modes gives rise to a doublet separation of 22 2 cm1 for these acceptor compounds. In contrast, the donor alkali metal compounds exhibit a downshift in these mode frequencies with increasing reciprocal stage (Solin 1977, 1980, Eklund et al. 1980a) as shown in gure 79. This mode softening is consistent with an in-plane lattice expansion with increasing (1=n) as reported by Nixon and Parry (1969). The sti ening of these lattice modes in the acceptor compounds and their softening in the donor compounds has been attributed to lattice strain associated with stagedependent changes in the lattice parameters by Underhill et al. (1979). For the case of graphiteFeCl3 compounds, the constancy of the linewidth for the E8g 2 interior 2 layer mode as a function of intercalate concentration and the identical frequency

158

M. S. Dresselhaus and G. Dresselhaus

Figure 80. Raman intensity for the stage 1 alkali metal compounds C8 K, C8 Rb and C8 Cs at T 4 K over a wide frequency range, as measured by Eklund et al. (1977).

Figure 81. Raman spectra of C8 Rb (stage 1), C24 Rb (stage 2) and C36 Rb (stage 3) at 77 K and over a wide frequency range. The data were taken by Eklund et al. (1977) in the Brewster angle, back-scattering geometry.

upshift for both interior and bounding Raman modes has been interpreted by Underhill et al. (1979) in terms of an identical strain within both the bounding and interior graphite layers for a given compound, a condition also necessary to prevent sample fracture upon intercalation. The magnitude of the strain increases as the number of interior graphite layers decreases, consistent with the idea that the stress introduced by intercalation is shared by fewer layers as the stage index n decreases. These Raman studies indicate that the acceptor compounds should experience an inplane lattice contraction in the graphite layers as a function of (1=n). Such a lattice contraction has recently been conrmed by direct X-ray measurements in NiCl2 (Flandrois et al. 1981), in FeCl3 and AlCl3 (Krapchev et al. 1981) and in AsF5 (Markiewicz et al. 1980b). Thus the addition of electron charge density expands the lattice while its removal results in lattice contraction. ^ A strong dependence of the relative intensities of the E8g2 and E2g2 components 2 on intercalate concentration is also observed. Increasing the intercalate concentration causes the intensity of the lower frequency component to decrease and the

Intercalation compounds of graphite

159

Figure 82. Raman scattering intensity at 77 K for stage 1 alkali metal compounds C8 K, C8 Rb and C8 Cs in the range 550 9 ! 9 610 cm1 using the Brewster angle, backscattering geometry. The polarization directions for incident and scattered light are indicated. The upper frequency component of the observed structure is identied with E2g symmetry and the lower frequency component with A1g symmetry. The dashed curves are a t by Eklund et al. (1977) to the data using a Lorentzian lineshape model.

intensity of the upper frequency component to increase. In this connection it is of interest to note in gure 77 that for stage 4 where the number of graphite bounding and interior layers is equal, the two peaks have approximately equal intensity. The discussion has so far focused on the distinctive similarities in the Raman spectra associated with the E2g 2 modes for many of the intercalation compounds. The Raman spectra for the alkali metal stage 1 compounds C8 K, C8 Rb, C8 Cs and C6 Li are, however, qualitatively di erent from spectra that have been reported for stage 1 acceptor intercalation compounds. For example, the spectra for C8 K, C8 Rb and C8 Cs are closely related to each other (Nemanich et al. 1977c, Eklund et al. 1977), as shown by the traces in gure 80, but are in sharp contrast with spectra observed for higher stage compounds with the same intercalant, as illustrated in gure 81 for the graphiteRb system. The spectra for the three stage 1 compounds in gure 80 show a broad, asymmetric line at high frequencies and a sharp structure at lower frequencies. Contributions to the broad line come for the high density of states in this phonon frequency range. Symmetry arguments have been given to account for the low and

160

M. S. Dresselhaus and G. Dresselhaus

high frequency features with particular reference to zone-folding e ects for a p2 2 superlattice which bring the M-point modes into the zone centre (Eklund et al. 1977). The broad lines around 1500 cm1 have the asymmetric BreitWigner lineshape (Scott 1974), which can arise through interaction between discrete Raman-active phonon modes and a continuum of phonon modes (Eklund et al. 1977). A strong interaction between the intercalate and graphite layers is also consistent with a variety of results on the electronic properties (see section 4) which provide strong evidence for a large charge transfer between the alkali metal and graphite layers. It is signicant that there are qualitative di erences observed between the Raman spectra for stage 1 alkali metal compounds and stage 1 acceptor compounds such as with the intercalants FeCl3 (see gure 77), AlCl3 (Gualberto et al. 1980) and AsF5 , HNO3 and SbCl5 (Eklund et al. 1979). This di erence in behaviour has been attributed by Dresselhaus and Dresselhaus (1979) to the greater coupling between intercalate and graphite layer planes in donor compounds as compared with acceptor compounds. The strong coupling is further documented by the values of the BreitWigner parameters for the asymmetric broad lines near 1500 cm 1 which have been given by Eklund et al. (1977) for C8 K, C8 Rb and C8 Cs. Referring to the spectra for the stage 1 alkali metal compounds shown in gure 80, a sharp spectral feature appears in the vicinity of 560 cm1 for C8 K, C8 Rb and C8 Cs. This spectral feature is shown in more detail in gure 82 using a polarization analysis to provide information on the two constituent components which exhibit di erent basic symmetries and are separated in frequency !. The two components are well resolved for C8 Cs, and appear as a di erence in linewidth for the (k; k) and (k; ?) polarization geometries for C8 Rb and C8 K. A weak component at higher frequency has also been reported by Caswell and Solin (1979) for C8 Cs (at 620 cm 21 ). It is of interest to note that for C8 Cs the component with (k; ?) polarization exhibits a discontinuous change in the BreitWigner frequency parameter at the melting temperature of Tm 608 K, though the other BreitWigner parameters and q show no temperature variation near Tm (Caswell and Solin 1979). This observation indicates a signicant coupling between the intercalant and graphite bounding layers for this donor compound, in contrast with the behaviour for the stage 1 acceptor C8 AsF 5 where no inuence on the graphite bounding layer mode is seen in the region of the orderdisorder transition (Eklund et al. 1979). For the p2 2R08 in-plane superlattice which characterizes the rst stage C8 K alkali metal compounds (see section 3.1), the M-point of the larger graphite Brillouin zone is folded onto the point of the smaller zone for the intercalation compound (see gure 41), thereby making additional graphite modes symmetry-allowed. According to this argument, zone folding of the phonon dispersion relations in gure 76 makes a number of additional modes symmetry allowed. Further work is necessary to explain why the modes observed in the vicinity of 560 cm1 are the only new modes that are observed out of a signicantly larger number of possible zonefolded modes. Caswell and Solin (1979) have argued that the modes in gure 82 are disorder-induced and reect maxima in the phonon density of states. However, a phonon density of states argument does not explain the absence of these modes in rst stage C6 Li and in rst stage acceptor compounds which also exhibit a dominant peak in the phonon density of states near this frequency. A direct calculation of the phonon dispersion relations for C8 K and C8 Rb has been carried out by Horie et al. (1980) using a Bornvon Karman model. In this

Intercalation compounds of graphite

161

model Horie et al. used the graphite force constants previously calculated by Maeda et al. (1979) and yielding the dispersion curves in gure 76. The force constants in the intercalate layer were taken as those in the pristine alkali metal, and the force constants between the carbon and alkali metal atoms were adopted from the neutron scattering studies on C8 Rb by Ellenson et al. (1977). In this calculation Horie et al. considered explicitly, the ; ; ; intercalant stacking sequence which is characteristic of well-annealed C8 K and C8 Rb (see gure 20). The calculations of Horie et al. (1980) show a high -point density of states in the vicinity of 580 cm1 , close to the experimentally observed mode frequencies in gure 82. However, the calculated point modes in the vicinity of 580 cm1 have displacements parallel to the c-axis, which could be excited in the experimental geometry used by Eklund et al. (1977) and Nemanich et al. (1977c) if the graphite layers become puckered by the intercalation process. On the other hand, the calculated -point modes with inplane displacements occur near 860 cm1 , but no Raman-active in-plane modes have been observed experimentally in this frequency range. Further work must be done to resolve these apparent inconsistencies between experiment and the model. The model of Horie et al. (1980) yields, in agreement with experiment, carbon atom -point modes that are higher in frequency for C8 Rb than for C8 K, because of the larger intercalategraphite interaction in the case of C8 Rb. The spectra of gure 82 have been interpreted by Eklund et al. (1977) in terms of modes that are upshifted by ! from a zone-folded M-point graphite mode frequency and exhibit a splitting of ! between the two components with di erent symmetries. Both the frequency upshift ! and the splitting ! increase with increasing size of the intercalate ion and the consequent increase of the graphiteintercalate interaction. The work of Horie et al. (1980) shows that the low frequency modes are more sensitive to the intercalate graphite interaction than are the high frequency modes. The model of Horie et al. gives in-plane intercalate modes with ! in the range 76 < ! < 106 cm1 for C8 Rb and 101 < ! < 120 cm1 for C8 K. However, no Raman-active alkali metal modes have been reported for C8 K, C8 Rb or C8 Cs. The frequencies for these intercalate modes are expected to be comparable to k D for the parent metals, where the Debye energies for the alkali metals are 63, 39 and 26 cm1 for metallic K, Rb and Cs, respectively. Some evidence for a soft intercalate mode is provided by the observation of a low temperature anomaly at 34 K (25 cm1 ) in the heat capacity of rst stage C8 Cs and at 48 K (36 cm 1 ) in stage 2, 3 and 4 graphiteCs compounds (Kondow et al. 1977, Mizutani et al. 1978). Information on the relative vibrational amplitudes of the alkali metal ions and the carbon atoms is provided by Mossbauer experiments on graphitecaesium compounds carried out by Campbell et al. (1977) on the 133 Cs nucleus. The large anisotropy that was observed in the DebyeWaller factor was interpreted to indicate that the motion of the Cs ions was tightly constrained along the c-axis, but relatively free in the intercalate layer with mean displacements for the Cs ions of hx2 iCs 35:8 104 A2 and hz2 iCs 18:9 104 A2 . This is in contrast to the carbon atoms which have large c-axis displacements and small in-plane displace ments, hx2 iC 12:5 104 A2 and hz2 iC 37:6 104 A2 . The broad asymmetric features of the Raman spectra for C8 K, C8 Rb and C8 Cs in the vicinity of 1500 cm1 have been interpreted in terms of an interaction between a Raman-active graphite phonon mode and a continuum of graphite phonon modes, which are coupled because of the reduction in symmetry introduced by the presence of the intercalate layer (Eklund et al. 1977). This interpretation has

162

M. S. Dresselhaus and G. Dresselhaus

Figure 83. Raman spectra for a graphiteBr 2 sample at 77 K in the frequency regions where the intercalate modes are prominent. The structure identied with the Br2 stretch mode in the intercalation compound is denoted by !0 , and its harmonics by n!0 . Fine structure in the vicinity of !0 is indicated by arrows. (From the work of Eklund et al. 1978.)

been substantiate d by a lineshape calculation of the asymmetric BreitWigner line in which the graphiteintercalate layer interaction is assumed to give rise to a coupling between the zone centre E2g2 mode and the continuum of modes arising from twophonon processes with frequencies extending to 3700 cm1 (Eklund and Subbaswamy 1979). Because of the high two-phonon density of states for ! > 1600 cm1 , this coupling gives rise to a downshift of the E2g2 mode at q 0. The Raman spectrum for C6 Li di ers from that of C8 K, C8 Rb and C8 Cs in two signicant ways. First, the room temperature spectrum of C6 Li exhibits (Zanini et al. 1978) no sharp line in the vicinity of 560 cm1 and the high frequency structure consists of a single line peaked at 1598 cm1 , lying very close to the E2g2 graphite line, and exhibiting only a slight BreitWigner asymmetry. At low temperatures, this single line is observed to be superimposed on a broad background structure (Eklund et al. 1980a). The absence of structure in C6 Li in the vicinity of 560 cm 1 is consistent with the in-plane zone-folding model, since for C6 Li it is the K-point modes that fold into q 0. From gure 76 it is seen that the K-point modes occur at di erent frequencies than the M-point modes. With regard to the Raman spectrum near 1600 cm1 , the smaller magnitude of the BreitWigner coupling parameter for C6 Li relative to the C8 K alkali metal compounds results in a smaller displacement of the peak frequency from the E2g2 graphite mode, a smaller departure from a Lorentzian lineshape, and a weaker coupling to the continuum of phonon states. This weaker coupling to the continuum of phonon states is consistent with the very small size of the Li ions which therefore have a smaller e ect on lowering the translational symmetry in the graphite layers. In contrast, stage 1 acceptor compounds (for example, with the intercalants AsF 5 , AlCl3 and FeCl3 ) exhibit Lorentzian lineshapes rather (Eklund et al. 1979) than the asymmetric BreitWigner lineshapes which occur in stage 1 alkali metal

Intercalation compounds of graphite

163

Figure 84. Loge dependence of Raman peak intensities on laser excitation energy for graphite-bromine samples of several intercalate concentrations (n denotes the stage of ^ the compound). The intensity ratio for the E2g2 mode on the graphite bounding layer ^ compared with that for the E8g2 mode in the interior layers, denoted by I E2g2 = 2 IE8g2 , is dependent on intercalate concentration, but independent of laser excitation 2 energy. In contrast, the intensity of the molecular strength mode shows large resonant enhancement e ects, but has the same dependence on intercalate concentration as the graphite bounding layer mode at 1600 cm1 . The intensity scale is logarithmic (base e) to cover the two orders of magnitude change in Raman intensity. (From the work of Eklund et al. 1978.)

compounds. Since acceptor compounds tend to form intercalate layers that are in most cases incommensurate with the graphite layers, the intercalate is not as e ective in lowering the translational symmetry of the graphite layers. Therefore the coupling between the Raman-active phonon modes and the continuum of phonon modes is not so important in stage 1 acceptor compounds. Additional evidence for a weak coupling between the graphite bounding layer and the intercalate layer in rst stage acceptor compounds comes from the absence ^ of any discontinuous change in either the peak frequency or the linewidth of the E2g 2 graphitic mode in C8 AsF 5 as the temperature is varied in the range 100 < T < 300 K where an orderdisorder transformatio n occurs in the intercalate layer. In addition to the high frequency modes identied with graphite layer vibrations, lattice modes at lower frequencies are also observed and are identied with the intercalate layer, because the reported spectra are di erent for each intercalantfor example, Br2 (Song et al. 1976, Eklund et al. 1978), IBr and ICl (Song et al. 1976) and FeCl3 (Caswell and Solin 1978). Except for the Br2 stretch mode some doubt remains about whether the reported low frequency modes are due to a surface deposit, to gas in the ampoule or to the intercalant (Eklund et al. 1979). As an example of Raman spectra associated with intercalate modes, the spectrum at 77 K for a graphitebromine compound (approximately stage 5) is shown in gure 83. The most pronounced low frequency feature in this spectrum is the strong, broad peak at !0 242 cm1 , which together with its harmonics at 2!0 and 3!0 , account for the strongest of the spectral features (Eklund et al. 1978). In addition, a number of smaller structures are observed, especially at low phonon frequencies. A correspondenc e has been made between the frequency of the main spectral feature at 242 cm1 and the molecular stretch mode occurring in solid Br2 at 300 cm1 (Cahill and Leroi 1966) and in the free Br2 molecule at 323 cm1 (Herzberg 1945). This interpretation of the 242 cm1 line provides support for the molecular identity of the intercalant in C8n Br2 compounds, consistent with X-ray and electron

164

M. S. Dresselhaus and G. Dresselhaus

di raction results (Eeles and Turnbull 1965). Consistent with this interpretation is the increase in intensity of all the spectral features of gure 83 with increasing intercalate concentration. In fact, a similar dependence on intercalate concentration ^ is found for the intensities of the !0 intercalate mode and of the E2g2 graphitic mode on the bounding graphitic layer (Eklund et al. 1978), while the mode frequencies for all the spectral features in gure 83 are essentially independent of intercalate concentration. This observation is interpreted to indicate that the intercalate force constants are principally determined by intralayer interactions with some perturbations introduced by the bounding graphite layers, but not by interior graphite layers. Further support for the molecular identity of bromine in C8n Br2 compounds comes from the observation of resonant Raman e ects associated with a rapid increase in the intensity of a Raman line when the laser excitation energy is close to an electronic transition. The observation of a strong resonant enhancement e ect for ^ the !0 242 cm1 intercalate line (shown in gure 84 as the I !0 =I E2g 2 curve) and ^ 2g 1600 cm1 graphitic line on the bounding the absence of such e ects for the E 2 ^ graphite layers (shown in gure 84 as the I E2g 2 =I E8g2 curve) indicates that the 2 electronic transition responsible for the resonant enhancement phenomenon is an electronic transition involving a bromine level. Similar resonant enhancement e ects are indicated by observations made on harmonics 2!0 and 3!0 . The insensitivity of ^ the resonant enhancement e ect (the curve labelled I !0 =I E2g 2 in gure 84) to intercalate concentration has been interpreted by Eklund et al. (1978) to indicate that the electronic transition in C8n Br2 is not between a bromine intercalate level and the Fermi level. The dissociation of the Br2 molecule has been suggested by Heinger (1979) as an intermediate state for the resonant enhancement process for the Br2 intercalate stretch mode. Consistent with this identication of the intercalate mode is the observation in gure 84 of a similar dependence on intercalate concentration of the intensity of the ^ intercalate line at !0 242 cm1 and of the bounding layer graphite E2g2 line at 1 for all laser excitation energies. It is also signicant that the intensities of 1600 cm the various spectral features in gure 83 decrease with increasing temperature, but the rate of decrease is strongly dependent on the type of spectral feature. Detailed study of the temperature dependence of the Raman spectra for the intercalate modes may provide a useful tool for the study of orderdisorder transformation s in molecular intercalation compounds. Also of interest are polarization studies of the intercalate modes to provide information on the alignment of the molecular axes. A polarization study of the graphiteBr2 spectrum (Chung 1977) is consistent with the Br2 molecular axis lying in the intercalate layer. The polarization analysis of lines attributed to the intercalate FeCl3 has been related to the polarization dependence of the Raman spectrum in solid FeCl3 by Caswell and Solin (1978), indicating that the intercalate layer in graphiteFeCl3 compounds maintains the same layer structure as found in solid FeCl3 . Other workers (Underhill et al. 1979) have identied these low frequency modes with FeCl3 surface layers. In this connection it is of interest to note that structure in the low frequency range was found by Eklund et al. (1979) in the case of the intercalant AsF5 , HNO 3 and SbCl5 , but by careful study of the di erence spectra, these authors were able to show that the low frequency Raman lines were due to gas in the ampoule in the case of HNO3 and AsF 5 , and to a surface deposit in the case of SbCl5 . Raman scattering has also been applied to the study of adsorbed molecules on graphite surfaces for vapour pressures below the threshold for intercalation. For the

Intercalation compounds of graphite

165

Figure 85. Room temperature infrared reectivity spectra taken by Underhill et al. (1979) for stage n 2, 4, 6 and 11 graphiteferric chloride compounds and for pristine graphite (HOPG). The t of the lineshape analysis to the experimetal data is given as the dotted curves. The infrared-active mode for graphite is the E1u mode shown in the inset which, in the intercalation compounds is designated as E8u for interior 1 ^ graphitic layers and as E1u for bounding graphite layers.

case of Br2 adsorbed on graphite surfaces, surface-specic Raman modes have been reported by Heinger (1979) with frequencies intermediate between 242 cm1 for the intercalate Br2 mode and 300 cm1 for solid Br2 . 5.4. Infrared studies of the lattice modes Being sensitive to odd-parity phonon modes, infrared (IR) spectroscopy provides a complementary tool to Raman spectroscopy, which is sensitive to evenparity modes. This complementarity applies to pristine graphite because the inversion symmetry of its crystal structure designates each lattice mode by either even or odd parity. Intercalation introduces a change in symmetry so that intercalation compounds have only approximate inversion symmetry. Nevertheless, it is found experimentally that the parity selection rules are preserved upon intercalation, so that IR spectroscopy only excites modes which are odd in the graphite host and likewise Raman spectroscopy only excites modes derived from even-parity graphite modes. This preservation of approximate inversion symmetry indicates that the coupling between the graphite and intercalate layers is weak, and that inversion symmetry should be used to model the phonon modes of intercalated graphite. Because of the high infrared absorption of graphite intercalation compounds associated with the electronic structure, infrared studies are carried out on reection. To obtain information on the lattice modes, the lattice contribution to the dielectric constant must be determined from the measured reectively. Therefore the interpretation of the IR lattice reectivity spectra requires a lineshape analysis to separate the lattice mode contributions elattice to the dielectric constant from the other contribu-

166

M. S. Dresselhaus and G. Dresselhaus

Figure 86. The dependence on inverse stage (1=n) of the mode frequencies for Raman E8 2 2g ^ ^ and E2g2 (squares) and infrared-active E8 and E1u (circles) modes in graphiteferric 1u chloride compounds. The solid points are identied with graphite interior layers and the open points with graphite bounding layers. The solid lines represent a least square t to the experimental points for n 3 and indicate a similar (1=n) dependence for all observed modes. The E2g2 and E1u mode frequencies for pristine graphite are given as 1=n ! 0. (From the work of Underhill et al. 1979.)

Figure 87. The reciprocal stage (1=n) dependence of the dipole moment hpa i in graphite ^ FeCl3 compounds for the infrared E8 and E1u modes normalized to that for the E1u 1u mode in pristine graphite, as obtained by Underhill et al. (1979). Of interest is the ^ signicantly smaller (1=n) dependence of hpa i for the E1u mode involving a graphite bounding layer as compared with hpa i for the E8 mode which involves only graphite 1u interior layers.

tions appearing in (4.29), including ecarriers due to free carriers and einterband due to interband transitions. In carrying out the lineshape analysis elattice is modelled by a set of oscillators. As described below, infrared structures associated with graphitic lattice modes exhibit both similarities and di erences according to whether the intercalant is a donor or an acceptor. Infrared reectivity results typical of acceptor compounds are shown in gure 85 for the graphiteFeCl3 system. Traces are shown for compounds with stage n 2, 4, 6 and 11 in the frequency range 1500 < ! < 1700 cm1 . The striking di erence in lineshape between the trace for graphite (labelled HOPG), for the stage 11 compound and for the lower stage compounds n 6, is due to the

Intercalation compounds of graphite

167

increasing magnitude of "carriers with decreasing stage and the relatively smaller contribution of "interband when the Fermi level falls below the K-point E3 band extremum for acceptor compounds (Dresselhaus and Dresselhaus 1977). The results of a lineshape calculation yielding lattice mode frequencies, linewidths and oscillator strengths are given in gure 85 by the dotted curves and yield one IR-active mode for stage 2 and two IR-active modes for stages n 3. We note that although cursory inspection of the data suggests a single IR-active mode, analysis of the observed lineshape requires the superposition of two unresolved Lorentzian infrared structures (Underhill et al. 1979). Analysis of the dependence of the mode frequencies (gure 86) and oscillator strengths (gure 87) on reciprocal stage (1=n) yields an identication of the observed structures with graphite interior layers (E8 u ) and 1 ^ graphite bounding layers (E1u ). The mode for which the peak frequency and oscillator strength versus (1=n) extrapolates to the graphite values as 1=n ! 0 is identied with the graphite interior layers. This mode is also not found for the stage 2 compound, consistent with the interpretation. The frequency for the other dominant mode versus (1=n) does not extrapolate to the E1u graphite mode and furthermore exhibits an oscillator strength with a weak dependence on (1=n). This mode is therefore identied with the graphite bounding layers since the environment of the carbon atoms on the graphite bounding layer is similar for stages n 3. The infrared spectra observed for other acceptor compounds, such as AlCl3 , Br2 , IBr and ICl, are qualitatively similar to the results obtained for the FeCl3 system. For all acceptor intercalants, a general increase in oscillator strength with increasing intercalate concentration is observed and this e ect can be understood physically on the basis of the following argument. Since the intercalation process introduces a charge density gradient in the c-direction (Pietronero et al. 1978, 1979), the magnitude of the dipole associated with the E1u mode is enhanced as shown in gure 87. Since the gradient in charge density is largest near the graphite bounding ^ layer, the E1u mode is expected to have a greater oscillator strength than the interior layer E8 mode, in agreement with observations. Furthermore, since the gradient in 1u charge density near the graphite bounding layer is only weakly dependent on stage for n 3, the oscillator strength for this mode is expected to be approximately independent of stage, also in agreement with observations. The IR-active mode for n 2 involves carbon atom vibrations on two neighbouring bounding layers, which is a di erent arrangement than that which occurs in the higher stage compounds. We note that IR-active modes have been reported for well-staged stage 2 acceptor compounds with the intercalants FeCl3 and AlCl3 , in contrast with the alkali metal C8 X compounds (X K, Rb, Cs) where no IR-active modes are found for stage 2 compounds. On the other hand, no IR-active modes are found for either acceptor or donor stage 1 compounds, in agreement with group theoretical considerations which predict that a single graphite layer cannot give rise to odd-parity modes. In this connection, the compelling experiment for stage 1 acceptors is for the graphiteAlCl3 system, which can readily be prepared as a single-staged n 1 compound, in contrast with graphiteFeCl3 , which does not readily form single-staged n 1 material, or graphiteBr2 , which does not form stage 1 compounds at all. It is therefore signicant that rst stage AlCl3 shows no IR-active modes (Gualberto et al. 1980). As in the case of Raman-active modes in acceptor compounds, the IR-active modes for acceptor intercalants upshift as a function of reciprocal stage (1=n), and for the graphiteFeCl3 system the upshifts of all Raman-active and IR-active modes

168

M. S. Dresselhaus and G. Dresselhaus

Figure 88. Room temperature c-face infrared reectivity spectra in the frequency range 1400 ! 1700 cm1 for stage 1, 3 and 6 graphiteRb compounds and for pristine graphite. The t of the lineshape analysis to the experimental data is shown by the dotted curves. No IR active modes are observed for stage 1 and 2 compounds with K, Rb and Cs. The IR-active E1u mode for graphite is shown in the inset. (From the work of Leung et al. 1980a.)

exhibit a similar (1=n) dependence (Underhill et al. 1979). Though the other acceptor compounds also exhibit a general frequency upshift of the IR-active modes, the behaviour does not appear to be as simple as for graphiteFeCl3 (Gualberto et al. 1980). This frequency upshift as a function of (1=n) has been attributed to a strain mechanism, implying an in-plane lattice contraction for the acceptor compounds. Infrared spectra observed in the alkali metal compounds with K, Rb and Cs are shown in gure 88 and are analysed in terms of an E8 mode associated with the 1u ^ graphite interior layers and a E1u mode associated with the graphite bounding layers ^ (Leung et al. 1980a). As also occurs in the acceptor compounds, the E1u mode lies lower in frequency than the E8 u mode. Whereas neither stage 1 donor nor acceptor 1 compounds exhibit IR activity, the stage 2 alkali metal (K, Rb and Cs) donor compounds exhibit no IR-active modes in contrast to the stage 2 acceptors. The ^ frequency separation betwen the E8 and E1u modes is signicantly larger in the 1u donor compounds than in the acceptors, indicating a large charge density in the graphite bounding layers in the case of the donors, resulting in an increased interaction between graphite bounding and intercalate layers, and consequently a ^ greater relative downshift of the E1u mode. Also in contrast with the behaviour in the

Intercalation compounds of graphite

169

Figure 89. Longitudinal lattice modes along the kz -axis for the rst stage compound C8 Rb as measured by Ellenson et al. (1977) using inelastic neutron scattering. The plot is made in terms of the dimensionless wave vector c kz Ic =2 where Ic is the c-axis lattice constant. Results are shown at 290 and 747 K. On the basis of observed neutron di raction patterns, the intercalate layer is identied to be liquid-like or disordered at 747 K. Also shown in this gure are results for graphite at 290 K (dotted curve). The dimensionless wave vector for graphite is in units of =c0 whereas for C8 Rb the unit is =Ic .

acceptors, the IR-mode frequencies downshift as a function of reciprocal stage (1=n), consistent with the behaviour of the Raman-active modes for the donors as a function of (1=n). These frequency downshifts are also consistent with a strain mechanism in so far as the in-plane lattice constants for graphiteK have been observed by Nixon and Parry (1969) to increase as a function of (1=n). The IR oscillator strength yields the dipole moment for the E1u mode and in turn the dynamic e ective charge associated with that mode. Since the graphite bounding and interior layer modes occur at di erent frequencies, the dynamic e ective charges associated with each mode can be determined independently. Analysis of the IR spectra for donor and acceptor compounds indicates that more dynamic e ective charge is associated with all E1u modes for donor compounds than for acceptor compounds of similar stage. The analysis further shows that relatively more e ective charge resides in the graphite interior layers for the donor compounds than for acceptor compounds of the same stage, in agreement with the interpretation of magnetoreection (Mendez et al. 1980a) and electrical conductivity (Dresselhaus and Leung 1979) results. The relation between the dynamic e ective charge and the charge available for electrical transport, however, requires further elucidation. 5.5. Inelastic neutron scattering studies Inelastic neutron scattering is a particularly useful tool for the study of phonon modes because large momentum transfers can occur when a neutron is scattered at a lattice site, creating or absorbing a phonon. Thus the entire Brillouin zone can in principle be probed by neutron scattering. However, the requirement for large single

170

M. S. Dresselhaus and G. Dresselhaus

crystals (1 mm3 ) imposes severe restrictions on the utility of this technique for graphite and its intercalation compounds, for which single crystal samples of such large sizes cannot be obtained. It has, however, been possible to study the phonon dispersion relations of graphite (Dolling and Brockhouse 1962, Nicklow et al. 1972) and of intercalated graphite (Ellenson et al. 1977) using samples based on highly oriented pyrolytic graphite (HOPG), which has a high degree (better than 18) of caxis alignment. Analysis of inelastic neutron scattering data by Nicklow et al. (1972) has provided phonon dispersion curves for the two lowest frequency phonon branches along M (see gure 76). Near the M-point the two lowest frequency modes are nearly degenerate at 466 cm 1 and correspond to z-axis displacements. In the analysis of the neutron scattering data, a Bornvon Karman model was used to t the measured low frequency phonon branches. Since the inelastic neutron scattering measurements have been limited to frequencies less than 470 cm1 , modelling procedures have been used to infer the phonon dispersion relations for the branches that are not measured experimentally. Such a model was developed by Nicklow et al. (1972), but was later modied by Maeda et al. (1979) and by Nemanich and Solin (1979) after the A2u mode was identied experimentally (Nemanich et al. 1977a, b). The dispersion curves of gure 76 are the results of the model developed by Maeda et al. (1979). Inelastic neutron scattering has also been successfully applied to the study of phonon dispersion relations for graphite intercalation compounds. Ellenson et al. (1977) have applied this technique to study the low frequency (9200 cm1 ) branches of C8 Rb and their results are shown in gure 89 at room temperature and at 747 K. In this gure, the phonon dispersion relations along A for pristine graphite are also shown, and these results are in good agreement with previous work by Nicklow et al. (1972). Of particular interest are the similarities and di erences between the dispersion relations for graphite and for C8 Rb along A, which can be understood in the following way. The basic unit cell along the c-direction in both cases involves two layers. Since C8 Rb has di erent atomic species on adjacent layers, two modications to the graphite dispersion relations are expected. First, since the structure factor for C8 Rb no longer vanishes at the A-point there is no symmetry requirement for a degeneracy of the LA and LO branches at the A-point on the Brillouin zone boundary. Thus a mode splitting is expected at the A-point. Because the mass of a Rb layer (mass proportional to atomic weight 85.5) is somewhat lower than the mass of a graphite layer (8 12 96), the LO branch for the intercalation compound is expected to be upshifted from the corresponding mode in pure graphite. That this upshift does not have a simple M 1=2 mass dependence is indicative of stronger interactions between the graphitic and intercalate layers in rst stage alkali metal compounds than between A and B layers of pristine graphite. Consistent with this interpretation are many of the observed electronic properties, including a large increase in the c-axis conductivity and a decrease in the anisotropy of the electrical conductivity as a function of reciprocal stage. To obtain information on the intercalate stacking arrangement, elastic neutron scattering experiments are required. In addition to study of the low frequency longitudinal branches, Ellenson et al. (1977) have also investigated the low frequency transverse branches using inelastic neutron scattering techniques. Of particular interest is their observation of a large temperature dependence of the TA mode near the A-point in the Brillouin zone

Intercalation compounds of graphite

171

(c 1=2), the mode corresponding to a translation of the Rb layer relative to the graphite layer. This mode would be expected to soften with increasing temperature, and to become overdamped at the orderdisorder transition of 747 K, above which the Rb layer has been reported to become liquid-like and disordered. These initial inelastic neutron scattering studies indicate great promise for this technique for detailed exploration of the phonon dispersion relations in graphite intercalation compounds.

Technological applications of intercalated graphite 6.1. Introduction The diverse properties of graphite intercalation compounds suggest a number of applications of these materials to mechanical, electrical and chemical engineering processes. Though graphite intercalation compounds are not yet widely used in industrial applications, the parent compound graphite is an extensively used industrial material. Intercalation o ers the interesting possibility of both major and minor modications to the properties of graphite. An attractive feature of graphite intercalation compounds is that the host material graphite can be prepared on a commercial scale from readily available raw materials. The preparation of intercalation compounds could also be carried out with reasonably modest equipment on a commercial basis. For these reasons, industrial use for graphite intercalation compounds should nd ready acceptance provided suitable applications can be identied. Some of the possible applications of graphite intercalation compounds that have been made or proposed are mentioned here though an extensive review of this subject has not been attempted. In particular this section contains a brief review of battery and electrode materials (section 6.2), chemical catalysis and catalytic applications (section 6.3), conductivity applications (section 6.4), carbon bres (section 6.5) and other applications (section 6.6). It should be noted that some of the uses mentioned in the other applications section may in time prove to be of greater industrial importance than those reviewed here in more detail. 6.2. Battery and electrode materials The use of graphite intercalation compounds for battery materials and electrodes has been discussed and specic applications have been attempted. The requirements of graphite intercalation compounds for application as electrodes in high density batteries have been considered in detail by Armand and Touzain (1977), and include the following. The cathode material must be rechargeable and exhibit structural integrity during both the charging and discharging cycles. The chemicalelectrical energy conversion process requires ion exchange with the electrolyte and electron exchange with the external circuit. To achieve high e.m.f. values, an intercalant with high electron a nity must be selected. Graphite intercalation compounds are attractive for cathode materials because of their potential for low weight relative to energy density and because the use of an intercalant species for ion exchange addresses the need for a wide range of accessible non-stoichiometries . Graphite uorides are of special interest because of their low equivalent weight and large free energy changes (Whittingham 1975) and have been studied for use with a lithium anode and a non-aqueous electrolyte. During the discharge process

6.

172

M. S. Dresselhaus and G. Dresselhaus

Li ions are supplied by the electrolyte and electrons by the external circuit, reducing the cathode which is an intercalation compound, to yield xLi xe hCn AXy i ! hLix Cn AXy i: The most serious technical problems associated with batteries based on intercalated graphite electrodes concerns the metastability of the electrolyte near the anode or cathode surfaces (where the graphite could act as a catalyst), and the leaching of intercalant from the edge planes by electrolyte activity. On the other hand, the large number of possible cathode materials (Armand and Touzain 1977) have attracted interest in investigating this applications area, with particular emphasis being given to the development of suitable and compatible electrolyte materials. The high conductivity of intercalated graphite electrodes is also an attractive materials property for this application. 6.3. Chemical catalytic applications A survey of the use of graphite intercalation compounds in chemical reactions has recently been given by Whittingham and Ebert (1979), and several examples of such reactions are summarized below. Although the specic role of the graphite intercalation compounds in fostering chemical reactions has not been identied, a number of examples of chemical catalytic activity have been given in the literature. For example, graphitealkali metalmetal chloride compounds have been used by Ichikawa et al. (1972a) and by Postnikov et al. (1975) to synthesize NH 3 from N 2 and H2 , and by Naito et al. (1972) and by Mashinskii et al. (1976) to form hydrocarbons from H2 and CO. Graphitealkali metal compounds have also been reported by Ichikawa et al. (1972b) to break hydrogencarbon bonds leading to hydrogen-deuterium exchange. Beguin and Setton (1976) have used graphitealkali metal compounds to promote organic reactions such as the production of biphenyl upon reacting benzene with tetrahydrofuran . The use of graphitebromine intercalation compounds for organic bromination has been considered by Page-Lecuyer et al. (1973), and this concept should be applicable to halogenation reactions using other graphitehalogen or halogen salt compounds. The promotion of organic reactions such as the esterication of ammonia in graphitesulphuric acid salts has been considered by Setton (1977) . Use of graphitealkali metal compounds for catalyzing polymerization reactions has been made by Gole (1977) and Panayotov et al. (1972, 1975), as for example the promotion of the polymerization of styreneisoprene into an alternating copolymer using a graphitelithium catalyst. In most cases, the distinct chemical role of the intercalant within the graphite intercalation compounds as opposed to its role as an isolated intercalant species has not been clearly identied. In an e ort to identify the catalytic activity of the intercalant in chemical reactions, Setton (1977) and Ebert and Selig (1977) have respectively demonstrate d that the chemical shift in the nuclear magnetic resonant eld and the asymmetry of the electron spin resonance lineshape could provide sensitive tools for monitoring the oxidation state of intercalants as chemical reactions proceed. 6.4. Conductivity applications The high room temperature conductivities achieved with such intercalation compounds as graphiteAsF5 (a 6:2 105 cm1 for a stage 2 compound) and graphite SbF5 (a 5:9 105 cm1 ) o er promise for practical applications

Intercalation compounds of graphite

173

of certain acceptor compounds as electrical conductors, where in comparison a for Cu is 5:8 105 cm1 (see section 4.2 and table 16). For most of the applications that have been envisaged, wires have to be fabricated. In this connection, two approaches have been taken, namely (1) the fabrication of composite wires in which the intercalation compound is placed in the core of a copper sheath, and (2) the use of bundles of intercalated carbon bres (see section 6.5). Considerable interest in the composite wire approach was stimulated by the report by Vogel (1977) of the fabrication of a composite wire consisting of a graphiteSbF5 core, embedded in a copper sheath, and exhibiting room temperature conductivity greater than that of copper. The copper sheath was used to provide encapsulation of a high conductivity intercalation compound, since compounds exhibiting the highest electrical conductivities tend to be highly reactive in air and to desorb readily. The copper sheath is also useful for several steps in the fabrication process, which started with the insertion of graphite powder and SbF5 into a copper tube which was then sealed. The loaded copper tube was then swaged or drawn into a wire, which also served to compact the core material. Heat treatment of the composite wire produced an intercalation compound in the core material, with a high degree of preferential orientation of the c-axis in the radial direction of the wire. For a wire, consisting of 2/3 (by volume) of Cu with the remaining 1/3 as the intercalated graphiteSbF5 core, Vogel (1977) reported a composite conductivity of 5:95 105 cm1 . Because this conductivity was comparable to that of copper (slightly greater in fact), considerable interest in these composite wires was stimulated. However, later work on a similar type of composite wire by Singhal (1980) failed to conrm these high composite conductivity values. Although Singhals graphiteSbF5 core material showed the preferential orientation of the c-axis, the maximum composite conductivity that Singhal obtained was 4:5 10 5 cm1 , from which he concluded that the intercalate core did not contribute signicantly to the composite conductivity. Although Vogels composite wire approach o ers considerable potential, it remains to be demonstrated that intercalated graphite cores enhance the composite wire conductivity or that such composites could be produced on a commercial basis with properties superior to conventional wires. Concurrently, considerable work is in progress on intercalated graphite bres (section 6.5) in an e ort to assess the potential of intercalated bre bundles for application as electrical conductors. Other promising applications areas for intercalated graphite involve exploitation of the high electrical anisotropy of these materials (see table 17). In this class of applications, high conductivity in one direction could be realized with poor conductivity perpendicular to this direction.

6.5. Carbon bres Carbon bres are materials of great commercial interest because of their high tensile strength (3 GPa) and high Youngs modulus (400 GPa). A good general survey of the carbon bre literature is found in the Proceedings of the Internationa l Carbon Fibers Conference of 1971 and 1974. High bre strength (1/2 of the theoretical crystalline graphite modulus in the basal plane) is achieved by aligning the graphitic basal planes parallel to the bre axis, and thereby exploiting the strong in-plane carboncarbon bonding in graphite.

174

M. S. Dresselhaus and G. Dresselhaus

Carbon bres are considerably less ordered than highly oriented pyrolytic graphite (HOPG), though the degree of ordering is strongly inuenced by the precursor bre that is graphitized to form the carbon bre and by the heat treatment temperature. Most commercial bres are prepared from graphitization of polyacrylonitrile (PAN), pitch or rayon bres. The precursor material plays a dominant role in determining the microstructure of the bre. Increasing the heat treatment temperature results in (1) an increase of the mean crystalline size of the ordered regions within the bre and (2) a decrease in the degree of misorientation of the basal planes with respect to the bre axis. Heat treatment temperatures 25008C yield more highly oriented bres and these are called graphite bres, in contrast to bres experiencing lower heat treatment temperatures. In the PAN-based bres the basal planes have a preferred orientation that is circumferential (in ribbons) with respect to the bre axis, with a relatively high degree of alignment near the skin of the bre and decreasing alignment as the core of the bre is approached. In contrast, in the pitchbased bres, the basal planes tend to be arranged radially. Two parameters characterize the dimensions of the ordered region: Lc and La for the width and length respectively of ordered ribbon regions. Typical ordered ribbon widths for highly graphitic PAN-based bres are Lc 100 A and the ordered lengths are signicantly longer, La 250 A. For both PAN- and pitch-based bres, higher electrical conductivity is achieved with increased structural ordering, increased bulk modulus, all of which result from higher heat treatment temperatures (Robson et al. 1972, 1973, Bright and Singer 1979). From a technological point of view, the commercial process for the fabrication of graphite bres is highly developed. Because of their great mechanical strength and their availability in the convenient form of bre bundles and yarns, graphite bres have been considered for use as electrical conductors, if their conductivities could be su ciently increased to become competitive with typical metal conductors. Intercalation o ers one method for achieving such high conductivities, and intercalated bres have attracted interest and attention in this context. The intercalation of graphite bres with both donor and acceptor intercalants has been demonstrated. Substantial intercalant uptake has been achieved, and in some cases concentrations comparable with stage 1 have been reported, though the extent to which staging occurs in intercalated graphite bres is still an open question. The potential of intercalated graphite bres for use as practical conductors was demonstrated by Herinckx et al. (1972) and Herinckx (1973) who showed that intercalation with potassium resulted in a large increase in electrical conductivity (by a factor of as much as 20) without signicant degradation of the mechanical properties. Intercalation of carbon bres with acceptors such as Br2 , AlCl3 and ICl has also been carried out successfully by Hooley (1975) and Hooley and Dietz (1978). In general, the threshold pressure for intercalation is signicantly greater in bres than in HOPG host materials, and the greater the degree of graphitization and crystallographi c ordering of the bre, the lower the threshold pressure for intercalation. Thus for application as electrical conductors, emphasis has been given to highly graphitized host bres which have a higher conductivity prior to intercalation and which also can be more readily intercalated. Also for conductivity applications, intercalation with acceptors such as HNO 3 (Vogel et al. 1976) and other acids (Besenhard and Fritz 1978) have been favoured because of their e ectiveness in increasing the conductivity of HOPG. Increasing attention is also being given to the degree of graphitization of the host bre for conductivity applications of intercalated

Intercalation compounds of graphite

175

bres. With presently available commercial bres, both the host material and the intercalated bres exhibit a decrease in conductivity with decreasing temperature, in contrast with the behaviour of HOPG and intercalation compounds based on HOPG. Thus structural defects in the bres dominate the mobility of presently available commercial bres. Important advances have recently been made by Endo et al. (1976, 1977, 1979) and Oberlin et al. (1976) in the preparation of a highly graphitized bre with signicantly increased crystallographic ordering. These bres are prepared by pyrolyzing a mixture of benzene and hydrogen at 11008C and graphitizing the bres by heat treatment at 280030008C. The ordering of the graphite basal planes is in concentric layers around the bre axis. These bres have been successfully intercalated with Br2 and HNO3 , and show a high degree of long-term stability, attributed to the graphite shell structure which inhibits desorption. In contrast with the behaviour shown by commercial bres, these highly graphitized pristine and intercalated bres show an increase in conductivity as the temperature is lowered. Also for these highly graphitized bres, intercalation does not signicantly degrade the mechanical properties. The intercalation of such bres could provide an attractive starting point for electrical conductivity applications. Present commercial use of carbon bres exploits their high strength to weight ratio in a variety of structural applications. Modications of the mechanical properties through intercalation are not expected to result in increased strength and modulus. On the other hand, little attention has so far been directed toward a detailed study of the e ect of intercalation on the mechanical properties. Studies on lattice properties of intercalated graphite (sections 5.3 and 5.4) suggest that acceptor intercalation would sti en bres whereas donor intercalation would soften bres. 6.6. Other applications A number of other applications are suggested by the unusual properties of intercalated graphite. Some of these properties and suggested applications are given below. Intercalate o ers a convenient method for the preparation of materials with large lattice constants (Ic 40 A for a dilute intercalation compound). This property can be utilized in such applications as in a low energy neutron monochromator. For this application high staging delity would be required. Pristine graphite is itself a material with a highly anisotropic electrical and thermal conductivity. This anisotropy in electrical conductivity can be further enhanced by intercalation with suitable intercalants. If increased anisotropy in the thermal conductivity could be achieved, it could be utilized in the construction of heat shields. The increased electrical conductivity could be utilized in the fabrication of infrared polarizers. The use of graphite as an infrared polarizer has already been demonstrated (Rupprecht et al. 1962). Since intercalation with acceptors enhances the electrical anisotropy ratio a = c , the use of intercalated graphite for infrared polarizers should result in improved performance. Intercalation can be used as a method for shifting the plasma frequency of a host material. This property could perhaps be utilized in the fabrication of optical lters, which could be tuned by changing the intercalate concentration through application of a voltage. The use of an applied voltage to change the plasma frequency has already been demonstrated by Puger et al. (1979) who exploited the colour change accompanying a stage change (gold for stage 1 and blue for stage 2 alkali metal

176

M. S. Dresselhaus and G. Dresselhaus

compounds) to make an electrochemical device for optical displays. The colour change associated with a stage transformation is due to the stage dependence of the plasma frequency and of the free carrier concentration. In the preparation of graphite intercalation compounds it is important to cool the ampoule from the side away from the intercalated sample to prevent exfoliation. On the other hand, exfoliation produced by a rapid cooling of intercalated graphite can be put to use in the preparation of very thin graphite sheets, or graphite material of high surface area. Such high surface area material is prepared commercially. When slightly compacted in sheet form it is called grafoil. A typical value of the surface area per unit weight is 100 m2 =g. Use of di erent intercalates could perhaps increase the value of the surface area per unit weight. The intercalation process allows for the storage of intercalants between the graphite layers. Subsequent recovery of the intercalant by desorption, and the use of intercalantdeintercalation cycles would be part of a practical storage/recovery system. In this connection it is of interest to note that higher intercalate concentrations can be achieved in intercalation compounds. For example, for stage 1 heavy alkali metal compounds the intercalant has a higher spatial density in the intercalation compound than in the free metal (see table 8). Of more practical interest is the use of intercalated graphite for hydrogen storage-release cycles, employing the hydrogenatio n of stage 1 alkali metal compounds to form the ternary compound C16 K2 H4=3 (Lagrange et al. 1978). The a nity of graphite intercalation compounds for H2 O and O2 could also be utilized in gas purication cycles. Although limited practical applications of graphite intercalation compounds have so far been made, this class of materials does exhibit a number of unusual properties which have signicant potential for commercial exploitation. Acknowledgements The authors wish to thank the members of our research group, both present and past for help with all aspects of this review article, which started as lecture notes for use by our research group. In particular we acknowledge S. Y. Leung, C. Underhill, R. Gerut and E. Shaw who worked long hours above and beyond the call of duty to complete the tedious task of preparing the manuscript. We wish to thank numerous colleagues who made their work available to us in preprint form. We wish to particularly thank Dr. P. Lauginie and Professor J. M. Thomas for valuable suggestions. We also gratefully acknowledge the nancial support from NSF (grant No. DMR 7810858), ONR (grant No. N0001477C0053), and AFOSR (grant No. 773391), which made the writing of this review article possible. References
Alexander, M. G., Goshorn, D. P., Onn, D. G., Gue rard, D. L., Lagrange, P., and El Makrini, M., 1980, Synth. Metals, 2, 203. Anderegg, H., Feuerbacher, B., and Fitton, B., 1971, Phys. Rev. Lett., 26, 760. Armand, M., and Touzain, P., 1977, Mater. Sci. Engng, 31, 319. Aronson, S. 1963, J. inorg. nucl. Chem., 25, 907. Aronson, S., Frishberg, C., and Frankl, G., 1971, Carbon, 9, 715. Aronson, S., Salzano, F. J., and Bellafiore, D., 1986, J. chem. Phys., 49, 434. Asher, R. C., 1959, J. inorg. nucl. Chem., 10, 238. Avogadro, A., Bellodi, G., Borghesi, G., Samoggia, G., and Villa, M., 1977, Nuovo Cim. B, 38, 403.

Intercalation compounds of graphite

177

Avogadro, A., and Villa, M., 1977, J. chem. Phys., 66, 2359. Ayache, C., 1980, Physica B, 99, 509. Ayache, C., Bonjour, E., Laguier, R., and Fischer, J. E., 1980, Physica B, 99, 547. Bach, B., 1971, C. r. hebd. seance Acad. Sci., Paris B, 272, 666. Bach, B., Evans, E. L., Thomas, J. M., and Barber, M., 1971, Chem. Phys. Lett., 10, 547. rold, A., 1963, C. r. hebd. Seance. Acad. Sci. Paris, 257, 681; 1968, Bull. Bach, B., and He Soc. chim. Fr., 5, 1978. Bak, P., and Domany, E., 1979, Phys. Rev. B, 20, 2818. Balzarotti, A., and Grandolfo, M., 1968, Phys. Rev. Lett., 20, 9. Barker, J. A., and Croft, R. C., 1953, Aust. J. Chem., 6, 302. Bassani, F., and Pastori Parravicini, G., 1967, Nuovo Cim. B, 50, 95. Basu, S., Zeller, C., Flanders, P., Fuerst, C. D., Johnson, W. D., and Fischer, J. E., 1979, Mater. Sci. Engng, 38, 275. Batallan, G., Bok, J., Rosenman, I., and Melin, J., 1978, Phys. Rev. Lett., 41, 330. Batallan, G., Rosenman, I., Simon, C., Furdin, G., and Fuzellier, H., 1980, Physica B, 99, 411. Beguin, F., and Setton, R., 1975, Carbon, 13, 293; 1976, J. chem. Soc. chem. Commun., p. 611. Bellodi, G., Borghesi, A., Guizzetti, G., Nosenzo, L., Reguzzoni, E., and Samoggia, G., 1975, Phys. Rev. B, 12, 5951. Bender, A. S., and Young, D. A., 1971, Phys. Stat. Sol. (b), 47, K95; 1972, J. Phys. C, 5, 2163. Berker, A. N., Kambe, N., Dresselhaus, G., and Dresselhaus, M. S., 1980, Phys. Rev. Lett., 45, 1452. Berney, C. V., and Underhill, C., 1980, Synth. Metals, 2, 149. Besenhard, J. O., and Fritz, H. P., 1978, Z. Naturf. (b), 33, 737. Bewer, G., Wickmann, N., and Boehm, H. P., 1977, Mater. Sci. Engng, 31, 73. rold, A., 1974, Bull. Soc. chim. Fr., 12, 2715. Billaud, D., and He Billaud, D., He rold, A., and Vogel, F. L., 1980, Mater. Sci. Engng, 45, 55. Blackman, L. C. F., Mathews, J. F., and Ubbelohde, A. R., 1960, Proc. R. Soc. A, 256, 15. Blinowski, J., Nguyen, H. H., Rigaux, C., Vieren, J. P., LeToullec, R., Furdin, G., rold, A., and Melin, J., 1980, J. Phys., Paris, 41, 47. He Bok, J., Batallan, F., and Rosenman, I., 1978, Proceedings of the International Conference on the Application of High Magnetic Fields to Semiconductor Physics, edited by J. R. Ryan (Oxford: Clarendon Press), p. 48. Bonnetain, L., Touzan, P., and Hamwi, A., 1977, Mater. Sci. Engng, 31, 45. Bottomley, M., Parry, G. S., and Ubbelhode, A. R., 1963a, Proc. R. Soc. A, 278, 291. Bottomley, M., Parry, G. S., Ubbelohde, A. R., and Young, D. A., 1963b, J. chem. Soc., p. 5674. Briggs, A., Moran, M. J., Holzwarth, N. A. W., and Fischer, J. E., 1980, Bull. Am. phys. Soc., 25, 299. Brillson, L. J., Burstein, E., Maradudin, A. A., and Stark, T., 1971, Proceedings of the International Conference on Semimetals and Narrow Gap Semiconductors, edited by D. L. Carter and R. T. Bate (New York: Pergamon Press), p. 187. Bright, A. A., and Singer, L. S., 1979, Carbon, 17, 59. Brockhouse, B. N., Hautecler, S., and Stiller, H., 1963, Interaction of Radiation with Solids, edited by R. Strumane (Amsterdam: North-Holland), p. 580. chner, U., 1977, Phys. Stat. Sol. (b), 81, 227. Bu Cahill, J. E., and Leroi, G. E., 1966, J. chem. Phys., 51, 4514. Campbell, L. E., Montet, L. G., and Perlow, G. J., 1977, Phys. Rev. B, 15, 3318. Carver, G. P., 1970, Phys. Rev. B, 2, 2284. Caswell, N., and Solin, S. A., 1978, Solid St. Commun., 27, 961; 1979, Phys. Rev. B, 20, 2557. Caswell, N., Solin, S. A., Hayes, T. M., and Hunter, S. J., 1980, Physica, B, 99, 463. Chieu, T. C., 1980, M.S. Thesis, MIT, Cambridge, Massachusetts, and unpublished results. Chinn, M. D., and Fain, S. C., Jr., 1977, Phys. Rev. Lett., 39, 146. Chung, D. D. L., 1977, Ph.D. Thesis, MIT, Cambridge, Massachusetts.

178

M. S. Dresselhaus and G. Dresselhaus

Chung, D. D. L., and Dresselhaus, M. S., 1976, Solid St. Commun., 19, 227; 1977a, Physica B, 89, 131. Chung, D. D. L., Dresselhaus, G., and Dresselhaus, M. S., 1977b, Mater. Sci. Engng, 31, 107. Clarke, R., Caswell, N., and Solin, S. A., 1979a, Phys. Rev. Lett., 42, 61. Clarke, R., Caswell, N., Solin, S. A., and Horn, P. M., 1979b, Phys. Rev. Lett., 43, 2018; 1980, Physica B, 99, 457. Conard, J., and Estrade, H., 1977, Mater. Sci. Engng, 31, 173. Conard, J., Estrade, H., Lauginie, P., Fuzellier, H., Furdin, G., and Vosse, R., 1980, Physica B, 99, 521. Cooper, J. D., Woore, J., and Young, D. A., 1970, Nature, Lond., 225, 721. Corbato, F. J., 1956, Ph.D. Thesis, MIT, Cambridge, Massachusetts; 1959, Proceedings of the Third Conference on Carbon (New York: Pergamon Press), p. 173. Coulson, C. A., 1947, Nature, Lond., 159, 265. Cowley, J. M., and Ibers, J. A., 1956, Acta crystallogr., 9, 421. Crespin, M., Tchoubor, D., Gatineau, L., Beguin, F., and Setton, R., 1977, Carbon, 15, 303. Crost, R. C., 1956, Aust. J. Chem., 9, 184; 1960, Q. Rev., 14, 1. Culik, J. S., and Chung, D. D. L., 1979, Mater. Sci. Engng, 37, 213. Daniels, J., Festenberg, C. V., Raether, H., and Zeppenfeld, K., 1970, Springer Tracts in Modern Physics, Vol. 54 (Berlin: SpringerVerlag), p. 77. Daumas, N., and He rold, A., 1969, C. r. hebd. Seanc. Acad. Sci., Paris C, 268, 373; 1971, Bull. Soc. chim. Fr., 5, 1598. de Gennes, P. G., 1974, Physics of Liquid Crystals (Oxford University Press). Delhaes, P., 1977, Mater. Sci. Engng, 31, 225. rard, D., and Herold, A., 1976, Delhaes, P. G., Rouillon, J. C., Manceau, J. P., Gue J. Phys. Lett., Paris, 37, 127. Dingle, R., 1975, Festiko rper Probleme, Vol. 15, Advances in Solid State Physics (Braunschweig: Pergamon-Vieweg), p. 21. Di Salvo, F. J., and Fischer, J. E., 1978, Solid St. Commun., 28, 71. Di Salvo, F. J., Safran, S. A., Haddon, R. C., Waszczak, J. V., and Fischer, J. E., 1979, Phys. Rev. B, 20, 4883. Doezema, R. E., Datars, W. R., Schaber, H., and Van Schyndel, A., 1979, Phys. Rev. B, 19, 4224. Dolling, G., and Brockhouse, B. N., 1962, Phys. Rev., 128, 1120. Doni, E., and Pastori Parravicini, G., 1969, Nouvo Cim. B, 64, 117. Dowell, M. B., 1977, Mater. Sci. Engng, 31, 129. Dowell, M. B., and Badorrek, D. S., 1978, Carbon, 16, 241. Dresselhaus, G., 1974, Phys. Rev. B, 10, 3602. Dresselhaus, G., and Dresselhaus, M. S., 1965, Rev. Rev. A, 140, 401; 1977, Mater. Sci. Engng, 31, 235. Dresselhaus, G., Dresselhaus, M. S., and Mavroides, J. G., 1966, Carbon, 4, 433. Dresselhaus, G., and Leung, S. Y., 1980, Solid St. Commun., 35, 819; 1981, Physica (to be published). Dresselhaus, G., Leung, S. Y., Shayegan, M., and Chieu, T. C., 1980, Synth. Metals, 2, 321. Dresselhaus, M. S., and Dresselhaus, G., 1979, Physics and Chemistry of Materials with Layered Structures, Vol. 6, edited by F. Levy (Dordrecht: Reidel), p. 497. Dresselhaus, M. S., Dresselhaus, G., Eklund, P. C., and Chung, D. D. L., 1977b, Mater. Sci. Engng, 31, 141. Dresselhaus, M. S., Dresselhaus, G., and Fischer, J. E., 1977a, Phys. Rev. B, 15, 3180. Dresselhaus, M. S., and Leung, S. Y., 1979, Extended Abstracts of the 14th Biennial Conference on Carbon, Pennsylvania State University, p. 496. Dresselhaus, M. S., and Mavroides, J. G., 1964, IBM Jl Res. Dev., 8, 262. Dworkin, A., and Ubbelohde, A. R., 1978, Carbon, 16, 292. Dyson, F. J., 1955, Phys. Rev., 98, 349. Dzurus, M. L., and Hennig, G. R., 1957a, J. Am. chem. Soc., 79, 1051; 1957b, ibid., 79, 5897.

Intercalation compounds of graphite

179

Dzurus, M. L., Hennig, G. R., and Montet, G. L., 1959, Proc. Fourth Carbon Conf. (New York: Pergamon), p. 165. Eberhardt, W., McGovern, I. T., Plummer, E. W., and Fischer, J. E., 1980, Phys. Rev. Lett., 44, 200. Ebert, L. B., 1976, A. Rev. Mater. Sci., 6, 181. Ebert, L. B., Huggins, R. A., and Brauman, J. I., 1974, J. chem. Soc. chem. Commun., p. 924. Ebert, L. B., and Selig, H., 1977, Mater. Sci. Engng, 31, 177. Eeles, W. T., and Turnbull, J. A., 1965, Proc. R. Soc. A, 283, 179. Eklund, P. C., Dresselhaus, G., Dresselhaus, M. S., and Fischer, J. E., 1977, Phys. Rev. B, 16, 3330; 1980a, Phys. Rev. B, 21, 4705. Eklund, P. C., Falardeau, R., and Fischer, J. E., 1979, Solid St. Commun., 32, 631. Eklund, P. C., Kambe, N., Dresselhaus, G., and Dresselhaus, M. S., 1978, Phys. Rev. B, 18, 7069. Eklund, P. C., Smith, D. S., and Murthy, V. R. K., 1981, Synth. Metals, 3, 111. Eklund, P. C., Smith, D. S., Murthy, V. R. K., and Leung, S. Y., 1980b, Synth. Metals, 2, 99. Eklund, P. C., and Subbaswamy, K. R., 1979, Phys. Rev. B, 20, 5157. El Makrini, M., Guerard, D., Lagrange, P., and He rold, A., 1980, Physica B, 99, 481. Ellenson, W. D., Semmingsen, D., Guerard, D., Onn, D. G., and Fischer, J. E., 1977, Mater. Sci. Engng, 31, 137. Elliott, R. J., 1954, Phys. Rev., 96, 266. Endo, M., Koyama, T., and Hishiyama, Y., 1976, Jap. J. appl. Phys., 15, 2073. Endo, M., Nakajima, H., Koyama, T., and Inagaki, M., 1979, Extended Abstracts of the 14th Biennial Conference on Carbon, Pennsylvania State University, p. 284. Endo, M., Oberlin, A., and Koyama, T., 1977, Jap. J. appl. Phys., 16, 1519. Ergun, S., 1968, Chemistry and Physics of Carbon, Vol. 3, edited by P. L. Walker (New York: Marcel Dekker), p. 45. Ergun, S., Yashinsky, J. B., and Townsend, J. R., 1967, Carbon, 5, 403. Esaki, L., and Chang, L. L., 1976, Thin Solid Films, 36, 285. Estrade, H., Conard, J., Lauginie, P., Heitjons, P., Fujara, F., Buttler, W., Kiese, rard, D., 1980, Physica B, 99, 531. G., Ackermann, H., and Gue Evans, E. L., and Thomas, J. M., 1975, J. Solid St. Chem., 14, 99. guin, F., and Setton, R., 1980, Facchini, L., Quinton, M. F., Legrand, A. P., Be Physica B, 99, 525. Falardeau, E. R., Foley, G. M. T., Zeller, C., and Vogel, F. L., 1977, J. chem. Soc. chem. Commun., p. 1315. Falardeau, E. R., Hanlon, L. R., and Thompson, T. E., 1978, Inorg. Chem., 17, 301. Feher, G., and Kip, A. F., 1955, Phys. Rev., 97, 337. Fischer, J. E., 1977, Mater. Sci. Engng, 31, 211; 1979, Physics and Chemistry of Materials with Layered Structures, Vol. 6, edited by F. Levy (Dordrecht: Reidel), p. 481. Fischer, J. E., Thompson, T. E., Foley, G. M. T., Gue rard, D., Hoke, M., and Lederman, F. L., 1976, Phys. Rev. Lett., 37, 769. Fischer, J. E., Thompson, T. E., and Vogel, F. L., 1975, Petroleum Derived Carbons, ACS Symposium Series, No. 12, p. 418. Fischer, R. M., Smith, D. J., Freeman, L. A., Pennycook, S. J., and Howie, A., 1979, Extended Abstracts of the 14th Biennial Conference on Carbon, Pennsylvania State University, p. 318. Flandrois, S., Masson, J., Rouillou, J., Gualtier, J., and Hauw, C., 1981, Synth. Metals (to be published). Foley, G. M. T., Zeller, C., Falardeau, E. R., and Vogel, F. L., 1977, Solid St. Commun., 24, 371. Forsman, W. C., Vogel, F. L., Carl, D. E., and Hoffman, J., 1978, Carbon, 16, 269. Fredenhagen, K., and Cadenbach, G., 1926, Z. anorg. Chem., 158, 249. Fukuyama, H., 1971, Prog. theor. Phys., 45, 704. Furdin, G., Bach, B., and He rold, A., 1970, C. r. hebd. Seanc. Acad. Sci., Paris, C, 271, 683.

180

M. S. Dresselhaus and G. Dresselhaus

Furdin, G., and Herold, A., 1972, Bull. Soc. chim. Fr., 5, 1768. Fuzellier, H., Melin, J., and He rold, A., 1977a, Carbon, 15, 45; 1977b, Mater. Sci. Engng, 31, 91. Galt, J. K., Yager, W. A., and Dail, H. W., 1956, Phys. Rev., 103, 1586. Ginderow, D., and Setton, R., 1963, C. r. hebd. Seanc. Acad. Sci. Paris C, 257, 687; 1970, Ibid., 270, 135. Goldsmith, M., 1950, J. chem. Phys., 18, 523. Gole , J., 1977, Mater. Sci. Engng, 31, 309. Greenaway, D. L., Harbeke, G., Bassani, F., and Tosatti, E., 1969, Phys. Rev., 178, 1340. Gualberto, G. M., Underhill, C., Leung, S. Y., and Dresselhaus, G., 1980, Phys. Rev. B, 21, 862. Gue rard, D., Foley, G. M. T., Zanini, M., and Fischer, J. E., 1977a, Nuovo Cim. B, 38, 410. Gue rard, D., and He rold, A., 1974, C. r. hebd. Seanc. Acad. Sci. Paris, C, 279, 455; 1975, Carbon, 13, 337. Gue rard, D., Lagrange, P., El Makrini, M., and He rold, A., 1978, Carbon, 16, 285. rard, D., Lagrange, P., and He rold, A., 1977b, Mater. Sci. Engng, 31, 29. Gue rard, D., Zeller, C., and He rold, A., 1976, C. r. hebd. Seanc. Acad. Sci., Paris C, Gue 283, 437. Guizetti, G., Nosenzo, L., Reguzzoni, E., and Samoggia, G., 1973, Phys. Rev. Lett., 31, 154. Haering, R. R., and Wallace, P. R., 1957, J. Phys. Chem. Solids, 3, 253. Halpin, M. K., and Jenkins, G. M., 1970, Proceedings of the Third Conference on Industrial Carbons and Graphite (London: Society for Chemistry in Industry), p. 53. Hanlon, L. R., Falardeau, E. R., and Fischer, J. E., 1977a, Solid St. Commun., 24, 377; 1977b, Mater. Sci. Engng, 31, 161. Hannay, N. B., Geballe, T. H., Matthias, B. T., Andres, K., Schmidt, P., and MacNair, D., 1965, Phys. Rev. Lett., 14, 225. Hastings, J. B., Ellenson, W. D., and Fischer, J. E., 1979, Phys. Rev. Lett., 42, 1552. Heald, S. M., and Stern, E. A., 1978, Phys. Rev. B, 17, 361; 1980, Synth. Metals, 1, 249. Heerschap, M. M., and Delavignette, P., 1967, Carbon, 5, 383. Heerschap, M., Delavignette, P., and Amelinckx, S., 1964, Carbon, 1, 235. Heflinger, B. L., 1979, Ph.D. Thesis, MIT, Cambridge, Massachusetts. Hennig, G. R., 1952a, J. chem. Phys., 20, 1438; 1952b, Ibid., 20, 1443; 1959, Progress in Inorganic Chemistry, Vol. 1, edited by F. A. Cotton (New York: Interscience), p. 125; 1965, J. chem. Phys., 43, 1201. Hennig, G. R., and McClelland, J. D., 1955, J. chem. Phys., 23, 1431. Hennig, G. R., and Meyer, L., 1952, Phys. Rev., 87, 439. Hennig, G. R., Smaller, B., and Yasaitis, E. L., 1954, Phys. Rev., 95, 1088. Herinckx, C., 1973, Carbon, 11, 199. Herinckx, C., Perret, R., and Ruland, W., 1972, Carbon, 10, 711. Herold, A., 1955, Bull. Soc. chim. Fr., p. 999; 1977, Mater. Sci. Engng, 31, 1; 1979, Physics and Chemistry of Materials with Layered Structures, Vol. 6, edited by F. Le vy (Dordrecht: Reidel), p. 323. Herzberg, G., 1945, Spectra of Diatomic Molecules (Princeton: Van Nostrand). Hibbs, J. D., and Young, D. A., 1978, Chem. Phys. Lett., 53, 361. Higuchi, K., Suematsu, H., and Tanuma, S., 1980, J. Phys. Soc. Japan, 48, 1532. Hirsch, B., Howie, A., Nicholson, R. B., Pashley, D. W., and Whelan, M. J., 1965, Electron Microscopy of Thin Crystals (London: Butterworths), Chap. 6. Hoffmann, U., and Frenzel, A., 1931, Z. Elektrochem., 37, 613. Hohlwein, D., Readman, P. W., Chamberod, A., and Coey, J. M. B., 1974, Phys. Stat. Sol. (b), 64, 305. Holzwarth, N. A. W., 1980, Phys. Rev. B, 21, 3665. Holzwarth, N. A. W., Girifalco, L. A., and Rabii, S., 1978a, Phys. Rev. B, 18, 5206. Holzwarth, N. A. W., and Rabii, S., 1977, Mater. Sci. Engng, 31, 195. Holzwarth, N. A. W., Rabii, S., and Girifalco, L. A., 1978b, Phys. Rev. B, 18, 5190.

Intercalation compounds of graphite

181

Hooley, J. G., 1972, Carbon, 10, 155; 1973, Ibid., 11, 225; 1975, Ibid., 13, 469; 1977a, Preparation and Crystal Growth of Materials with Layered Structures, edited by R. M. A. Leith (Dordrecht: Reidel), p. 1; 1977b, Mater. Sci. Engng, 31, 17; 1978, Bull. Am. phys. Soc, 23, 185. Hooley, J. G., and Bartlett, M., 1967, Carbon, 5, 417. Hooley, J. G., Bartlett, M. W., Liengme, B. V., and Sams, J. R., 1968, Carbon, 6, 681. Hooley, J. G., and Deitz, V. R., 1978, Carbon, 16, 251. Hooley, J. G., Garby, W. P., and Valentin, J., 1965, Carbon, 3, 7. Hooley, J. G., and Soniassy, R. N., 1970, Carbon, 8, 191. Horie, C., Maeda, M., and Kuramoto. Y., 1980, Physica B, 99, 430. Horn, D., and Boehm, H. P., 1977, Mater. Sci. Engng, 31, 87. Hove, J. E., 1955, Phys. Rev., 100, 645. Hwang, D. M., Solin, S. A., Utlaut, M., and Isaacson, M., 1981, Synth. Metals (to be published). Hwang, D. M., Utlaut, M., Isaacson, M. S., and Solin, S. A., 1979, Phys. Rev. Lett., 43, 882; 1980, Physica B, 99, 435. Ichikawa, M., Kawase, K., and Tamaru, K., 1972a, J. chem. Soc. chem. Commun., p. 177. Ichikawa, M., Kondo, T., Kawase, K., Sudo, M., Onishi, T., and Tamaru, K., 1972b, J. chem. Soc. chem. Commun., p. 176. g, Inagaki, M., Rouillon, J. C., Fu G., and Delhaes, P., 1977, Carbon, 15, 181. Inoshita, T., Nakao, K., and Kamimura, H., 1977, J. phys. Soc. Japan, 43, 1237. Inoue, M., 1962, J. phys. soc. Japan, 17, 808. Isaacson, M., Utlaut, M., and Ohtsuki, M., 1979, Extended Abstracts of the 14th Biennial Conference on Carbon, Pennsylvania State University, p. 105. Iye, Y., Takahashi, O., and Tanuma, S., 1980, Solid St. Commun., 33, 1071. Johnson, L. G., and Dresselhaus, G., 1973, Phys. Rev, B, 7, 2275. Juza, R., Luebbe, H., and Heinlein, L., 1979, Z. anorg. allg. Chem., 258, 105. Juza, R., Schmidt, P. S., Schmeckenbecker, A., and Jonk, P., 1955, Naturwissenshaften, 42, 124. Juza, R., and Wehle, V., 1965, Naturwissenschaften, 52, 560. Kahn, A. H., and Frederikse, H. P. R., 1959, Solid State Physics, Vol. 9, edited by F. Seitz and D. Turnbull (New York: Academic Press), p. 257 Kambe, N., 1981, Ph.D. Thesis, M.I.T., Cambridge, Massachusetts. Kambe, N., Dresselhaus, G., and Dresselhaus, M. S., 1980, Phys. Rev. B, 21, 3491. Kambe, N., Dresselhaus M. S., Dresselhaus, G., Basu, S., McGhie, A. R., and Fischer, J. E., 1979, Mater. Sci. Engng, 40, 1. Kamimura, H., Nakao, K., Ohno, T., and Inoshita, T., 1980, Physica B, 99, 401. Karimov, Yu. S., 1973, Zh. eksp. teor. Fiz., 65, 261 (1974, Soviet Phys. JETP, 38, 129); 1974, Ibid., 66, 1121 (1975, Ibid., 39, 547); 1975, Ibid., 68, 1539 (1976, Ibid., 41, 772). Karimov, Yu. S., Zvarykina, A. V., and Novikov, Yu. N., 1971, Fizika tverd. Tela, 13, 2836 (1972, Soviet Phys. Solid St., 13, 2388) Kawamura, K., Ouchi, Y., Oshima, H., and Tsuzuku, T., 1979, J. phys. Soc. Japan, 46, 587. Kawamura, K., Saito, T., and Tsuzuku, T., 1975, Carbon, 13, 452. Khanna, S. K., Falardeau, E. R., Heeger, A. J., and Fischer, J. E., 1978, Solid St. Commun., 25, 1059. Kittel, C., 1956, Introduction to Solid State Physics, second edition (New York: Wiley), pp. 240, 352; 1978, Solid St. Commun., 25, 519. Klotz, H., and Schneider, A., 1982, Naturwissenschaften, 49, 448. Kobayashi, M., and Tsujikawa, I., 1979, J. phys. Soc. Japan, 46, 1945. Koike, Y., Suematsu, H., Higuchi, K., and Tanuma, S., 1978, Solid St. Commun., 27, 623; 1980, Physica B, 99, 503. Koike, Y., Tanuma, S., Suematsu, H., and Higuchi, I., 1981, J. Phys. Chem. Solids (to be published). Kondow, T., Mizutani, U., and Massalski, T. B., 1977, Mater. Sci. Engng, 31, 267. Krapchev, T., Ogilvie, R., and Dresselhaus, M. S., 1981, Extended Abstracts of the 15th Biennial Conference on Carbon, University of Pennsylvania.

182

M. S. Dresselhaus and G. Dresselhaus

Lagrange, P., El Makrini, M., Gue rard, D., and He rold, A., 1980, Physica B, 99, 473. Lagrange, P., and He rold, A., 1978, Carbon, 16, 235. Lander, J. J., and Morrison, J. M., 1967, Surf. Sci., 6, 235. Lau, C. L., and Dresselhaus, M. S., 1980, Phys. Rev. B, 21, 3635. Lauginie, P., 1981 (private communication). rard, D., El Makrini, M., Lagrange, P., Lauginie, P., Conard, J., Estrade, H., Gue Fuzellier, H., Furdis, G., and Vasse, R., 1979, Extended Abstracts of the 14th Biennial Conference on Carbon, Pennsylvania State University. Lauginie, P., Estrade, H., Conard, J., Guerard, D., Lagrange, P., and El Makrini, M., 1980, Physica B, 99, 514. Lazo, R. M., and Hooley, J. G., 1956, Can. J. Chem., 34, 1574. Leung, S. Y., 1980c, Ph.D. Thesis, MIT, Cambridge, Massachusetts. Leung, S. Y., and Dresselhaus, G., 1981a, Phys. Rev. (to be published). Leung, S. Y., Underhil, C., Dresselhaus, G., and Dresselhaus, M. S., 1980a, Solid St. Commun., 33, 285. Leung, S. Y., Underhill, C., Dresselhaus, G., Krapchev, T., Ogilvie, R., and Dresselhaus, M. S., 1979, Solid St. Commun., 32, 635; 1980b, Physics Lett. A, 76, 89. Leung, S. Y., Underhill, C., Krapchev, T., Dresselhaus, G., Dresselhaus, M. S., and Wuensch, B. J., 1981a (to be published). Maeda, M., Kuramoto, Y., and Horie, C., 1979, J. phys. Soc. Japan, 47, 337. Maire, J., and Mering, J., 1959, Proceedings of the Third Carbon Conference (London, New York: Pergamon Press), p. 337. Marchand, A., Rouillon, J. C., and Courtois DArcollieres, F., 1973, Carbon, 11, 113. Markiewicz, R. S., Hart, H. R., Jr., Interrante, L. V., and Kasper, J. S., 1980a, Synth. Metals, 2, 331. Markiewicz, R. S., Kasper, J. S., and Interrante, L. V., 1980b, Synth Metals, 2, 363. Mashinskii, V. I., Postnikov, V. A., Novikov, Yu. N., Lapidus, A. L., Volpin, M. E., and Eidus, Ya. T., 1976, Izv. Akad. Nauk. SSSR. Ser. Khim., 9, 2018. McClure, J. W., 1956, Phys. Rev., 104, 666; 1957, Ibid., 108, 612; 1960, Ibid., 119, 606; 1964, IBM Jl Res. Dev., 8, 225; 1969, Carbon, 7, 425; 1971, Proceedings of the International Conference on Semimetals and Narrow Gap Semiconductors, edited by D. L. Carter and R. T. Bate (New York: Pergamon Press), p. 127. McClure, J. W., and Spry, W. J., 1968, Phys. Rev., 165, 809. McClure, J.. W., and Yafet, Y., 1962, Proceedings of the Fifth Conference on Carbon, Vol. 1 (New York: Pergamon Press), p. 22. McDonnell, F. R. M., Pink, R. C., and Ubbelohde, A. R., 1951, J. chem. Soc., p. 191. McFeeley, F. R., Kowalczyk, S. P., Ley, L., Cavell, R. G., Pollak, R. A., and Shirley, D. A., 1974, Phys. Rev. B, 9, 5268. McGovern, I. T., Eberhardt, W., Plummer, E. W., and Fischer, J. E., 1980, Physica B, 99, 415. McMillan, W. L., 1968, Phys, Rev., 167, 331. ^ , rold, A., 1980a, Physica B, 99, 489. McRae, E., Billaud, D., Mareche J. F., and He McRae, E., and Herold, A., 1977, Mater. Sci. Engng, 31, 249. rold, A., 1980b, Physica B, 99, 541. McRae, E., Metrot, A., Willman, P., and He Mele, E. J., and Ritsko, J. J., 1979, Phys. Rev. Lett., 43, 68. Melin, J., and Herold, A., 1975, Carbon, 13, 357. Mendez, E., Chieu, T. C., Kambe, N., and Dresselhaus, M. S., 1980a, Solid St. Commun., 33, 837. Mendez, E., Misu, A., and Dresselhaus, M. S., 1980b, Phys. Rev. B, 21, 827. Merle, G., Rashkov, I., Mai, C., and Gole , J., 1977, Mater. Sci. Engng, 31, 39. Metz, W., and Hohlwein, D., 1975a, Carbon, 13, 84; 1975b, Ibid., 13, 87. Metz, W., and Siermoglu L., 1978, Carbon, 16, 225. ss, Misu, A., Mendez, E., and Dresselhaus, M. S., 1979, J. phys. Soc. Japan, 47, 199. Mizutani, U., Kondow, T., and Massalski, T. B., 1978, Phys. Rev. B, 17, 3165. Moore, A. W., 1973, Chemistry and Physics of Carbon, Vol. 11, edited by P. L. Walker and P. A. Thrower (New York: Dekker), p. 69.

Intercalation compounds of graphite

183

Morris, R. C., and Coleman, R. V., 1973, Phys. Rev. B, 7, 991. Morris, R. C., Coleman, R. V., and Bhandari, B., 1972, Phys. Rev. B, 5, 895. Mukaibo, T., and Takahashi, Y., 1963, Bull. chem. Soc. Japan, 36, 625. Mu ller, K. A., and Kleiner, R., 1962, Physics Lett., 1, 98. Murray, J. J., and Ubbelohde, A. R., 1969, Proc. R. Soc. A, 312, 371. Murthy, V. R. K., Smith, D. S., and Eklund, P. C., 1980, Mater. Sci. Engng, 45, 77. Nagayoshi, H., 1977, J. phys. Soc. Japan, 43, 760. Nagayoshi, H., Tsukada, M., Nakao, K., and Uemura, Y., 1973, J. Phys. Soc. Japan, 35, 396. Nagayoshi, H., Nakao, K., and Yemura, Y., 1976, J. phys. Soc. Japan, 41, 1480. Naito, S., Ogawa, O., Ichikawa, M., and Tamaru, K., 1972, J. chem. Soc. chem. Commun., p. 1266. Nakao, K., 1976, J. phys. Soc. Japan, 40, 761; 1979, Ibid., 47, 208. Nemanich, R. J., Lucovsky, G., and Solin, S. A., 1975, Proceedings of the International Conference on Lattide Dynamics, edited by M. Balkanski (Paris: Flammarion Press), p. 619; 1977a, Mater. Sci. Engng, 31, 157; 1977b, Solid St. Commun., 23, 117. Nemanich, R. J., and Solin, S. A., 1977, Solid St. Commun., 23, 417; 1979, Phys. Rev. B, 20, 392. Nemanich, R. J., Solin, S. A., and Guerard, D., 1977c, Phys. Rev. B, 16, 2965. Nicklow, R., Wakabayashi, N., and Smith, H. G., 1972, Phys. Rev. B, 5, 4951. Nicolson, A. P. P., and Bacon, D. J., 1977, J. Phys. C, 10, 2295. Niess, R., and Stumpp, E., 1978, Carbon, 16, 265. Nixon, D. E., 1966, Ph.D. Thesis, Imperial College, London. Nixon, D. E., and Parry, G. S., 1968, J. Phys. D, 1, 291; 1969, J. Phys. C, 2, 1732. Nixon, D. E., Parry, G. S., and Ubbelohde, A. R., 1966, Proc. R. Soc. A, 291, 324. Novikov, Yu. N., Kazakov, M. E., Zvarykina, A. V., Astakhova, I. S., and Volpin, M. E., 1971, Zh. struct. Khim., 12, 486 (1972, J. struct. Chem., 12, 446). Novikov, Yu. N., Postnikov, V. A., Salyn, Ya. V., Nefedev, V. A., and Volpin, M. E., 1973, Izv. Akad. Nauk. SSR, Ser. Khim., p. 1968 (Bull. Acad. Sci. SSR, Ser. Chem., p. 1653). Oberlin, A., Endo, M., and Koyama, T., 1976, Carbon, 14, 133. ntherodt, H.-J., 1980, Phys. Rev. Lett., Oelhafen, P., Pfluger, P., Hauser, E., and Gu 44, 197 (1979, Solid St. Commun., 32, 885). Ohhashi, J., and Tsujikawa, I., 1974a, J. phys. Soc. Japan, 36, 422; 1974b, Ibid., 37, 63. Ohno, T., Nakao, K., and Kamimura, H., 1979, J. phys. Soc. Japan, 47, 1125. Onn, D. G., Foley, G. M. T., and Fischer, J. E., 1977, Mater. Sci. Engng, 31, 271; 1979, Phys. Rev. B, 19, 6474. Ono, S., 1976, J. phys. Soc. Japan, 40, 498. Postlund, S., and Berker, A. N., 1980, Phys. Rev. B, 21, 5410. Page-Lecuyer, A., Luche, J. L., Kagan, H. B., Calin, G., and Mazieres, C., 1973, Bull. Soc. chim. Fr., p. 1690. Painter, G. S., and Ellis, D. E., 1970, Phys. Rev. B, 1, 4747. Panayotov, I. M., Berlinova, I. V., and Rashkov, I. B., 1975, J. Polym. Sci., 13, 2043. Panayotov, I. M., and Rashkov, I. B., 1972, J. Polym. Sci., 10, 1267. Parkash, S., Chakprabartty, S. L., and Hooley, J. G., 1978, Carbon, 16, 231. Parry, G. S., 1971, Third Conference on Industrial Carbon and Graphite (London: Society of Industrial Chemistry), p. 58; 1977, Mater. Sci. Engng, 31, 99. Parry, G. S., and Nixon, D. E., 1967, Nature, London, 216, 909. Parry, G. S., Nixon, D. E., Lester, K. M., and Levene, B. C., 1969, J. Phys. C, 2, 2156. Perrachon, J., 1978, Ph.D. Thesis, University of Pennsylvania. Perrachon, J. B., Zeller, C., and Vogel, F. L., 1979, Extended Abstracts of the 14th Biennial Conference on Carbon, Pennsylvania State University, p. 304. Pfluger, P., Ku nzi, H. U., and Gu ntherodt, H.-J., 1979, Appl. Phys. Lett., 35, 771. Pfluger, P., Oelhafen, P., Ku nzi, H. U., Jeker, R., Hauser, E. Ackerman, K. P., ller, M., and G untherodt, H.-J., 1980, Physica B, 99, 395. Mu Philipp, H. R., 1977, Phys. Rev. B, 16, 2896.

184

M. S. Dresselhaus and G. Dresselhaus

ssler, S., 1979, Solid. St. Commun., 32, 1337. Pietronero, L., and Stra ssler, S., and Zeller, H. R., 1979, Solid St. Commun., 30, 399. Pietronero, L., Stra ssler, S., Zeller, H. R., and Rice, M. J., 1978, Phys. Rev. Lett., 41, Pietronero, L., Stra 763; 1980a, Physica B, 99, 499; 1980b, Phys. Rev. B, 22, 904. Platts, D. A., 1975, Ph.D. Thesis, MIT, Cambridge, Massachusetts. Platts, D. A., Chung, D. D. L., and Dresselhaus, M. S., 1977, Phys. Rev. B, 15, 1087. Poitrenaud, J., 1970, Rev. Phys. Applic., 5, 275. Poquet, E., Lumbroso, N., Hoarau, J., Marchand, A., Pacault, A., and Soule, D. E., 1960, J. Chim. phys., 57, 866. Postnikov, V. A., Dmitrienko, L. M., Ivanova, R. F., Dobrolyubova, N. L., Golubeva, M. A., Gapeeva, T. I., Novikov, Yu. N., Shur, V. B., and Volpin, M. E., 1975, Izv. Acad. Nauk. SSSR, Ser. Khim., 24, 2529. Potter, M. E., Johnson, W. D., and Fischer, J. E., 1979, Extended Abstracts of the 14th Biennial Conference on Carbon, Pennsylvania State University, p. 300. Proceedings of the International Carbon Fibres Conference, 1971, 1974 (London: The Plastics Institute). Resing, H. A., Vogel, F. L., and Wu, T. C., 1979, Mater. Sci. Engng, 41, 113. Ritsko, J. J., and Mele, E. J., 1980a, Phys. Rev. B, 21, 730; 1980b, Physica B, 99, 425. Ritsko, J. J., and Rice, M. J., 1979, Phys. Rev. Lett., 42, 666. Robinson, A. K. V., 1974, M. S. Thesis, University of Maryland. Robson, D., Assabghy, F. Y. T., Cooper, E. G., and Ingram, D. J. E., 1973, J. Phys. D, 6, 1822. Robson, D., Assabghy, F. Y. T. , and Ingram, D. J. E., 1972, J. Phys. D, 5, 169. Rosenman, I., Batallan, F., and Furdin, G., 1979, Phys. Rev. B, 20, 2373. Ru dorff, W., 1939, Z. phys. Chem. B, 45, 42; 1941, Z. anorg. allg. Chem., 245, 383; 1959, Adv. inorg. Chem. Radiochem., 1, 223. Ru dorff, W., and Hoffmann, U., 1938, Z. anorg. allg. Chem., 238, 1. dorff, W., and Schulz, H., 1940, Z. anorg. allg. Chem., 245, 121. Ru dorff, W., and Schulze, E., 1954, Z. anorg. allg. Chem., 277, 156. Ru dorff, W., Schulze, E., and Rubisch, O., 1955, Z. anorg. allg. Chem., 282, 232. Ru Ru dorff, W., and Siecke, W. F., 1958, Chem. Ber., 91, 1348. Ru dorff, W., and Zeller, R., 1955, Z. Anorg. allg. Chem., 279, 182. Ruprecht, G., Ginsberg, D. M., and Leslie, J. D., 1962, J. opt. Soc. Am., 52, 665. Saehr, D., 1964, Bull. Soc. chim. Fr., p. 1287. Safran, S. A., and Di Salvo, F. J., 1979, Phys. Rev., 20, 4889. Safran, S. A., Di Salvo, F. J., Haddon, R. C., Waszczak, J. V., and Fischer, J. E., 1980, Physica B, 99, 494. Safran, S. A., and Hamann, D. R., 1979, Phys. Rev. Lett., 42, 1410; 1980, Physica B, 99, 469; 1981, Phys. Rev. B, 23, 565. Saito, N. H., and Tsuzuku, T., 1973, Carbon, 11, 469. Salzano, F. J., and Aronso, S., 1966a, J. chem. Phys., 44, 4320; 1966b, Ibid., 45, 2221; 1966c, Ibid., 45, 4551. Samuelson, L., Batra, I. P., and Roetti, C., 1980, Solid St. Commun., 33, 817. Sasa, T., Takahashi, Y., and Mukaibo, T., 1970, Bull chem. Soc. Japan, 43, 34; 1971, Carbon, 9, 407; 1972, Bull. chem. Soc. Japan, 45, 2250. Saunders, G. A., Ubbelohde, A. R., and Young, D. A., 1963, Proc. R. Soc. A, 271, 499. utl, P., 1841, J. prakt Chem., 21, 155. Schaffa Schleede, A., and Wellmann, M., 1932, Z. phys. Chem., 18, 1. gl, R., and Boehm, H. P., 1979, Extended Abstracts of the 14th Biennial Carbon Schlo Conference, Pennsylvania State University, p. 308. Schoppen, G., Meyer-Spasche, H. Siemsglu ss, L., and Metz, W., 1977, Mater. Sci. Engng, 31, 115. Schroeder, P. R., Dresselhaus, M. S., and Javan, A., 1968, Phys. Rev. Lett., 20, 1292; 1971, Proceedings of the International Conference on Semimetals and Narrow Gap Semiconductors, edited by D. L. Carter and R. T. Bate (New York: Pergamon Press), p. 139. Scott, J. R., 1974, Rev. mod. Phys., 46, 83.

Intercalation compounds of graphite

185

Setton, R., 1964, Les Carbones, edited by A. Pacault (Paris: Masson), p. 643; 1977, Mater. Sci. Engng, 31, 303. Sharma, M. P., Johnson, L. G., and McClure, J. W., 1974, Phys. Rev. B, 9, 2467. Shayegan, M., 1979, B.S. Thesis, MIT, Cambridge, Massachusetts; 1981, M.S. Thesis, MIT, Cambridge, Massachusetts. Shieh, C. C., Schmidt, R. L., and Fischer, J. E., 1979, Phys. Rev. B, 20, 3351. Shoenberg, D., 1952, Phil. Trans. R. Soc., 245, 1. Singhal, S. C., 1980, Physica B, 99, 536. Slater, J. C., and Koster, G. F., 1954, Phys. Rev., 94, 1498. Slonczewski, J. C., and Weiss, P. R., 1958, Phys. Rev., 109, 272. Smaller, B., Hennig, G. R., and Yasaitis, E. L., 1954, Phys. Rev., 97, 239. Solin, S. A., 1977, Mater. Sci. Engng, 31, 153; 1980, Physica B, 99, 443. Song, J. J., Chung, D. D. L., Eklund, P. C., and Dresselhaus, M. S., 1976, Solid St. Commun., 20, 1111. Soule, D. E., 1958, Phys. Rev., 112, 698; 1964, IBM Jl Res. Dev., 8, 268. Soule, D. E., McClure, J. W., and Smith, L. B., 1964, Phys. Rev. A, 134, 453. Spain, I. L., 1971, Proceedings of the International Conference on Semimetals and Narrow Gap Semiconductors, edited by D. L. Carter and R. T. Bate (New York: Pergamon Press), p. 177; 1973, Chemistry and Physics of Carbon, Vol. 8, edited by P. L. Walker (New York: Marcel Dekker), pp. 105, 110. Spain, I. L., and Nagel, D. J., 1977, Mater. Sci. Engng, 31, 183. Stein, C., Bonnetain L., and Gole , J., 1966, Bull. Soc. chim. Fr., p. 3166. Stephens, P. W., Hwiney, P., Birgeneau, R. J., and Horn, P. M., 1979, Phys. Rev. Lett., 43, 47. gl, R., and Pentenreider, R., 1979, Carbon, 17, Streifinger, L., Boehm, H. P., Schlo 195. Stumpp, E., 1977, Mater. Sci. Engng, 31, 53. Suematsu, H., Higuchi, K., and Tanuma, S., 1980a, J. phys. Soc. Japan, 48, 1541. Suematsu, H., and Tanuma, S., 1972, J. phys. Soc. Japan, 33, 1619. Suematsu, H., Tanuma, S., and Higuchi, K., 1980b, Physica B, 99, 420. Sugihara, K., Kawamura, K., and Tsuzuku, T., 1979, J. phys. Soc. Japan, 47, 1210. Suzuki, M., Ikeda, H., Suematsu, H., Endoh, Y., Shiba, H., and Hutchings, M. T., 1980, J. phys. Soc. Japan, 49, 671. Syme-Johnson, A. W., 1967, Acta crystallogr., 23, 770. Taft, E. A., and Philipp, H. R., 1965, Phys. Rev. A, 138, 197. Tanuma, S., Suematsu, H., Higuchi, K., Inada, R., and Onuki, Y., 1978, Proceedings of the Conference on the Application on High Magnetic Fields in Semiconductor Physics, edited by J. F. Ryan (Oxford: Clarendon Press), p. 85. Tashiro, K., Saito, M., and Tsuzuku, T., 1978, Carbon, 16, 292. Thomas, J. M., Millward, G. R., Davies, N. C., and Evans, E. L., 1976, J. chem. Soc. Dalton Trans., p. 2443. gl, R. F., and Boehm, H. P., 1980, Mat. Res. Thomas, J. M., Millword, G. R., Schlo Bull., 15, 671. Thompson, T. E., Falardeau, E. R., and Hanlon, L. R., 1977, Carbon, 15, 39. Toy, W. W., Dresselhaus, M. S., and Dresselhaus, G., 1977, Phys. Rev. B, 15, 4077. Toyota, N., Nakatsuji, H., Noto, K., Hoshi, A., Kobayashi, N., Muto, Y., and Onodera, Y., 1976, J. low Temp. Phys., 25, 485. Tricker, M. J., Evans, E. L., Cadman, P., Davies, N. C., and Bach, B., 1974, Carbon, 12, 499. Tsang, D. Z., and Dresselhaus, M. S., 1976, Carbon, 14, 43. Tsuzuku, T., 1979, Carbon, 17, 293. Tsuzuku, T., and Sugihara, K., 1975, Chemistry and Physics of Carbon, Vol. 12, edited by P. L. Walker and P. A. Thrower, p. 109. Tuinstra, F., and Koenig, J. F., 1970, J. chem. Phys., 53, 1126. Ubbelohde, A. R., 1968a, Proc. R. Soc. A, 304, 25; 1968b, Carbon, 6, 177; 1969, Proc. R. Soc. A, 309, 297; 1971, Ibid., 321, 445; 1972, Ibid., 327, 289; 1976, Carbon, 14, 1. Ubbelohde, A. R., and Lewis, F. A., 1960, Graphite and its Crystal Compounds (Oxford: Clarendon Press).

186

M. S. Dresselhaus and G. Dresselhaus

Underhill, C., Krapchev, T., and Dresselhaus, M., 1980, Synth. Metals, 2, 47. Underhill, C., Leung, S. Y., Dresselhaus, G., and Dresselhaus, M. S., 1979, Solid St. Commun., 29, 769. Ushio, H., Uda, T., and Uemura, Y., 1972, J. phys. Soc. Japan, 33, 1551. Van der Hoeven, B. J. C., and Keesom, P. H., 1963, Phys. Rev., 130, 1318. Van Haeringen, W., and Junginger, H.-G., 1969, Solid St. Commun., 7, 1723. Vangelisti, R., and Herold, A., 1976, Carbon, 14, 333. Venghaus, H., 1975, Phys. Stat:Sol. (b), 71, 609; 1977, Ibid., 81, 221. Vogel, F. L., 1976, Carbon, 14, 175; 1977, J. Mater. Sci., 12, 982; 1979, Molecular Metals, edited by W. E. Hateld (New York: Plenum Press), p. 261. Vogel, F. L., Foley, G. M. T., Zeller, C., Falardeau, E. R., and Gan, J., 1977, Mater. Sci. Engng, 31, 261. Vogel, F. L., Fuzellier, H., Zeller, C., and McRae, E. J., 1979, Carbon, 17, 255. Volpin, M. E., Novikov, Yu, N., Lappina, N. D., Kasatochkin, V. A., Struchkov, Yu. T., Kazakov, M. E., Stukham, R. A., Povitsky, V. A., Karimov, Yu. S., and Zvarikina, A. V., 1975, J. Am. chem. Soc., 97, 3366. Wagoner, G., 1960, Phys. Rev., 118, 647. Wallace, P. R., 1947, Phys. Rev., 71, 622. Weinberger, B. R., Kaufer, J. Heeger, A. J., Falardeau, E. R., and Fischer, J. E., 1978a, Solid St. Commun., 27, 163. Weinberger, B. R., Kaufer, J., Heeger, A. J., Fischer, J. E., Moran, M., and Holzwarth, N. A. W., 1978b, Phys. Rev. Lett., 41, 1417. Whittingham, M. S., 1975, J. electrochem. Soc., 121, 1308. Whittingham, M. S., and Ebert, L. B., 1979, Physics and Chemistry of Materials with Layered Structures, Vol. 6, edited by F. Levy (Dordrecht: Reidel), p. 533. Williamson, S., Foner, S., and Dresselhaus, M. S., 1965, Phys. Rev. A, 140, 1429. Williamson, S., Surma, M., Praddaude, H. C., Patten, R. A., and Furdyna, J. K., 1966, Solid St. Commun., 4, 37. Willis, R. F., Fitton, B., and Painter, G. S., 1974, Phys. Rev. B, 9, 1926. Woollam, J. A., 1970, Physics Lett. A, 32, 115. Woollam, J. A., Haugland, E., Dowell, M. B., Kambe, N., Mendez, E., Hakimi, F., Dresselhaus, G., and Dresselhaus, M. S., 1979, Extended Abstracts of the 14th Biennial Conference on Carbon, Pennsylvania State University, p. 320. Woollam, J. A., and Somoano, R. B., 1976, Phys. Rev. B, 13, 3843. Wyckoff, R. W. G., 1964, Crustal Structures, Vol. 1 (New York: Interscience). Young, D. A., 1977, Carbon, 15, 373. Young, J. A., and Koppel, N. U., 1965, J. chem. Phys., 42, 357. Zabel, H., Jan, Y. M., and Moss, S. C., 1980, Physica B, 99, 43. Zabel, H., Moss, S. C., Caswell, N., and Solin, S. A., 1979, Phys. Rev. Lett., 43, 2022. Zanini, M., Basu, S., and Fischer, J. E., 1978a, Carbon, 16, 211. Zanini, M., Ching, L.-Y., and Fischer, J. E., 1978b, Phys. Rev. B, 18, 2020. Zanini, M., and Fischer, J. E., 1977, Mater. Sci. Engng, 31, 169. Zanini, M., Grubisic, D., and Fischer, J. E., 1978c, Phys. Stat. Sol., 90, 151. Zeller, C., Denenstein, A., and Foley, G. M. T., 1979a, Rev. scient. Instrum., 50, 602. Zeller, C., Foley, G. M. T., Falardeau, E. R., and Vogel, F. L., 1977, Mater. Sci. Engng, 31, 255. Zeller, C., Foley, G. M. T., and Vogel, F. L., 1978, J. Mater. Sci., 13, 1114. Zeller, C., Pendrys, L. A., and Vogel, F. L., 1979b, J. Mater. Sci., 14, 2241. Zunger, A., 1978, Phys. Rev. B, 17, 626.

Potrebbero piacerti anche