r'"#8-.'9:-"-’ ., . . . THE’QYE'S’ LIBRARY 3 1293 01104 4173 Michigan State University This is to certify that the dissertation entitled The Effect of Composition on Charge Exchange, Lattice Expansion, and Staging in Potassium—Ammonia Graphite Intercalation Compounds presented by Brian R. York has been accepted towards fulfillment of the requirements for Ph.D . degree in Physics W ‘ Major professor Stuart A. Solin Date March 22, 1985 MSU is an Affirmative Action/Equal Opportunity Institution 0-12771 MSU RETURNING MATERIALS: Place in book drop to LIBRARIES remove this checkout from .—:—. your record. FINES will be charged if book is returned after the date stamped below. 'if,a. .‘MAGICZ 9 JAN 24"}99 -0 THE EFFECT OF COMPOSITION ON CHARGE EXCHANGE, LATTICE EXPANSION, AND STAGING IN POTASSIUM-AMMONIA GRAPHITE INTERCALATION COMPOUNDS 31! Brian 8. York A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Physics and Astronomy 198M wwwx / (.2 ABSTRACT THE EFFECT OF COMPOSITION ON CHARGE EXCHANGE, LATTICE EXPANSION, AND STAGING IN POTASSIUM-AMMONIA GRAPHITE INTERCALATION COMPOUNDS 33' Brian R. York We have studied the charge exchange and compositional dependence of the sandwich thickness of stage-1 alkali-ammonia ternary graphite intercalation compounds K(NH3)ny, O _<_ x _<_ 14.33, 12 _<_ y 5 2M. A model of the sandwich energy is presented which explicitly accounts for x- dependent charge exchange and size or stiffness effects and is in excellent agreement with experimental measurements of the dependence of the (002.) x-ray diffraction patterns on ammonia vapor pressure. From this model we find that for the stage-1 compound K(NH ) f - 0.95 3 n.3302n' and that the NH3 molecules solvate some of the electron charge which was originally donated to the carbon layers in the KC starting material. 2“ In addition, the NH molecules form planar l8-fold coordinated K(NH3)u 3 clusters and hence also solvate the K+ ions in graphite galleries. We suggest that the K(NH3)u clusters together with "spacer" N113 molecules constitute the two-dimensional analogue of the well-studied bulk, three- dimensional metal-ammonia solutions. DEDICATION This work is dedicated to my loving and understanding wife, Linda, for her support and friendship, to my adorable son, Bryan, for giving me such wonderful memories of parenthood, and to my father, mother, and family for their continued support for so many years. ACKNOWLEDGEMENTS It is a pleasure to acknowledge Professor Stuart A. Solin for his helpful criticism and invaluable support at all stages of this work and for introducing this author to James Joyce. I have also benefited from: numerous fruitful discussions with Professors J.L. Dye, S.D. Mahanti, T.J. Pinnavaia, and Drs. S.K. Hark and P. Vora. I gratefully thank IProfessors R.D. Spence, J.A. Cowen, and G.L. Pollack for their help and useful advice which was sought at many stages of this study. Thanks areralso due to Y.B. Fan and x.w. Qian for their assistance and for useful discussions. There are many others, too numerous to mention, to whom I owe thanks, but in particular, M. Langer who has provided the technical glassblowing support that was essential to the success of this project and for remaining remarkably tolerant of demands for items to be finished "yesterday." I am appreciative of the support of this work by the NSF under grant #DMR 82-1155”. Finally, I sincerely thank my typists, L. York and S. Lindsay, for their good nature and uncompromising quality of work, with special thanks going to L. York, who also devoted a large portion of her time and artistic talents to the many figures contained in the body of this thesis. ii FORWARD Our interest has been, and continues to be, the study of the structural phase transitions and related phenomena of layered solids. The term layered solids, as used here, refers to materials for which the interatomic forces within layers of atoms are stronger than the forces between layers. These materials inherently exhibit a high degree of anisotropy in their electronic and/or chemical and physical properties. The theoretical and experimental interest in layered solids stemsfrom the fact that they provide a means of exploring quasi-two-dimensional phenomena. This general class of layered solids can be broken down, based on the atomic thickness of the layer making up the solid, into the following three subclasses. Class I layered materials are formed by layers one atom thick; a typical example of this class would be graphite. Class II layered substances are formed by layers which are a few atoms thick; a common example of this class would be the dichalcogenide TaS Finally, class III layered solids are formed by 2. layers many atoms thick. A representative example of this class would be the sheet silicate clays. While my recent research activities have included investigations of class III materials, the bulk of my activity has been related to the class I materials, in particular, graphite intercalation compounds (GIC's) which are the focus of this dissertation. I present here a collection of studies on the structural characterization and phase iii transition phenomena of several recently-developed, novel alkali-alkali and alkali-ammonia GIC's. The studies of the potassium-ammonia GIC's represents the main body of the dissertation with other investigations described in the appendices. iv TABLE OF CONTENTS List of Figures I. II. III. IV. Introduction Experimental Details and Results Discussion and Analysis A. Comparison with Rudorff's Results on K-NH GIC's 3 B. The Absorption Isotherm C. The Residue Compound K500A) distances. There is a charge transfer, in the intercalation process, between the intercalant layer and the graphite layers. Thus, in addition to characterization according to stage number, the intercalants in GIC's are also classified as donors or acceptors according to whether electron charge is transferred to or fvwnn the graphite sheets, respectively. Detailed theoretical models outlining the physics of the staging 1’2'3 An important phenomena have only recently been presented. ingredient of those models is the contribution of the carbon- intercalant-carbon sandwich energy to the free energy, which establishes the equilibrium stage. The sandwich thickness ds (which is the perpendiclar distance between carbon layers that flank an intercalant layer) can be determined by minimizing the free energy with respect to d . s Safran and Hamann considered both elastic and electrostatic contributions to the sandwich energy. They concluded that the elastic contribution to the sandwich energy was the dominant mechanism for the kinetics of intercalation while the electrostatic interactions were important in determining the equilibrium state or stage of the GIC. The 3 is somewhat similar to that of recent model of Hawrylak and Subbaswamy Safran (for modifications see reference 3 and references therein), but takes into account volume and inplane intercalant density variations, 6'7 in and is capable of explaining experimentally observed effects binary alkali GIC's which had not been addressed in earlier models. A binary GIC is one in which a single chemical species has been intercalated into the galleries between the layers of the graphite host. While a theoretical description of binary GIC's based on elastic and electrostatic energies can explain many experimentally observed phenomena, other effects which couple those energies are also significant. The coupling of elastic and electrostatic forces would be experimentally visable in a variation of ds with composition and/or charge exchange (the amount of charge per intercalant exchanged between the intercalated species and the carbon sheets). Thus, any acceptable model of staging must adequately account for the variation in the sandwich thickness dS with composition and charge exchange. To date, the experimental data necessary to test theoretical predictions for the variation of ds with those parameters has been at best, sparse. Furthermore, the composition of the binary alkali GIC's, which are the most widely-studied and well-understood GIC's, cannot be conveniently varied at constant stage. The variation in the composition at a fixed stage allows a more detailed study of the role of charge exchange in determining the sandwich energy. For example, Woo and coworkers8 reported a small (.31) composition dependent change in as for stage 2 LiC12 in comparison with stage 2 LiC18, which they attribute to a competition between electrostatic and elastic effects. Similarly, Metrot et al.9 noted a weak dependence of ds on the amount of chemical overcharging of stage-1 H280“ GIC's, but chose to focus on the chemical aspects of their measurements rather than to explore in detail the coupling of ds to the charge exchange f. It should also be mentioned that large variations (~101) in (18 with composition x have been reported for the ternary GIC's MxM'1-xc81o’“ where M and M' are alkali metals, but such variations are predominantly elastic in origin10 and preclude a study of (18 vs. f. A ternary GIC is one in which two distinct guest species simultaneously occupy the carbon interlayer space. We have shown1‘2 that the K-NH ternary GIC's permit an indepth 3 study of the variation of sandwich thickness ds with composition and charge exchange f, at constant stage. The K-NH intercalants form a 3 two-dimensional (2D) liquid in the graphite galleries,13 which represents a natural 2D counterpart of the well-studiedm'15’16 3D K-NH 3 solution. In particular, the M-NH3 (M = alkali metal) GIC's offer the possibility of exploring the nonmetal-metal transition17 in 2D. In view of this, iizsis interesting to note that several decades ago Rudorff et al.18 prepared limited compositional forms of those compounds at low pressure, but they have only recently been investigated in detail.13’19’20 In this paper we will discuss and extend the coupled elastic energy-charge exchange model which has to date only been addressed in a preliminary report.12 We will provide here new experimental details and additional physical insight into the K-NH ternary GIC's. Finally, the 3 approach taken here includes modifications that we feel provide a more global understanding of the K-NH3 GIC's. II. Experimental Details and Results All K(NH3)XC2u samples were prepared by using a well characterized stage-2 KC 11 sample made in the usual manner21 from the host material 2 highly-oriented pyrolytic graphite (HOPG). The KC samples were 21: rapidly transferred from their preparation vessel to another container, which was ultimately exposed to previously cleaned NH3 gas. The transfer process was carried out in a Vacuum Atmospheres model M0-N0-1 glove box maintained at <.5ppm 02/H20 levels. Many KC samples were prepared during the course of this work and 2H their stoichiometry was carefully determined by accurately weighing the starting HOPG and the resultant intercalation compound. We found that the actual composition of the pure stage-2 compound was KCZIHG’ where -2.20 :1: .05 < 6 _<_ 0. Nevertheless, we will follow custom and hereinafter often designate specimens by their nominal compositions, e.g. K(NH3)x02u, but the actual composition, will be used in any relevant analysis. Commercial grade NH (main impurities are: “Oppm 02, 5ppm H20) was 3 purified by condensing it onto Na metal with liquid nitrogen. The NH3 was then warmed to -70°C with a dry ice/alcohol mixture. The solution was again frozen and any residual H gas was pumped away. Hydrogen may 2 be formed from the decomposition of NH3 to the amide NH2 by the reaction of NH3 with impurities. This "freeze-pump-thaw" procedurez‘2 was repeated until there was no evidence of the evolution of H determined 2! by monitoring the pressure with a standard vacuum-thermocouple gauge. Absorption isotherms or weight uptake measurements of KC for NH 2" 3 gas were determined by two techniques: The first technique employed an 23 archaic but quite useful device now referred to as a McBain balance. This balance consisted of a very sensitive Hook's law quartz spring (force constant k - 1mg/mm, equilibrium length 9.0 - 300m), that was calibrated with premeasured wire weights. The spring was vertically suspended from a specially-designed flange in a glove box with the low H20/02 levels indicated above. Extreme care was used in attaching the KC specimen (mass ~100 mg) to the base of the spring. A thick-walled 211 glass tube designed to withstand >10 atms pressure was used to enclose the specimen-spring assembly and was connected to the flange through an 0-ring teflon coupling. The entire apparatus was removed from the glove box, connected to a vacuum system, and then exposed to clean NH gas 3 from a reservoir containing excess NH liquid. 3 Vertical displacements from the KCZII equilibrium position which resulted from the intercalation of NH3 vapor were monitored as a function of NH3 pressure with a Wild KM-326 cathometer with an accuracy of <.05mm or equivalently (Song. By controlling the temperature of the excess liquid in the NH reservoir, the ammonia gas pressure, PNH , 3 could be determined from the equation of state for the saturated vapor 3 pressure of NH (in atms), as a function of temperaturezu given in Eq. 3 1. _ 191u.9569 1°31OPNH (T) - 27.37600N T - 8.H5983 log1OT 3 + 2.39309 x 10’3T + 2.95521u x 10'61'2 . (1) Corrections of the McBain balance data for bouyancy, significant only for P > 1 atm were made by determining the density of NH gas NH 3 from Van der Waals equation of state25 with PNH given in Eq. 1, and the 3 volume of the sample as follows: The area of the sample could be 3 calculated from the weight, density, and d-spacing or c-axis repeat distance of the host material HOPG. [Note that if the intercalant layers are disordered and/or unregistered with the graphite layers, then d - [ds + hrfi) 3.35]A, where n is the stage number and 3.35A is the interplanar'distance in pristine graphite.] The volume of the specimen can then be determined from the d-spacing of the NH3 intercalated material which in turn is obtained from (002) X-ray diffraction measurements. After the sample was initially intercalated with NH at 3 room temperature, the bulk composition of the specimen was reversibly varied over the range 1.149 < x < 11.33 (where x is the mole ratio of NH3 to K) for 10-3 atms < PNH < 10 atms. The McBain balance absorption 3 data, corrected for bouyancy, is shown in Figure 1. The second technique for weight-uptake measurement3»utilized1a specially-designed high-pressure glass manifold, which was connected through a 3mm Ace glass-teflon valve to a vacuum system with an ultimate 7 Torr. An NH reservoir, a sample chamber, a standard 3 volume, and a Datametrics Barocell Type 590D-100P-3P1-H5X-HD capacitance vacuum of 5 x 10- manometer were also connected to the manifold each through a separate 3mm glass-teflon valve. The volume of the manifold and each chamber was determined by a series of expansions at a known initial pressure of an inert gas such as argon from the standard volume into a previously evacuated manifold or chamber. From a measurement of the final pressure MOLE FRACTION X N 9° P 5" b o o lo 7‘ (D 9 o I I I 1 l 0.0 1.0 2.0 3.0 7.0 8.0 9.0 10. J 4.0 5.0 6.0 PRESSURE NH3 (81”) Figure 1. The dependence of the mole fraction x (x - mole ratio of NH3/K) of K(NH3)xC2u with PNH3’ determined by a McBain balance19 method. Experimental points are indicated by solid dots (o), the solid line is a guide to the eye. resulting from the expansion, the volume of each chamber and manifold could be calculated and then averaged over several trials to yield an accuracy of < 1%. By monitoring the pressure difference before and after intercalation, together with the predetermined volumes, the number of moles of NH3 taken in by the graphite can then be calculated from Van der Waals equation.25 This "gas-handling" technique, the results of which are shown in Figure 2 as solid dots for intercalation and open circles for deintercalation, compliments the McBain balance technique, the data of which are also shown as open triangles in Figure 2. The former allows access to the low P, low x region of the initial intercalation of KCZN with NH3. In fact, for our typical samples (mass ~50mg), the inclusion of the low pressure region of Figure 2 necessitated measuring the NH3 content intercalated into graphite to an accuracy of .1u mole. However, the "gas-handling" technique has a slightly higher error (11 vs. .2” compared with the McBain balance method at the high pressure end. Note also that there was a small amount of adsorption of NH3 onto the walls of the glass manifold that will affect the accuracy of x at low pressures (<10D2 atms), but this was minimal and should not change the qualitative feature of the weight-uptake curve of Figure 2. All x-ray measurements and sample characterization reported here were accomplished utilizing a Huber model l130-11110-512 lI-circle diffractometer coupled through a vertically-bent graphite monochrometer to a Rigaku 12-Kw rotating anode x-ray source equipped with a Mo anode. The x-ray signal from a Bicron NaI detector was fed into a Tracor- Northern 1710 Multichannel analyzer (MCA). Both the MCA and Huber diffractometer were controlled using a DEC-PDP-1103 digital computer, 10 1 10 A 9 oA°A¢:°. on.“ oRooooo°O°°°°° o 10 ° X 2 -1 010 . l: . ,. . ' o o .. . .00 o < E10 Ill . _l g. 16 ' 16 1 111111111 141111111 111111111 111111111 111111111 ' _g .73 —2 -1 O 1 1O 10 10 10 1O 10 PRESSURE (atm8-1 Figure 2. The variation of the mole fraction x with PNH for the 3 ternary GIC K(Nfi3)xC2u as measured by a "gas-handling" technique (see text). Experimental points indicated with a solid dot (0) represent the initial intercalation of NH3, those with open circles (o) are for deintercalation. The McBain balance data from Fig. 1 has been replotted as open triangles. 11 with specially-designed software for automatically repeated scans along any predetermined path in real or reciprocal space. In situ x-ray diffraction studies as a function of PNH were also 3 performed on the ternary GIC K(NH A KC sample (~.7mm x 6mm x 3’xC2u' 2h 10mm) prepared in the manner described above, was placed in an 8mm pyrex tube with excess liquid NH The KC sample was initially intercalated 3' 2A with NH3 at room temperature, and was found to be a pure stage-1 compound (d - 6.633A) with a stoichiometry K(NH ) determined by 3 u.33czu weight-uptake measurements. We then maintained the sample at room temperature for all subsequent measurements, but in order to change the sample composition, i.e. change x in K(NH3)xCZA’ we varied the temperature of the excess NH3 as follows: A dewar of ethyl alcohol was placed around the NH3 end of the sample tube and the alcohol was slowly (~20°K/hour) cooled (to -110°C) to avoid exfoliation of the specimen. The ethyl alcohol was then replaced with liquid nitrogen as the low temperature bath. The dewar of liquid nitrogen was removed and a dewar containing an ethanol slush at - 110°C was placed over the NH end of the sample tube and allowed to warm 3 up slowly (~11.14°K/hour) to room temperature. The ammonia pressure was again determined by monitoring the temperature of the excess NH 11 id 3 Q“ and applying Eq. 1. In situ x-ray (002) diffraction scans were taken at one hour intervals as the NH warmed. Representative scans are shown in Figure 3 3 which depicts the pressure evolution of the K-NH GIC from stage 2 to 3 stage 1. The pattern of Figure 3(d) is associated with a change in the color of the sample from dark metallic blue to a yellow-gold and reveals that the sample consisted of a mixed phase system composed mainly of the 12 ‘1 3.8.3 3.5232. 1.1a; ES: . MW Es: . 1W m. :3: mm m. I\ m. l.\ m. ASSN 11\ +3. 89ch (1 3a 1 t t H + I . as: m as: . m 3%: i as: m a 9. a. as; . um: 22 . :3 as: 1 1 __ 3 33: l 3 38: i N 63: I H 32.3: . EN“ M :83 . P PN P P 22 is: .2. is. 1. 2212...: 1 as: 1 an. . a? 1 as: i I . 2% 1 22 . IVER... as: he: . .ame. . . . . .. 2.2. s22. ... .. .. . . :4AMIQTQVN 1. . s . r .f AfiaavN 0.. . . . . s... AMQGVN‘ .. . . . 38;... . as: . .82 . $221 3 :99; I.. o .. . 1.“ :69: 1.3.5; =ch— 3...i c1619.!— ==SN I ASSN i :SENK EEK L b — _ — _ b _ _ _ b h _ L p _ _ .15 (:13.- (a), (b), (c), and (d) represent the PNH evolution of the 3 Figure 3. (002) x-ray diffraction patterns of K(NH3)XC2u- Reflections Those associated with the binary potassium GIC's associated with the stage n K-NH3 ternary GIC's are labeled n(002). are labeled nK(00£). 13 stage-2 K-NH ternary GIC with admixtures of the binary GIC's KCZM (d - 3 8.6711) and KC8 (d - 5.33A). The "yellowish" appearance of the sample surface is due to the presence of KC8, which is normally gold in color. Observed reflections in each pattern are of instrument-limited width and thus, for the 1-mm slit settings used, indicate a c-axis correlation range of ~350A. Shown in Figures 11 and 5 are the d-spacings of stage 1 (d1) and NH3' The NH3 pressure dependence of the intensities of the (001) reflection of stage 1 and the stage 2 (d2) K-(NH3) GIC's as a function of P (002) reflection of stage 2 is shown in Figure 6 and clearly indicates a phase transition from stage-2 to stage-1 with increasing pressure. The inflections visable in the intensity data of Figure 6 at P - 5.5 atms NH3 are due to compositional changes and will be discussed below. For PNH 3 > 6.5 atms, only the (001) and (002) reflections of the stage-2 region are detectable in the (00$) x-ray diffraction patterns. The (003) reflection of stage 2 is masked by the strong (002) reflection of stage 1. The large fluctuations visable in (12 vs. PNH of Figure 5 are due to 3 inaccuracies associated with determining d from only the two observable 2 low-angle reflections cited above. The dashed line in Figure 5 is an extrapolation of the d2 vs. PNH curve to PNH - 9.5 atms (corresponding 3 3 to T room temperature) with d2 - 9.930A. Both the d-spacings of 3 Figures 11 and 5 and the weight-uptake measurements of Figures 1 and 2 show rapid changes for PNH < 1 atm and establish that the c-axis 3 expansion of the K-NH3 GIC is due to the influx of NH3. 6.630 6.620 6.610 6.600 6.590 6.580 6.570 6.560 6.550 6.540 6.530 6.520 6.510 d, (A) Figure u. 14 1 [VjTIIIWF'I‘I'II1'V'fi ‘- ' I V r V V i 1 l . l 1 l 1 l i I . l I l 1 l . 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 Pressure N143 (atms.) The variation of the d-spacing of stage-1 K-NH3 GIC (d1) with PNH . The experimental points are shown as solid dots 3 (o) with the indicated error bars, and the solid line represents a theoretical fit to the data using Eq. 18. <12 (A) 15 9.99 9.96 l 9.94 . 9.92 ’ 9.90 9.88 9.86 i 9.84 ’ 9.82 h 9.80 C- 9.78 l 9.76 L i J l 1.0 2.0 3.0 4.0 5 .O 6.0 7.0 8. J O 9.0 Figure 5. Pressure NH:3 (atms.) The dependence of the d-spacing of stage-2 (d2) K-NH3 GIC with PNH . Experimental points are indicated with a solid 3 dot (o), and the solid line is a guide to the eye. The - 9.5 dashed line is an extrapolation of the curve to PNH 3 12 16 10- (cos) I Inuuudhr x10 1(001) x10 2(002) Figure 6. PhH(abn) The dependence of the intensities of the (001) reflection of stage-1 and the (002) reflection of stage-2 K-NH3 GIC's with PNH (indicated with solid dots o). 3 III. Discussion and Analysis A. Comparison with Rudorff's Results on K-NHa GIC's .2 Before proceeding with the analysis of the above experimental data, it will be useful to discuss the stage-1 and stage-2 K-NH GIC results 3 of Rudorff and coworkers18 in order that comparisons can be made with our results. It will also be necessary to discuss the physical implications arising from the weight-uptake curve of Figure 2 and the x- ray diffraction pattern of Figure 3d before a method of analysis can be established. Rudorff et al.18 showed, over 30 years ago, that metal ammonia solutions (M-NH where M is any alkali metal) could be readily 3 intercalated into graphite. They prepared a stage-1 K-NH3 GIC via two techniques: The first technique was to directly immerse the binary GIC KC into liquid NH They found that upon intercalation of NH K was 8 3' 3r expelled from the specimen. A second technique was to directly immerse graphite powder in a metal rich K-NH3 solution. Both techniques, after the removal of excess absorbed NH3 at 0°C, were reported to yield the ternary GIC K(NH3)2 0012 with a d-spacing of 6.5A. A stage-2 K--NH3 GIC was also prepared by controlling the concentration of the K-NH3 solution into which the graphite powder was immersed. Rudorff et al.18 found, after the evacuation of NH that only for a carbon/potassium molar 3! ratio of 28 was a pure stage-2 compound formed with an average stoichiometry of K(NH ) 3 2.3028 and d-spacing of 9.9A. 17 18 From the results of Rudorff and coworkers, it seems energetically favorable at low NH3 pressures, for K(NH3 carbon/metal ratio of 12 for stage 1 and 28 for stage 2, and an 1 )ny GIC s to acquire a NH3/metal ratio of 2.0 for both (stage 2 may have a slightly higher NH3/metal ratio). Thus, potassium was expelled from the binary GIC KC8 upon submersion into liquid NH according to the reaction 3 3KC8 + “(NH3) + 2KC12(NH3)2.O + K. (2) Upon the removal of the absorbed NH in a vacuum, both stage 1 and stage 3 2 K-NH3 GIC's still retain large amounts of NH3 respective galleries. Thus, at low pressures these K-NH (x - 2.0) in their 3 ternary GIC's are residue compounds. We have also prepared pure stage-1 K-NH GIC's in a manner similar 3 to that of Rudorff18 (i.e. HOPG + K-NH liquid and K08 + NH liquid), 3 except our samples have excess liquid NH 3 3 in the sample vessel. The (009.) x-ray diffraction patterns of KC8 + NH liquid and HOPG + K-NH 3 3 liquid are shown in.Figures 7a and 7b, and indicate that the resultant stage-1 ternary GIC's have very similar d-spacings of 6.516A and 6.586A, respectively. These values are in reasonable agreement with Rudorff's results of d - 6.5A for both methods. Furthermore, our KC8 + NH liquid 3 samples upon intercalation exfoliated, a process which appears to be associated with the rapid expulsion of potassium as reported by Rudorff.18 Nevertheless, it is reasonable to question whether the stoichiometry of our stage-1 compounds prepared from HOPG + K-NH liquid 3 8 + NH3 liquid is in fact the same as reported by Rudorff,18 since at high NH and KC pressure it is possible to either expell potassium or 3 intercalate additional ammonia. Figure 7. 19 33 o 9.— (a) :: 0 El 13 3 - z '. . C . °. A 5 r E F? g; ,l a !- 1. 8 ° 8 N '3 ' . V T O .9 l l 1 l 1 l l v I a v E (b) m I: 0 a A 3 cu - 0 El 3 :3 o 51 . F? Q? —~ a~ .. O O m (D . ' 9... 9— 8 ° 1 1 1 J I 1 26 (Degrees) The room temperature (00%) x-ray diffraction patterns of (a) stage-1 KC8 + NH3 liquid, and (b) stage-1 HOPG + K-NH3 solution. The notation is the same as that of Fig. 3. 20 Given a fixed carbon/potassium ratio of 12, it is possible to estimate the maximum intercalation of NH3 at 9.5 atms. The "effective" area of an NH3 molecule can be estimated on the asumption that NH3 completely fills the available area of the graphite gallery for stage-1 K(NH ) The "available area" for the unit of composition 3 H.3302u' K(NH3) of stage 1 is A = 56A2 and it was determined by u.3302u subtracting the area of a potassium ion of radius 1.46A (see below) from the area of 211 carbon atoms. From A - 56112 and xmax - 11.33. the "effective area" of an NH3 molecule ANH is 13.0713 which following a 3 similar analysis leads to a value x - 2.0 for the compound max K(NH3)XC12. Thus, if no potassium is ejected, the K(NH3)XC1 compound 2 becomes saturated at the low pressure (vacuum) at which x - xmax - 2 and ingests no more NH3 even when PNH is raised to 9.5 atm. 3 The d-spacing of our stage-1 compounds prepared from HOPG + K-NH3 liquid is ~6.5A even with PNH - 9.5 atm and lies very close to that 3 8 (d - 6.5A). In contrast, our stage-1 reported by Rudorff1 u 33C2u 2“ with NH3 has a d-spacing of 6.633A. Therefore, it is indeed evident that large K(NH3) compound which was prepared by reacting KC amounts of potassium are not expelled from the K(NH3)XC system when it 12 is exposed to elevated NH pressures. 3 B. The Absorption Isotherm We will now describe and explain the physical implications of the weight-uptake curve of Figure 2 and the x-ray diffraction pattern of the residue compound of Figure 3d. 21 As PNH is increased, the weight-uptake curve in Figure 2 (solid 3 dots) shows that little intercalation takes place until a pressure of 10-3 atms is reached, then rapid intercalation sets in. This rapid influx of NH3 could be attributed to an activation energy associated with separating the graphite layers enough to allow a small amount of NH to enter. For ammonia pressures >10-3 3 reached with x - .03. This plateau region extends from 10- atms a plateau region is 3 atms to .6 atms and is followed by another rapid influx of NH reaching another 3 plateau region at x - 11.0. This transition at PNH - .6 atms is also 3 concurrent with a stage transformation from stage 2 to stage 1 (refer to Fig. 6) and is accompanied by a dilution of the inplane potassium density from KC12 to “211' At 9.5 atms the sample is a pure stage 1 with a stoichiometry of KC2u(NH3) As the pressure is lowered from 11.33' 9.5 atms (open circles of Fig. 2), x now follows a new path and as PNH 3 + 0, x + 1.H9, which is consistent with the fact that the compound which 18 results from pumping off the NH is a stage-2 residue compound. 3 Akuzawa et al.26 have also prepared K-NH GIC's using quite 3 different preparation techniques. They first prepared a stage-2 18 K(NH3) using the methods of Rudorff, then evacuated the sample 2 .0C28 chamber at an elevated temperature and reintercalated to stage 1 with potassium vapor. Akuzawa and coworkers report a stage-1 compound with a d-spacing of 5.11:1, slightly larger than that of KC of 5.3511 and a 8 stoichiometry from chemical analysis of K(NHB) 0306' The exact composition of the samples prepared by Akuzawa et al. is somewhat in doubt since NH3 has a tendency to decompose at the elevated temperatures used to reintercalate potassium. The similarity in the K/NH3 ratio x a 22 .03 of Akuzawa's sample and the x value of the first plateau region in Figure 2 is striking in light of the totally different preparation techniques involved. This similarity indicates that once this first plateau region at x - .03 is reached, the weight-uptake curve (see Fig. 2) is no longer reversible and subsequent desorption of NH yields a 3 stage-2 residue compound, K(NH ) 3 .03C2u° Although we have not as yet measured the d-spacing in that first plateau region of Fig. 2 at composition K(NH3) 11’ we expect to find .03C2 a carbon-intercalant-carbon sandwich thickness similar to that of ) K(NH which according to Akuzawa et a1.26 is approximately 3 .03C6 d - 5.111: as noted above. This similarity is not surprising since NH3 sparsely populates the graphite galleries (for x - .03 the NH3-NH3 distances are ~100A) and graphite can easily accommodate a sufficiently small density of local elastic distortions (due to intercalation) without altering significantly its sandwich thickness.27 From the above discussion, a plausible explanation of the first rapid influx of NH at 10"3 3 the carbon-(K-NH3)-carbon sandwich is small since the sandwich atms is as follows. The elastic energy of thicknesses of K(NH ) A (ds - 5.AA) and KCZu (ds - 5.35A) are nearly 3 .03C2 the same. Thus, even at these very low pressures, the NH pressure is 3 still large enough to overcome the energy associated with creating sparsely-populated elastic distortions, thereby intercalating .03 moles of NH per mole of K into the galleries. The plateau region of x - .03 3 and the second rapid intercalation of NH3 at PNH - 0.6 atms can be 3 explained in a similar fashion. Since NH3 has a high mobility in the galleries of graphite,19 to significantly increase x beyond 0.03 may require the sandwich thickness on the average to be much larger (for 23 example, ds - 6.5A as measured on deintercalation at 0.6 atms.) than that of K(NH3) 11 (d3 - 5AA). Thus, the elastic energy involved .0302 with increasing x is large and the pressure to overcome this energy is also large. Thus, x remains constant at x ~ .03 for several decades of ammonia pressure until rapid intercalation sets in at the critical pressure of 0.6 atms. C. The Residue Compound K(NH3)1.A9EQA As noted above, the residue compound represented in the pattern of Figure 3d contained three phases: stage-1 KC8, stage-2 KC , and the 211 stage-2 K(NH The bulk stoichiometry of the latter was 3)1.u9C2u° established from the weight-uptake measurements shown in Figures 1 and 2. These results at first glance are not consistent with the stoichiometry of the stage-2 residue compound K(NH3) 8 prepared by 2.3C2 Rudorff, but at this point, it is not clear that such a comparison can be made given the difference in the initial conditions of sample preparation. Since no potassium is expelled from K(NH )xc211 as x is reduced the 3 presence of KC8 in the x - 1.119 residue compound requires that for a fixed carbon/potassium ratio of 2“ there must also exist expanded regions of KCy where y > 211. It is interesting to note that the formation of 75% of K(NH3)ZC28 is entirely possible Just based on balancing the NH content for the two compounds (i.e. 75% of x - 2.0 + x 3 - 1.5). From the above observations it is possible to describe the three phase system of Fig. 3d with the following reaction K(NH ) 3 1.u9C2u + a K(NH3)ny + b KC8 + 0 K02“ (3) 24 The relative magnitudes of a, b, and c and the mole fractions x, y can be determined by simultaneously solving the three mass equations Aa, Ab, and lie that balance Eq. 3 for K, C, and NH plus Eqs. 5a and 5b, 39 which represent the ratio of the intensity of the (003) reflection of K(NH3 reflection of KCZU to the (002) reflection of KC8 respectively. )ny to the (002) reflection of KC8, and the ratio of the (003) 1-a+b+c(11a) 2M . ay + 8b + Zflc (Nb) 1.H9 - ay (“0) v1 2' " a'sz(qoo3,x,y)l LP(Q003) - 12(0003) (53) b|s1(q002)|2LP(q002) 11(3002) 2 c|82(q003)| LP _ 12(qoo3) (5b) 2 b|S1(0002)| LP(qooz) 11(q002) Here q002’ q003' and q003 are the wave vectors (q - "Heine/A) associated with the (002) reflection of KC8, the (003) reflection of K(NH )ny, and 3 the (003) reflection of KC211’ respectively. Also shown in Eqs. 5a and 5b are the structure factors S1(q), Sé(q), 82(q) (which is also dependent on composition x,y), and the intensities I1(q). Ié(q). 12(q) for KC8, K(NH and Kczu, respectively, as well as the combined 3)xcy’ Lorentz polarization factor LP(q). Debye-Waller and absorption corrections were omitted from Eqs. 5a and 5b and are epected to produce only marginal differences in the results since q002 - C1003 and (1002 0 q003' 25 The parameters that result from simultaneously solving the above five equations are: a - .71, b - .211, c = .05, x - 2.1, and y - 29.11, whicni to within experimental uncertainty are consistent with a - .75, b - .19, c - .063, x - 2.0, and y - 28, so that Eq. 2 can be rewritten in the following form: K(NH3)1.5C2u - .75 K(NH3)2.0028 + .25 KC12 (6) where KC12 - .75 KC8 + .25 K02“. Equation 6 suggests that our stage-1 K(NH3)u 33C2A sample evolved as the pressure was decreased to the stage-1 and stage-2 Rudorff compositions given below. K(NH ) (7) 3 2.0C2N - .75 KC28(NH + .25 KC12(NH 3)2.o 3)2.o This reaction is expected to occur at PNH - 10"2 atm, as estimated from 3 Figure 2 (i.e. for PNH - 10-2 atms, x ~ 2.0). 3 Upon further lowering of the ammonia pressure below 10-2 atms, a dynamic exchange of NH between stage 1 and stage 2 and the environment 3 may have caused the depletion of NH in the stage-1 region yielding the 3 unstable KC12 component which ultimately phase separated into KC8 and KCZN' A possible senario for the depletion of NH3 in stage 1 is the following: As PNH is lowered below 10-2 atms, the composition evolves 3 to x < 2.0; this is an energetically unfavorable situatnnifor boui stages. However, if as we show below, stage 2 has a higher affinity for NH3 than does stage 1, it will "pull" the NH3 gallery from the stage-1 region. Then any excess in the stage 2 region, across the graphite 26 13,15,16 above x - 2.0, will be desorbed. It has been established that 18 stage-1 K(NH3) is a donor GIC, thus, Rudorff's K-NH compounds u.33czu 3 are also likely to be donors. If a direct comparison can be made between ternary alkali-ammonia and binary alkali donor GIC's which have charge exchanges of f - 1 and f . .86 for the stage-2 and stage-1 compounds, respectively, then the charge on the potassium ion of stage-2 K(NH3)2.0C28 3)2C12. A greater charge on the potassium ion results in a stronger charge-dipole bond is greater than that for stage-1 K(NH with NH3, which in turn requires that the stage-2 K-NH affinity for NH 3 GIC has a higher 3 than does stage-1. This would also explain the significantly higher NH /metal ratios of the other stage-2 alkali and 3 alkaline earth NH3-GIC's also reported by Rudorff.18 D. The Generalized Reaction of KOO“ with NHa (.w 3 From the above discussion it is likely that the compositions of the two stages are pressure dependent and Eq. 7 should be generalized in the following manner K(NH3)xC2u - p1(NH3)x1Cy1 + DZK(NH3)x2Cy2 (8) where yn, x , and pn are the mole ratios of carbon/potassium, n NH /potassium, and relative fraction of each stage n, all of which are 3 pressure dependent and have yet to be determined. To solve unambiguously for the above variables would require six independent equations. There are five equations that are readily at our disposal. These are the three mass equations that balance equation 8 for potassium, carbon, and NH listed in equations 9a, 9b, and 9c 30 below, plus the two intensity equations 10a and 10b for the (001) 27 reflection of stage-1 K(NH )x C and the (002) reflection of stage-2 31y1 K(NH) C . 3 x2 3'2 P1 1‘ P2 '1 (9a) p13!1 + pzy2 - 2“ (9b) p1x1 + p2x2 - x (90) 2 . 10 p1|S1(qOO1,x1,y1,Y1)| LP 6.0 atms since, to a very good approximation, F’1 - 1.0 in 3 that region (see Fig. 6). The notation used in Eqs. 9 and 10 is the following: is the wavevector associated with the (002.) reflection q009. of stage n, In’ and Pn are the intensities and fraction of each stage, LP(q) is the combined Lorentz polarization factor, and Sn (q002,xn,yn,Yn) is the structure factor for stage n, which depends on the wave vector q009.’ the composition (xn,yn) and a charge exchange parameter (Yn). Incorporated into the structure factors Sn(qn) are the charge exchange terms f - 1 - Yn(Xn - A) where Y - 0 and Y - .uzu were 2 1 15 of K(NH3)11.33C211' terms will be considered in more detail later. inferred from NMR measurements The charge exchange 28 In order to solve for the variables in eqs. 9 and 10 as a function of ammonia pressure an additional constraint is required. Such a constraint can be found by setting x - 2.3. We have shown in earlier 2 work12 that x2 is essentially pressure independent and x2 pressures. However, small pressure dependent compositional changes in - 2.1 at low y1 and y2 were not previously considered, but do not alter the basic pressure insensitivity of x Therefore, since x is pressure 2' 2 insensitive, y2 can be expected to follow suit (this assumes of course that NH3 completely fills the available gallery area determined by yz). Rudorff reported that at very low ammonia pressures the stage-2 K- 18 NH ternary GIC had an average stoichiometry of K(NH which is 3 3)2.3C28 consistent with our ternary GIC represented in Fig. 3d. If y2 remains constant at 28, we can estimate the value of (x2)max by assuming that that is NH 3’ 3 completely fills the available interlayer space. From the effective the stage-2 ternary GIC is maximally ocupied by NH area of an NH3 molecule previously calculated, we find (x2)max - .101 , y2 - .512, and (x2) - 2.3 for y2 - 28. Therefore, to a good max approximation, x is pressure independent and equal to 2.3. 2 With the above constraint, yn,pn for each stage n, and x1 can be determined by simultaneously solving Eqs. 9 and 10. This is accomplished by reducing these equations to a 6th degree polynomial of one variable. The resulting solutions for the variations of y1 and y2 with PNH are plotted in Fig. 8. It can be seen from Fig. 8 that as 3 PNH is increased, y1 increases from Rudorff's result of y1 - 12 (open 3 circle) to y1 - 211, while y2 remains relatively constant at Rudorff's value y2 - 28 (open circle at low pressure) until P < 5.0 atms, where N H3 29 36 34 32 30 28 26 24 22 20 1 8 1 6 14 12 I I 1 I T l U Ii 'I'IIIIU Carbon/Potassium Ratio Yn '_ L L L Lit. O- J .0 c Figure 8. l . l 1 l L l J l i l i l . l . J 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 Pressure NH3 (atms.) The dependence on PNH of the carbon/potassium ratios yn for 3 each stage n K-NH3 GIC. Analyzed data points are indicated with a solid dot (o). The solid lines are a guide to the eye. Thefdashed line for the y1 vs. P curve is an NH 3 extrapolation to the Rudorff18 C/K ratio y1 - 12. The open (circles represent the C/K ratios reported by Rudorff18 for stage 1 and 2 K-NH3 GIC's. The solid dot with a concentric circle (9) represents an estimate of y2 at PNH - 9.5 atms 3 based on an analysis of the x-ray data shown in Fig. 10 (see text). 30 it rises sharply. This sharp rise in y2 for P ~ 5.0 atms implies NH3 that our original assumption of the pressure insensitivity of y? is valid only in the range PNH 3 is the pressure dependence of x < 5 atms. However, our immediate concern With the present approach, x1 should .3 6.0 atms. (see 1. be accurate up to P < 5.0 atms and alsoforPNH 3 discussion of eqs. 9 and 10). Thus, the sharp rise in y2 does not pose NH3 a serious problem for determination of x1(PNH ). 3 Small fluctuations of y1 and y2, visible in Fig. 8, are due to the artificial constraint of holding x constant at 2.3 and can be largely 2 removed by allowing small variations in x "“3 the (001) reflection of stage 1 and the (002) reflection of stage 2 2. The larger fluctuation in y2 at P - 3.2 atms corresponds to an inflection in the intensities of NH3 deduced from the solutions of eqs. 9 and 10 is shown in Fig. 9. The shown in Fig. 6. The dependence of the fractions p1 and p2 on P rapid changes in which p1 + 1 and p2 + 0 at PNH - 5.5 atm are 3 associated with the sharp rise of y2 at that pressure. Furthermore, the dashed line indicates an extrapolation to p2 - .75 and p1 - .25 (at low NH3 pressure) which would be consistent with the formation of the stage 1 and 2 Rudorff compositions given in Eq. 6. There are other indications that the value of y2 should be in excess of 37 as indicated in Fig. 8. The room temperature (P - 9.5 NH3 36 + NH3 liquid are shown in Fig. 10 and represent a two-phase system composed mostly of a stage-2 K- atms) (002) x-ray diffraction patterns of KC NH3 ternary GIC, but with small amounts of a stage-1 K-NH3 GIC. An analysis identical to that used on the 3-phase system of Fig. 3d (see 31 1.C)- .9 ' .81- : XI: o. 9 1 .7 H 5 1 3.6-1 g p a: .5 r g .431 P .9; .- gsa' /_2 16 DC .2 - T .1 r (3" 1 1 l L 1 1.() 2LC1 1313 45¢) £513 (51) Pressure NH3 (atms.) Figure 9. The dependence of pn , the fraction of each stage n on PNH . 3 Analyzed data points are indicated with a solid dot (o), and the solid line is a guide to the eye. The dashed line indicates an extrapolation from the pn vs. PNH curves to 3 the values of p1 - 0.25 and p2 - 0.75 at PNH ~ 0 atms 3 (indicated with open circles), which would be consistent with the formation of the stage-1 and stage-2 Rudorff18 K- NH3 GIC's (see Eq. 7 of the text). 32 600:. + amoovm A 808m 300: COOVN . 1. AVOO: + AQOOVN Goem P 88: 258m x. 8.8: + 80on 2 seem as; 3.523 >m._._wzw._.z_ Figure 10. The room temperature (PNH - 9.5 atms) (009.) x-ray 3 . diffraction pattern of KC36 + NH3 liquid. The notation is the same as that of Fig. 3. 33 Eqs. ’4 and 5) was applied to determine the values of x and y2 for this 2 2-phase system. It was found by assuming for stage 1, y1 - 211 and x1 - 11.33 and for stage 2 a maximally-occupied gallery (i.e. x - 101 , y2 - 2 .512). that y2 - 37.2 (shown as a solid dot with a concentric open circle in Fig. 8), which may indicate that at P - 9.5 atms, NH 3 K NH C is a re sta e-2 K-NH GIC. ( 3)3.2 37 p” 3 3 E. x1 vs. PNH3 and the Fowler-Guggenheim Form Figure 11 shows the variation with PNH of x1 (solid dots) fix" 3 K(NH3)XC2u and the corresponding fit to a Fowler-Guggenheim (F0) adsorption equation23 (solid line in Fig. 1) given below 0X 1 wx P -———_ e 1 (11) with a - 0.1196 x10-5 1 0.077 X105, 3 - 11.110 1 0.06, w = 2.143 1 0.0". It is interesting to note that because the carbon/potassium ratios have been allowed to vary, the PG function now provides a better fit to the data of Fig. 11 than does the Langmuir function?8 which was previously applied.12 The superiority of the F0 form is not surprising since the Langmuir function is a zero-order approximation to adsorption and does not include the interaction between adsorbed atoms. However, the Fowler-Guggenheim equation given above includes this interaction and reduces to the Langmuir form where w - 0. The fit of x1 vs. PNH shown in Fig. 11 yields the parameter to 3 which measures the interaction energy between NH3 molecules. Fowler and Guggenheim derived the relation 5 - BkTw/Z where for our use, Z is the 34 9° 0 I .N (D l Mole Fraction X1 1.0 r ........AAJ:A - C O "V'V' A Figure 11. 0.0 ‘ 1.0 ‘ 2.0 J 3.0 L 4.0 ‘ 5.0 ‘ 6.0 L 7.0 L 8.0 J 9.0 Pressure NI-l3 (atms.) The variation of the mole fraction of stage-1 X1 with PNH . 3 Analyzed data points are indicated with solid dots (e). The solid line represents a least squares fit to X vs. PNH 3 using Eq.‘Vlcfi'the text with a - 0.u96 x 10-5, 8 - u.uo, and w - 2.U3. 35 number of nearest neighbor NH molecules. The interaction energy 3 between NH3 molecules is e, kT is the usual thermal energy, and B and w are given above, winiua> 0 indicating a repulsive interaction between molecules. We find to > 0 and for a maximum close-packed structure of six nearest neighbors 5min = +0.0N eV. An estimate can be made of the maximum repulsive energycfi‘two interacting dipoles from emax - uZ/r3.29 Here, u is the dipole moment and we have taken r as the close-packed hard sphere diameter of the NH3 molecule. By using the known value for “NH . 1.5 Debye25 and r - 3.M1A 3 from diffuse x-ray scattering measurements of liquid NH3,30 emax - .035eV which is, within error, in agreement with the repulsive we find energy found from the parameter w. A repulsive interaction between NH molecules is contrary to 3 physical intuition, since one would expect dipole alignment to be energetically favorable. However, it is known from NMR results19 that the NH3 3-fold axis which is the dipole axis is dynamically tilted toward the potassium ion to which it is bound, thereby restricting the molecule's ability to orient itself. There are also steric constraints on the NH molecule which are imposed by the graphite lattice as can be 3 seen owMn Fig. 12. The inability of the NH molecule to freely orient 3 means that the probability for dipole-dipole alignment is low and that the dipole-dipole interaction could be repulsive. The accuracy of the magnitude of the interaction energy derived from Eq. 11 is not high, but its repulsive character (w > 0) is unambiguous. 36 o 5.63 A '— Figure 12. O do. K A K A The K-NH3 intercalant of the stage-1 ternary GIC has been established to be of the donor type from x-ray measurements of the a- 13 19 axis lattice parameter and from NMR and reflectivity measurements.2O Most of the charge donated by potassium remains on the carbon sheets after NH insertion, and for the purposes of this discussion, we assume 3 that this charge is distributed uniformly, leaving behind a positively- charged K-NH layer which is "sandwiched" between negatively charged 3 carbon sheets. In actuality, for donor compounds such as the alkali- 31 GIC's it has been shown that the charge on the carbon layers is not completely delocalized, but screens the positive ions by accumulating 38 around them. The carbon layers are also not rigid sheets, but can 27 accommodate distortions perpendicular to the layer. Nevertheless, a model which describes the K-NH ternary sandwich as uniformly charged 3 rigid carbon sheets separated by "atomic" springs, constitutes a sufficiently reasonable approximation to provide useful information. 3 32 Following Hawrylak and Dahn the stage-1 sandwich energy E per carbon atom can be expressed as E - --o:x3/2 + 9d x2f2 + 1/2 k (d - d )2 + K 31 K 0 s1 0 2 1/2 kaK(ds1 dK) + 1/2 xAkA(ds1 d 2 ) (12) A Here a and Q are constants, dS is the equilibrium sandwich thickness of 1 stage 1 and xK and xA are the concentrations of K and NH3 in units of the number of carbon atoms, i.e. xK - NK/NC and xA - NA/NC' where NJ is the number of atoms of type J in the bulk specimen for a particular stage. In EH1. 12 the first term represents the binding or cohesive energy of a 2-dimensional free electron gas while the second term is the electrostatic energy associated with separated layers of charge. The remaining terms represent the elastic energies associated with pure - 0) and ammonia (x = K for the potassium graphite (xK - x - 0) and potassium (x - 0, x A A K 0, x = 0) GIC's. The equilibrium length d A K intercalant is related to the size of the K species which in turn depends on the charge transfer, f, from the potassium ion to the carbon sheets through the relation33 rK = er+ + (1 - f)rKo (13) 39 where rK is the radius of the K ion with charge exchange f, and rK+ and rKo are the ionic and atomic radii of potassium. Enoki et al.33 have used this relationship to determine the size of K and Rb ions in alkali binary GIC's and were able to explain the differences they observed in the ESR signals of hydrogenated KC and RbC8. In order to relate r in 8 K Eq. 13 to dK’ the ionic and atomic equilibrium lengths (or sandwich thicknesses) d + and d 0, respectively, must be established. To a very K K good approximation dKO and dK+ can be determined from the following relations: dKo . 2rKo + Ro (1ua) dK+ . 2rx+ + Bo (1ub) where the geometrical factor R0 - 2.69A was obtained from the experimental value of d + - 5.35A with rK+ - 1.33A.33 Using Eqs. 111a K and 1Nb, Eq. 13 can be rewritten as dK = de+ + (1-f)dKo (15) where dK - 2rK + Ro . It can be expected that the charge exchange f in Eq. 15 is 1‘1,15,16 dependent on the NH concentration x since it is well known 3 A' that NH3 in metal-ammonia (M-NHB) solutions has a strong affinity for electrons. Therefore, it is useful to review briefly below the properties of bulk 3-dimensional (30) metal ammonia solutions which illucidate the relationship between f and x in the 2D analogue A K(NH3)XC2u. As alkali metal is added to liquid NH3 to form a dilute 3D M-NH3 solut ion1u’15 the metal relinquishes its valance electron to the liquid NH3 and (for M . potassium) is solvated or surrounded by six octahedrally coordinated NH3 molecules. Solvation of the expelled 40 electron also occurs, and it forms what will be referred to as an electron-cage with surrounding NH3 molecules. However, the number of NH3 molecules required for electron-cage formation on the basis of energy considerations is not known exactly, but is estimated to be 311 between four and six. Ammonia is bound in M(NH3)6 complexes by charge-dipole forces, with each NH molecule oriented so that its dipole 3 points away from the metal ion. In contrast, the binding of NH in 3 electron cages is much weaker. Copeland et al.3u have shown that electron cage formation may be brought about by the self-trapping of an electron in a cavity produced by NH molecules through short-range 3 interaction with NH3 dipoles and by the long-range polarization of the NH medium. When the metal concentration is increased enough to allow 3 significant overlap in the wavefunctions of the individual electron cages, electron "hopping" from cage to cage occurs and a nonmetal to metal transition takes place. 35 the critical electron 1 /3 an aH is the effective hydrogen-like radius of the electron-cage. For K- NH solutions, the nonmetal-metal transition occurs at ~11 mole percent 3 metal (MPM - moles of metal/(moles of metal + moles + NH According to the Mott criterion, concentration n for a nonmetal-metal transition is n - .25, where 3)x1001). Increasing the metal concentration eventually leads to an insufficient amount of NH3 molecules with which to solvate the expelled valance electron of the metal. For example, in the very dilute region, MPM < 10-5, there is essentially an infinite number of NH3 molecules for each electron, but at 10 MPH there are only 9 NH3 molecules for each metal ion. Assuming 6 NH molecules are bound to the metal ion, only three 3 can couple to the electron. Thus, at high concentrations the expelled 41 valance electrons which cannot be solvated are forced th>the conduction band, causing the M-NH 36 3 solution to behave as a liquid metal. In fact it has been shown from reflectivity experiments that there is a gradual transitiion between electron cage formation and the conduction band regime. Also note that solvated cations and solvated electrons may 37.38 interact to produce neutral species either by direct cation- electron interaction, in which there are no NH3 molecules in between their respective centers, or by NH -shared ion-electron pairs, in which 3 the solvated cation and solvated electron share one or more NH3 molecules, but still retain their own identities. It is clear from the above discussion that the amount of electron solvation in M-NH3 solutions depends on the relative concentration of NH3 to that of K which can be written as X, where X - xA/xK. exists a minimum concentration Xc below which electron solvation cannot There also occur. As noted above, Rudorff showed several decades ago that metal- ammonia solutions could be interclated into graphite.18 It is surprising that neither the structural or molecular form of the intercalant, nor its 2D metal-ammonia character viz-a-viz a 2D nonmetal- metal transition, have yet been explored. The metal hexamine K(NH3)6 cannot be intercalated into graphite, since that would require a sandwich thickness of at least 1011 and the maximum value observed for the K-NH ternary GIC‘s is 6.6A. However, if 3 a pair of NH3 molecules which lie on the same A-fold octahedral axis of K(NH3)6 are removed, leaving a planar H-fold coordinated K(NHB),4 cluster, intercalation is possible. Given an interlayer space of 3.28A, which is defined to be the difference between the carbon-K(NH3)-carbon 42 sandwich thickness and the Van der Waals diameter of the carbon atoms that flank an intercalant layer, only a monolayer of NH can occupy the 3 interlayer gallery. This point is illustrated in Fig. 12. It is interesting to note that fl-fold coordination is also suggested by weight-uptake measurements of stage-1 K(NH ) with the .33 NH 3 u.33czu' 3 "spacer" molecules per K(NH3)IJ.0 cluster possibly involved in electron cage formation. Ihipndnciple, other planar N-fold coordination possibilities exist with N = 2,3,5,6. However, such N-fold configurations are very unlikely because of steric limitations and/or the fact that the potassium-NH components do not adopt such planar 3 arrangements in bulk M-NH3 solutions. Moreover, only fl-fold coordinatiom is consistent with recent inplane diffuse x-ray scatteirng 13 experiments on K(NH3)xC2u° The entrance of NH3 into the graphite galleries results:h1the solvation of both potassium and of electrons originally donated by potassium to the graphite layer. The electron solvation is a manifestation of the collateral extraction of charge out of the carbon sheets together with the formation of electron cages. fim~the most dilute case, which corresponds to stage-1 K(NH ) the 3 u.33C2u' concentration of potassium relative to NH is still - 19 MPM, well into 3 the metallic region of a bulk 30 M-NH3 solution with the same concentration. Thus, it is very likely that the K-NH intercalant layer 3 acts as a 20 liquid metal with delocalized electrons, which are collectively shared by each potassium ion. From the above discussion, it can be expected that as NH is added 3 to KC little if any electron solvation takes place, since NH 2N’ 3 preferentially solvates the available potassium ions until a critical 43 concentration X0 = u.o is exceeded. It can be assumed to a first approximation and for simplicity that the charge exchange, f, varies linearly with the excess NH above X - Xc. Thus, 3 1forX 1-Y(X-X)forX>X C C where Xc - 1.0 and the constant Y has yet to be determined. By incorporating Eqs. 15 and 16 into Eq. 12 and minimizing the sandwich energy E with respect to the sandwich thickness (18 for a given 1 x and x we find: K A k d do k d do 0 0 K A A K d [1 x +x1(1-—)]+[—— Y(—-1)]X s1 - c kaK dK+ c dK+ kK dK+ dK+ 1 + dK+ k kA 1+ 3 +-—x1 "K K K 2 n xK[1 - Y(X1 - Xc)] k k x:+%x1) KK K (17) kK(1 + When pristine graphite is intercalated with potassium and ammonia, the resultant carbon interlayer distance increases by about a factor of two. As a result of this large expansion the carbon force constant ko appropriate to Eq. 17 is not equal to that of pure graphite. In order to properly determine kc and k in Eq. 17, a micrOSCOpic theory with k appropriately chosen interaction potentials would have to be constructed and the force constants evaluated. However, a crude approximation can be made for k0 by describing the carbon-carbon interlayer interaction with a Lennard-Jones potential, the minimum of which is at r0 = 3.35A, 44 the interlayer distance in pristine graphite, and calculating the force constant (curvature) at r - 6.63A, the interlayer distance N1 K(NH ) Using this approach, we estimate that ko for 3 u.33czu° K(NH3) is a factor of ~200 smaller than that of pristine u.33czu graphite. Similarly, we find that the value of kk appropriate to K(NH ) decreases only about a factor of 2 from that for KC 3 u.33czu 2a which is not surprising since the sandwich thickness of K(NH3)u 33Czu'is only about 20% larger than d +. The above method for estimating force K constants is of course over-simplified and ignores effects such as charge exchange, which would tend to decrease even further the interlayer interaction. Thus, it is clear that to a good approximation ko/kaK << 1, and that the value of kK may be approximately determined from those measured for the potassium GIC's. It should also be mentioned that the results of Hawrylak and 3 Subbaswamy for KCx are also consistent with ko/k << 1. In addition, xxx they found that the electrostatic contribution to (18 (second term in Eq. 17) is quite small and we also find it to be negligible. With a . 0 and ko/kaK << 1 equation 17 can be rewritten. d o d o k d K K A A d [1-Yxt —1)]+[1( -1)+—- 1x 51 _ ° dx+ dk+ kx dK+ 1 (18) d + k K 1 + kA X1 K where X0 = ”.0. The force constant ratio kA/kK of the K-NH3 GIC in Eq. 18, can be estimated from measurements of bulk potassium and bulk ammonia. The bulk compressibilities of the alkali metals K, Rb, and Cs, and of NH3 are at least an order of magnitude higher than that of pristine 45 graphite. Therefore, a c-axis compression of the GIC sandwich results primarily in a deformation of the intercalant in K-NH3 GIC's. The effective c-axis force constant of the sandwich can thus be approximated by the interatomic force constants determined from measurements of the bulk properties of the intercalants. For example, the c-axis force 39 constant ratio deternuned from neutron scattering experiments on KC8 and RbC8 is, within experimental error, equal to the value kK/ka - 1.A6 obtained from ultrasonic measurements of the room temperature elastic constants of bulk potassium and rubidium metal.“O From the compressibility of liquid potassium“ (CK) and liquid 15 NH 3 (CA/CK) (CA) we find an approximate force constant ratio (kA/kK) 1 - 0.56. In addition, from the longitudinal sound velocities 15 3 determined by ultrasonic v1 for liquid potassium”1 and for NH N2 measurements and the relation - v’ki/p1 where p1 is the density of vi the liquid species 1, the force constant ratio is found to be kA/kK = .77. An independent estimate of the force constant ratios k /k can be A K 39 By fitting the made from an expression given by Dresselhaus et a1. phonon dispersion curves of several GIC's, they found an empirical relation between the force constant of an intercalantanuithe corresponding equilibrium sandwich thickness. Their result can be rewritten to determine the force constant ratio kA/kK as a function of dA (defined above) and the inplane intercalant density, S. Thus, k PA. 17.11 e5.58 (19) K 46 39 Equation 19 was derived using a value of k - 2.3 (normalized to K k of pristine graphite). To our knowledge NH 0 cannot be solely 3 intercalated into graphite, therefore, we must estimate the values of (1A and S to be used in Eq. 19. 19 ) We know from recent NMR measurements of stage-1 KC2u(NH 3 “-33 that the NH3 molecule is rapidly rotating (on NMR time scales) about its 0 axis which simultaneously precesses about the graphite c-axis (see 3 Fig. 12). The motion of the NH3 molecule is nevertheless constrained (i.e. it does not tumble freely) by the charge-dipole forces which bind it to the potassium ion and by steric forces imposed by the bounding graphite planes. In a ficticious binary "NH3-GIC" the NH3 molecules could freely rotate in the graphite galleries. Free rotations of the NH molecules can be insured if the molecule occupies a volume 3 determined by inscribing its surface of revolution about the 03 axis (see Fig. 12) in a sphere of radius rNH - 2.1“. If such a sphere fit 3 precisely into the double hexagonal cavity of the bounding graphite planes, the resultant value of d would be (1 - 7.08A. With that value A A of dA and S - 5.N6 established from the inplane area “rNH 2 we find from 3 Eq. 19 that kA/kK - 0.65. This value of kA/kK - 0.65 is in excellent agreement with those obtained from measurements of the GIC compressibilities (kA/ - 0.56) and of the sound velocities of the “K intercalants K and NH3 in bulk form (kA/kK - 0.77). and constitutes the average of those measurements. We will, therefore, assume for the remainder of this paper that kA/kK . 0.65. 47 G. Comparison with Experiment We have fit d1 vs. PNH3 in Fig. 11 (dS1 - d Fowler-Guggenheim form of x1 vs. PNH (see Eq. 11), kA/kK - 0.65, and 3 Eq. 18, with only two adjustable parameters, which are the bracketed 1 for stage 1) using the terms in the numerator of Eq. 18. The results of that fit are shown as a solid line in Fig. A and constitute excellent agreement between theory and experiment. The apparent cusp in the fit of <13 vs. PNH at PNH < 1 3 3 1 atm is expected, since we have assumed a charge exchange f that inherently has a discontinuity in slope at x - “.0. This cusp is unphysical and should be smoothed (see dashed line in Fig. A). From the fit of Eq. 18 to our data we find directly Y - .143 1 .013 and d - 7.05 A t .OAA. The value of dA - 7.05A is surprisingly close to our estimate of dA - 7.08A which indicates a self-consistency in our approach. From Y - 0.1A3 we find for stage-1 K(NH3) the charge exchange f - 0.95 1.33021 which, given the uncertainty in determining f by any method, is in reasonable agreement with values obtained from measurements of 20 19 reflectivity, f - 0.73 and NMR, f - 0.86. It should be mentioned that different values of kA/kK were tried and the resultant value of f was insensitive to the choice of k /k i.e. f - 0.95 for A K' 0.3 < kA/k < 0.9, but a least squares fit with f - 1.0 could not be K obtained. The agreement between the charge exchange f obtained from NMR and reflectivity measurements, and that extracted from the fit of dS vs. 1 PNH , can be improved by considering other mechanisms for electron 3 solvation such as NH3 shared ion-electron pairs. That is, it is possible for the NH3 molecules of different ll-fold K-NH3 clusters to 48 orient themselves to form a cavity in which they can bind an electron. This would imply that for X < Xc (where presently Xc - 11.0) the charge exchange f < 1. It is probable that f varies nonlinearly with the number of A-fold K-NH3 clusters. This nonlinearity could be treated by adding additional terms and parameters to Eq. 16 or equivalently to Eq. 18. This would marginally improve the fit to the data of Fig. 11. However, such a procedure is not warranted because Fig. 18 contains the essential physics of the charge exchange process and the data of Fig. A does not vary rapidly enough with PNH to allow additional parameters to 3 be accurately determined. For the purposes of illustration, let us assume that the current description of the charge exchange f given in Eq. 16 is applicable but because of the existence of other electron solvation mechanisms, the constraint Xc - 11.0 must be relaxed to permit smaller values of Xe. From the current fit of ds1 vs. PNH3 in Fig. 11, we find YXC - .57, and for a charge exchange f - .80 we estimate that Xc - 3.2. This indicates that the inclusion of other solvation mechanisms tends to lower X0 and brings the extracted charge exchange f into better agreement with values determined in other measurements. From the parameter (dA/dK+)(kA/kK) + Y(dKo/dK+ - 1) - .908, and Y - .1113, obtained from the fit to d1 vs. PNH , we estimate that the 3 relative elastic contribution of the charge transfer term Y(dKo)(dK+ —1) to the sandwich expansion is at least 6%. Furthermore, the fractional ) increase in the sandwich thickness of our stage-1 K(NH3 11 33C” (d - 6.633A) relative to Rudorff's stage-1 K(NH3)2 0012 (d - 6.5A) is 2%, which can be fully accounted for by a charge transfer coupled 49 elastic expansion. As a final point, x is pressure insensitive (at 2 least at low pressure), therefore, the variation of d with P 2 NH3 (see Fig. 5) must depend more on the variation with x1 than with x2. The "keying" of d to <11 is not surprising in view of the high mobility of 2 NH in the galleries of laterally contiguous stage-1 and stage-2 3 regions, as would exist in the Daumas-Herold (DH) model'43 of the layer structure of GIC's which is shown schematically in Fig. 13. In the DH model each gallery contains some intercalant which, however, may show lateral inhomogenities in its distribution. Macroscopic regions of different stage are separated by kink dislocations as shown in Fig. 114. But an NH molecule can dynamically sample contiguous stage-1 and stage- 3 2 regions while still remaining in the same gallery. The expanded stage-1 sandwich (ds ) imposes both a dynamic and 1 static strain on the stage-2 sandwich (ds ) resulting in a small 2 decrease in d as compared with d (d - d = 0.0511). In fact the 82 s1 s1 32 strain between the stage-1 and stage-2 regions, as measured by the difference between <12 and d1, remained constant over the pressure range 0.1 atms < PNH < 6.5 atm, with d2 - d1 - 3.30A. The extrapolated value 3 of d2 - 9.93A (see dashed line in Fig. 5) also indicates that this strain persists even at P - 9.5 atms (of course the extrapolation NH3 assumes that stage 2 still exists at that pressure). It is interesting to note that higher stage mixed phase K-NH3 GIC's also show the same difference between dn and an for stage n. For example, Fig. 10 shows +1 the room temperature x-ray diffraction pattern of KC36 + NH3, which indicates a stage-1 and stage-2 mixture with d =- 6.638A, d - 9.93811, 1 2 50 A —————— Figure 13. Daumas-Herold model“2 of laterally contiguous layers of stage-1 and stage-2 regions present in a mixed stage graphite intercalation compound. 51 and dz-d1 - 3.300A. Figure 1A shows that the room temperature (001) x- ray diffraction pattern of KCNB + NH also indicates a mixed stage 3 compound which in this case is composed mostly of stage-2 material (d2 . 9.98AA) with some stage-3 admixture (d = 13.29A). Again, we find that 3 d3-d2 = 3.306A. It is clear from the above discussion that for mixed phase K-NH 3 GIC's which are composed of 2 stages, n and n+1, the lower stage n imposes a strain on the higher stage n+1, and that this strain is insensitive to variations in PNH . Moreover, the sandwich thickness of 3 the stage n phase is essentially independent of that strain. To support this point, we note that the value of dB for KC 2 identical to the strain-free value of ds for K(NH3) 1 M8+NH3 is 6.634A and is 1.33C21° 52 Amoovm + “NF 005 A: 09m 8? 09m Anoovm 309m 809m + 808m Coovm. Amoovm 803m «voovm $005 80on + 2.8.3 808m 808m 808m . :83 :0on .1_ L 1 A a '1 AthD >m..1_mzm.r2_ a (71—1) Figure 111. The room temperature (PNH - 9.5 atms) (009.) x-ray 3 The notation is diffraction pattern of KCAB + NH3 liquid. the same as that of Fig. 3. IV. Summary_and Concluding Remarks We have found that as PNH is lowered, the stage-1 K(NH3) 3 ternary GIC evolves to a mixed phase stage-1 K(NH3)x Cy and stage-2 1 1 K(NH3)x Cy system, with unexpected variations of the carbon/potassium 2 2 u.33czu ratios yn (for stage n) with P At P - 9.5 atms we find that x - NH ’ NH 1 3 3 n.33 and y1 - 2A and estimate that y2 > 37 and x2 > 3.2. As the ammonia pressure is lowered, both y1 and y2 decrease and at PNH ~ 10-2 atms 3 could be extrapolated to y1 - 12 and y2 - 28, which is consistent with the stoichiometry of compounds prepared by Rudorff,” namely stage-1 K(NH3)2.OC12 and stage-2 K(NH3 2.0028' It was found that the sandwich of stage 1 imposed an elastic strain ) on the sandwich of stage 2 such that ds - ds - 0.05A and that this 1 2 strain was insensitive to variations in P . Both dS and d8 were NH3 1 2 found to depend on the variations of x1 with PNH , which was fit to a 3 Fowler-Guggenheim equation for adsorption. From x1 vs. PNH in Eq. 11 and the independently determined value 3 of k /k - 0.65 we fit the d vs. P data in Fig. 11 using Eq. 18 with A K 1 N113 only two adjustable parameters. The resultant fit constitutes excellent agreement between theory and experiment. Both the sandwich thickness of the NH3-GIC and the charge exchange f could be independently extracted from the fit to d1 vs. PNH . The deduced value of f . 0.95 is in 3 53 54 reasonable agreement with values obtained from other measurements. The deduced charge exchange can be brought into closer agreement with reflectivity16 and NMR15 measurements by incorporating multiple electron solvation mechanisms into Eq. 16. The physical origin of our results may be separated into three major effects: First, as NH3 is added to the interlayer space containing K+ ions, an expansion results from the size difference of NH3 relative to K+ (1.6» d /d + > 1). Second, as NH is ingested, some of A K 3 the charge originally donated to the carbon layers is extracted back into the K-NH3 layer, concurrent with electron cage formation, and is delocalized. Third, much of the delocalized charge resides near the potassium ions and increases their effective radii. Thisznumeased potassium radius generates additional elastic energy causing an increase in dS . 1 The fractional increase of our stage-1 K(NH3)u 3302” (ds - 6.633A) ' 1 from rudorff's stage-1 K(NH (d=6.5A) is 2%, which is very close 3)2.oC12 to our estimate of the elastic contribution of the charge exchange tern: in Eq. 18 and suggests that the difference is entirely due to an elastic-charge transfer coupling. Finally, the experimental results from several measurements1 1’12’13’19 can also be separated into three important effects: First, there is a significant extraction of charge from the graphite layers (f - 1.0 to f .. 0.80) to the intercalant layer due to intercalated NH3, attributable to electron solvation. Second, the K-NH3 intercalant forms a 20 liquid in the graphite gallery with the K ion forming a planar A- fold coordinated cluster with the NH3 molecule and the excess NH3 55 molecules (0.33 Per K(NH3)u cluster at PNH = 9.5 atms) lying in the 3 plane defined by those clusters. Third, the relative potassium/ammonia concentration can be continuously varied from 19 MPM to 100 MPM with indications (see Fig. 8) that higher stages may provide a lower inplane potassium density than that corresponding to y1-211 for stage 1. This would leave additional room for NH3 and yield a lower minimum concentration. These results for the K-NH3 GIC are consistent with the formation of a 20 metal-ammonia solution in the gallery of graphite. Higher stages may provide the means of accessing the dilute region < .1 MPM, in which case the K-NH3 ternary GIC's could be a novel and intrinsically interesting system for studying the 20 nonmetal/metal transition. LIST OF REFERENCES 10. 11. 12. 13. 14. 15. 16. 17. 18. LIST OF REFERENCES S.A. Safran, Phys. Rev. Lett. 33, 927 (1980). S.E. Millman and G. Kirczenow, Phys. Rev. B26, 2310 (1982). P. Hawrylak and K.R. Subbaswamy, Phys. Rev. B (in press). S.A. Safran and D.R. Haman, Phys. Rev. B22, 606 (1980). S.A. Safran and D.R. Haman, Phys. Rev. Lett. 32, 1410 (1979). R. Clarke, N. Wada, and S.A. Solin, Phys. Rev. Lett. 53, 1616 (1980). C.D. Fuerst, J.E. Fischer, J.D. Axe, J.B. Hastings, and D.B. McWhan, Phys. Rev. Lett. 29, 357 (1983). K.D. Woo, W.A. Kamitakahara, D.P. DeVincenzo, D.S. Robinson, H. Mertwoy, J.W. McViker, and J.E. Fisher, Phys. Rev. Lett. _5_g, 182 (1983). A. Metrot and J.E. Fischer, Syn. Metals 3, 201 (1981). D.A. Neuman, H. Zabel, J.J. Rush, and N. Berk, Phys. Rev. Lett. 2;, 56 (198A). A. Herold, Proceedings of the International Conference on the Physics of Intercalation Compounds, edited by L. Pietronero and E. Tosatti (Springer-Verlag, New York, 1981), p. 7. S.K. Hark, B.R. York, and S.A. Solin, Solid State Comm. S.A. Solin, Y.B. Fan, and B.R. York, Bull. Mat. Res. $00., in press. J.L. Dye, Progress in Inorganic Chemistry 22 (198A). J.C. TWuanson, Electrons in Liquid Ammonia (Clarendon Press, Oxford, 1976). J. Jortner and N.R. Kester, editors, Electrons in Fluids: The Nature of Metal-Ammonia Solutions (Springer-Verlag, New York, 1973). J.C. Thompson, Rev. Mod. Phys. 39, 70A (1968). W. Rudorff and E. Schultze, Angew. Chem. 66, 305 (195A). 19. 20. 21. 22. 23. 2”. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 57 H. Resing, B.R. York, and S.A. Solin, to be published. 0. Hoffman, A.M. Rao, G.L. Doll, P.C. Eklund, B.R. York, and S.A. Solin, Bull Mat. Res. 800., in press. A. Herold, Bull. Soc. Chim. Fr. 999 (1955). B.R. Dewald and J.H. Roberts, J. Phys. Chem. 12, A224 (1968). J.W. McBain and A.M. Baker, J. Am. Chem. Soc. flfi, 690 (1926). E.W. Washburn, edittnu International Critical Tables of Numerical Data, Physics, Chemistry, Technology. first edition, Vol. III (McCraw-Hill, New York, 1928). R.C. Weast, editor, Handbook of Chemistry and Physics 55th Ed. (CRC Press, Boca Raton, Florida, 197"). N. Akuzawa, M. Ikeda, T. Ameniya, and Y. Takahashi, Syn. Metals 1, 65 (1983). P. Chow and H. Zabel, Syn. Met. 1, 2M3 (1983). R. Fowler and E. Guggenheim, Statistical Thermodynamics (Cambridge University Press, 19N9). Jackson, Classical Electrodynamics 2nd Ed. (John Wiley and Sons, New York, 197“)- A.H. Narten, J. Chem. Phys. 32, 1692 (1968). N.A.W. Holzwarth, S. Louie, and S. Rabii, Phys. Rev. Lett. El, 1318 J.R. Dahn, D.B. Dahn, and B.R. Haering, Solid State Comm. 32, 179 T. Enoki, M. Sano, H. Inokuchi, J. Chem. Phys. 2g, 4 (1983). D.A. Copeland, N.R. Kestner, and J. Jortner, J. Chem. Phys. 23, 37A1 (1975). N.F. Mott and E.A. Davis, Electronic Processes in Non-Crystalline Materials (Clarendon Press, Oxford, 1971). T.A. Beckman and K.S. Pitzer, J. Phys. Chem. 62, 1527 (1961). J.L. Dye, Pure and Appl. Chem. 32, 3 (1977). W.A. Siddon, J.W. Fletcher, R. Catterall, and F.C. Sopchyphyn, Chem. Phys. Letters 3g, 584 (1977). G. Dresselhaus, R. Al-Jishi, J.D. Axe, C.F.11ajkrzak, I“. Passell, and S.K. Satija, Solid State Comm. fig, 229 (1981). 58 I10. K.H. Hellwege, editor, Landolt-Bornstein Numerical Data and Functional Relationships in Science and Technology, Vol. II (Springer-Verlag, New York, 1979). H1. O.J. Kleppa, J. Chem. Phys. 18, 1331 (1950). H2. C. Kittel, Introduction to Solid State Physics, 2nd Ed. (John Wiley and Sons, Inc., New York, 1957). N3. N. Daumas and A. Herold, Bull. Soc. Chim. Fr. 2, 1598 (1971). APPENDICES VOLUME 50. NUMBER 19 PHYSICAL REVIEW” LETTERS 9 Maw 1083 Ternary Graphite intercalation Compound KCsCM: An ideal Layered Heterostructure B. R. York, 3. K. Hark, and S. A. Solin Department of Physics and Astronomy. Michigan State University, East Lansing. Artichigan 48824 (Received 28 February 1983) An ideal ternary heterostructure graphite intercalation compound, KCsCm, which at- hibits a stage-l c-axis stacking sequence, . . . CKCCsCKCCs. . . . has been prepared. The stage, stacking sequence, and 12 x2)R0° in-plane structure have been confirmed by high- resolution x-ray diffraction studies which reveal a c-axis correlation range of ~ 350 .3. The Structure factor and widths of the 1001) reflections calculated on the basis of the pro- posed stacking sequence are in excellent agreement with experimental observations. PACS numbers: 61.60.+m, 61.10.?1' Heavy-alkali-metal ternary graphite intercala- tion compounds (GIC’s), which were first synthe- sized more than a decade1 ago using powdered graphite, have become the focus of consider- able recent interest"3 in part because they rep- resent potential model systems with which to study the structural properties of binary alloys in two dimensions. The ternary GIC’s prepared to date have always been planar solid solutions (A ,B,.,),_ C," , where n/m is the stage corre- sponding to the number of carbon layers separa- ting nearest layers of intercalate ions, and Us is the areal number density of intercalate atoms relative to carbon atoms.‘ We report in this Letter the first successful preparation of a ternary heterostructure GIC. The general form of such a compound can be stoichiometrically represented by the notation (A_C,,,),(B_.C,.,,.),., where m(m' ), s(s' ), and n(n’) retain the definitions inferred above and ((1’) corresponds to the number of times the cor- responding bracketed unit repeats in the minimum size c-axis stacking unit which defines the struc- ture. The ternary GIC which we have prepared, (KC,,,),(CsC,, ,), or equivalently l(CsC,,H is the simplest such structure and can be represented by the repeat sequence . . . CKCCsCKCCs. . . . It is an ideal heterostructure because the potassium and cesium layers are commensurate with (i.e., epitaxial to) the carbon layer; the carbon, potas- sium, and cesium layers are atomically flat; there is no interdiffusion at the layer interfaces; and the stacked layers exhibit long- range order in both the c-axis and the u-axis directions. To our knowledge, such an ideal manmade hetero- structure has not to date been prepared by use of SOphisticated techniques such as molecular beam epitaxy.’ Samples were prepared in Pyrex with highly ordered pyrolytic graphite (HOPG) (~ 5 x 102’- 0.5 mm“) by a sequential intercalation procedure. 1470 A well-ordered pure stage-2 CsC“ compound“ was first obtained.7 The sample was then rapidly transferred in a glove box (s 0.5 ppm 0,) to an- other Pyrex tube containing pure potassium met- al. The tube was evacuated and sealed off in the usual manner. The tube was uniformly heated to 70‘C and the CsC“ sample was immersed in liquid potassium. Periodically, to facilitate x- ray examination, the sample tube was placed in a centrifuge which was housed in an oven at 70 °C, the excess liquid potassium was Spun off, and the tube was quenched in air to room temperature. The diffraction patterns reported here were re- corded using a 12-ltW Rigaku rotating-anode x- ray source, a molybdenum target, and a Huber Model No. 430-440-512 four-circle diffractome- ter equipped with a vertically bent graphite mon- ochromator. The maximum resolution of this instrument was measured with use of the (400) reflection of single-crystal germanium and found to be 0.003 A“. Figure 1 shows the (001) x-ray diffraction pat- terns of the CsC“ starting material, after im- mersion and quenching for two hours, two days, and 12 days. The small diffuse background notice- able in Fig. 1 is due to the Pyrex envelope. The patterns of Fig. 1 have been indexed with the no- tation (n/m)9(001), where n/m is the stage and I is the order of the c-axis reflection. The symbol 9 denotes the intercalate Species. Thus 9 =K, Cs, or H, where H represents the heterostructure KCsC,,. As can be seen from Fig. 1, the mono- phase KC“ evolved to a three-phase system con- taining Cst KC“, and CsC,; then to another three-phase system containing the KCsC,, heter- ostructure, KC,” and KC“; and finally to a two- phase system of KC, and KCsCu. in Fig. 2 we show a plot of the wave vector, q , versus the order of the reflections shown in Fig. 1(c). For a prOperly identified stacking structure such a plot should yield a straight line passing © 1983 The American Physical SOCiClY REVIEW LETTERS ‘) MAY 1983 VOLUME 50.Nu.~ts£a 19 PH YSIC AL LSF t (Cl “ 12OAYS 1.0+- g . a g ems; as ~=§~ é‘z‘ E xii? ii? i. i t g . cs: - 3 arms tor § E .- g g F: , as 8 g .- é " ’5 a g E i . «E . l 1 4 . 2.0- w E g g 2% w. 9% - 33 3” = g - ti. 5. .i . t" 3“" g “ L ' '. 1 5 20 as 25 30 20 mm FIG. 1. (001) diffraction patterns of stage-2 CsCu after immersion in liquid potassium. The reflections 40 labeled H (001) correspond to the heterostructure KCscn. Those labeled astoon correspond to a stage-a binary GIC intercalated with species 9 . The diffraction pat- terns were acquired with Mo Kc radiation. Reflections which appear in more than one trace have been labeled only once. The ordinates of curves A and B have been truncated for presentation. through the origin and with a slope 2u/d, where d is the minimum distance along the c axis (or sandwich thickness) which defines the stacking sequence.“ From Fig. 2 we find ch. :- 5.35 A. I(q)°=';f,(q)exp[- W, (3%)? expl-i (4.2,) - (4231an N' can (— %)]L,(qi. Here L ,(q) is the combined Lorentz-polarization factor; 2“ is a unit vector; 2, is the distance of the _ith layer from the origin of the c-axis cell; _r’,(a), j =C,Cs,K, are the q-dependent form fac- tors”; and the reciprocal lattice vectors are a, = (2a /d,,)lz‘. The apprOpriate values of W“ the Debye-Waller factor, have been obtained from measurements of KC, and CsC._,.lo The square- bracketed term in Eq. (1) is the Gaussian approx- imation to the Bragg scattering term.‘ The full width at half maximum of each reflection is 1" 2 4(1.4) ’.\'d,,. Here N is the number of cells of length d” that are correlated along the c axis, the correlation range of which is .\'d,,. In Fig. 3 we have plotted [(7), derived from I 70- CC... — CC... ’ m 6:5.35A 6‘0“ "' d":1127a 5.0-- 4.0% °§ 30‘ 2.0-i C— K e Cg o 1.01 l 1 l 1 4 L 1 i 1 l i O 2 4 6 8 0 12 1 FIG. 2. A plot of the wave vector vs the order of the corresponding (0011 reflection for the KC. (dots) and KCsC,, (circled dots) components of Fig. lie). The basal spacings d, and d,‘ indicated were determined from the slapes of the straight (solid) lines which are least-squares fits to the data. Similarly, from our measurements of CsC, we find that dc,c.- 5.94 A. Thus, to within experi- mental error, d, =dxc.+dc.c. as would be expect- ed for the stacking sequence . . . CKCCsCK. . . . To further confirm the heterostructure stack- ing sequence of KCsCm we have calculated the (001 ) x-ray intensity distribution I ((1) vs q from the following equation': (1) I Fig. 1(c), and the theoretical diffraction pattern of Eq. (1) obtained with .V =31 and a l" corrected for the instrument resolution. With the exception of the (001) peak, the experimental and theoreti- cal curves of Fig. 3 are in excellent agreement. The discrepancy at low angle arises from the fact that at grazing incidence [P = 1.3" for the (001) reflection] the effective sample cross section is only a small fraction of the incident beam cross section. For our sample configuration this ge- ometrically induced cross section mismatch is absent for 6 > 9° and has only a marginal effect on reflections which occur in the range 3’s 6 a 9°. The value of N used to compute I (q) was in fact I471 V'OLL'ME 50. NUMBER 1‘) PHYSICAL REVIE\\" LETTERS OMM-wst [(004) ~’ ' . ,1 , I i NRNSHHMENWWVUHS a 8 8 “El a l ,' .7. “ g g ‘2 '5 E )- 8 3 ‘ g g : 1 8 8 8 ° m iwr'f:*I—:"T“"‘“r—'~r—'¢7* 10 10 10 to 50 60 70 FIG. 3. The intensity of the (00!) diffraction pattern of KCsC,‘ vs 4. The solid line is the calculated pattern [see tart, Eq. (1)]. The dots correspond to experimen- tal results and are obtained from Fig. 1(c) after cor- rection for the measured glass envelope background, subtraction of the stage-1 KC. pattern, and normaliza- tion to the theoretical curve at (004). The inset shows a high-resolution scan of the (004) reflection (vertical bars indicate the instrument resolution, 0.0045 A"). deduced from the high-resolution x-ray scan shown in the inset of Fig. 3. From a Gaussian fit to the observed (004) heterostructure reflection we find, taking account of the instrument resolu- tion of 0.0045 A", a c-axis correlation range of ~ 350 A. The correSponding correlation range for the KC, component of our sample was found from the (002) reflection to be “ 260 A. We also found that the width of the (001 ) reflections of KCsC,, were, to within experimental error, in- dependent of q as expected when N is large.’ This q independence of the linewidth indicates that the staging of the heterostructure is “pure" over the ~ 350-A correlation range with little or no statistical admixture of other stacking se- quences (e.g., pristine HOPG or KC,) as was re- cently found in pressure-induced fractional stages of KC..‘ The observation of phase-separated KC, and KCsC,, in the saturated compound which we have prepared is quantitatively consistent with two conditions which appear to govern the intercala- tion process during the initial period (0 to ~ 15 days) of intercalation, and at the immersion temperature of 70 'C: (1) No cesium is expelled from the sample, and (2) potassium does not al- loy with the Cs present in the sample. We have carried out a neutron activation analysis of the residual potassium used for immersion of CsC,, and find no cesium in excess of the impurity amount, a 50 ppm, present in our potassium. 1472 Thus, the neutron activation analysis results support condition (1). Condition (2) is supported by the fact that intralayer Cs in CSC“ does not form a solid solution with potassium, but densi- ties to CsC, during the early stages of exposure [see Fig. 1(a)]. ' Conditions (1) and (2) above, together with our x-ray results, indicate that immersion intercala- tion of CsCu by potassium is initially controlled by the reaction CsCu+ 2K ° KCsC,, + KC” , (2) the right-hand side of which is a saturated non- equilibrium state which appears to be metastable when the sample is cleansed of liquid potassium and quenched to room temperature. If the inter- calation process was allowed to proceed to com- pletion with the sample immersed in liquid po- tassium, the equilibrium product would probably be a ternary solid solution K, Cs, -, C, But there is evidence that such compounds are ultimately unstable“ and further evolve into a phase-sepa- rated system composed of a ternary solid solu- tion, KC,, and/or CsC,’ Note, however, that the ternary solid solution and binary GIC com- ponents cited above would yield (00!) diffraction patterns which are easily distinguished from those of the KCsC,6 heterostructure. Using Eq. (2) we calculate that the ratio of the peak intensity of H(004) to 1K(002) should be 7.2. The corresponding observed ratio [see Fig. 1(c)] is 5.4, in good agreement with theory. In principle an alternative metastable saturated heterostructure of the form (KCJ,(CsC,), or equivalently K,CsC,. with the stacking sequence . . . CKCKCCsCKCKCCs. . . could be generated by the admixture of CsC“ and potassium. Ap- parently such a structure has a higher free en- ergy than the admixture of the higher-symmetry components KCsC,, and KC, The in-plane diffraction patterns of our ternary heterostructure KCsC,, have been carefully meas- ured and confirm that both the potassium and the cesium layers form (2 X 2)R0° (Ref. 6) commen- surate superlattices with respect to the graphite host. We find no evidence for any other in-plane structure. We do, however, find a slight expan- sion of the a-axis lattice parameter of the host material from 2.46 A in the pristine form to 2.47 A when intercalated. While KC, does make a contribution to the (2 x 2)R0° in-plane diffraction pattern, we calculate on the basis of Eq. (2) that the contribution would be down by a factor of 17 for the (100) and (110) VOLUME 50. Numsea l9 PHYSICAL REVIE\V LETTERS ‘) M \\ I‘lh'i reflections compared to the contribution from KCsC,,. These two reflections contain no carbon layer contributions.‘5 Therefore, we are confident that the KCsC,, does indeed exhibit a (2X 2)R0" in-plane superlattice structure. As a final note on this point, Raman scattering studies carried out on KCsC,, (Ref. 12) yield the unique Spectra characteristic of an ordered (2 X 2)RO° super- lattice.” The authors gratefully acknowledge useful dis- cussions with P. Vora, T. Kaplan, and S. Mahan- ti. Thanks are also due to A. W. Moore for pro- viding the HOPG used in this study. This work was supported by the National Science Foundation under Grants No. DMR80-10486 and No. DMR82- 11554. The x-ray equipment was provided by the U. S. Office of Naval Research under Grant No. N00014- 80-C-0610. ’A. Hérold, in Proceedings of the Intemational Con- ference on the Phys ics qr‘Ihtercalatiou Compounds. Trieste, 1981, edited by L. Pietronero and E. Tosatti (Springer-Verlag, New York, 1981), p. 7. 2M. E. Misenheimer, P. Chow, and H. Zabel, to be published. 3a. York, P. Vora. s. K. Hark, and s. A. Solin, Bull. Am. Phys. Soc. gs, 348 (1983). ‘c. n. Fuerst, J. E. Fischer, J. D. Axe, J. B. Hast- ings, and D. B. McWhan, Phys. Rev. Lett. 5_0, 357 (1983). 5M. s. Panish, Science 2_o_s, 916 (1950). is. A. Solin, Adv. Chem. Phys. 4_9_. 455 (1932). 'A. Herold, Bull. Soc. Chim. 53-. £53, 999. 'A. Guinier, .Y-Ray Diffraction (Freeman, San Fran- cisco, 1963). ’International Tables for X—Ray Crystallography (Kynoch, Birmingham, England, 1962), Vol. III. “1.. 2. Campbell, 0. L. Montet, and o. J. Perlow, Phys. Rev. B 1_5, 3313 (1977). “A. Hérold, private communication. ”P. Vora, B. York, and S. A. Solin, unpublished. ”s. A. Solin, Physica (Utrecht) as, 443 (1979). J 147: Structure and Synthesis of the Ternary Alkali Graphite Intercalation Compound KCsC l 6’ An ideal Layered Heterostructure s. K. Hark: B. a. York and s. A. Solin Department of Physics and Astronomy Michigan State University East Lansing, MI 4882!) Abstract We have prepared a stable two We Graphite Intercalation Compound (GIC) containing KC8 and l90 and has only a marginal effect on reflections which occur in the range 3°<_ e <_ 9°. With the observations that 1) dH = cll Cs + d1K and 2) the theoretical and experimental (001) diffraction patterns are in very good agreement, the heterostructure nature of our sample is thus clearly established. The x-ray results cited above indicate that the immersion intercalation of CsCza by liquid K is best described by the chemical reaction CsC 1+ + 2K + KCsC + KC (7) 2 16 8 The right hand side of Eq. 7 represents a metastable state since we expect the equilibrium phase to be a ternary solid solution of K 5Cs 5C8 + KC8 or the single phase substance 15 K.66C5.3a 8' Several facts indicate the Eq. 7 is applicable in the time period during which our C sample synthesis was carried out. First note that Cs is not expelled from the CSCZQ specimen on exposure to liquid K. Instead the entry of potassium causes the Cs to laterally coalesce into regions of CsC8 (see Fig. 1A) from the less dense CsCza. In 16 of the residual potassium in which addition, careful neutron activation analysis studies CsCza was immersed reveal that the level of cesium detected, <50 ppm, was the same as that expected from the original purity of the potassium. Thus the bulk C/Cs ratio is preserved on exposing CsCza to liquid K. On the basis of Eq 3 we should be able to reasonably predict the relative intensity of say the (001;) reflection of the heterostructure phase to the (002) reflection of the KC8 phase. That relative intensity is given by 2 . - F 2 -W.(q/lm) -l(d Z.)-(q Z) f.e ’ e H ’ H L (q ) I P H moon) _ L _72 Iu<(ooz) ' -W( In )2 -'(d E) ( “)2 ‘ ' (8) 'q " ‘ 1K j ‘ quZ Effie ’ e Lp(q1K) 1 In fact the measured value of IH(004)/IIK(002) as deduced from Fig. 1C 15 5.4 in good agreement with the theoretical estimate based on Eq. 7. While it is now obvious that Eq. 7 governs the chemical reaction of CSCZQ with liquid potassium, it is worth mentioning the alternative reaction: CsC 4 + 2K + K CSCZQ (9) 2 2 where the right hand side of Eq.4 also represents a ternary alkali GIC heterostructure but with the stage 1 stacking sequence . . . K C K C Cs C K C K C Cs C . . . Such a structure, while chemically viable, clearly has a higher degree of structural asymmetry than does the KCsCl6 structure. Since chsczu does not form it must also have a higher intercalant- intercalant layer interaction energy and thus a higher free energy of stacking. The in plane (hkO) diffraction pattern of the metastable two phase sample containing KC8 and KCsC16 is shown in Fig. 6 and has been indexed according to the reflections of the (2x2)RO° superlattice. We find no evidence of other in plane reflections. A plot of (h2 + k2 + hk)”z versus the q values of each reflection of Fig. 6 is shown in Fig. 3. The slope of this plot yields the in plane lattice parameter, a, of the (2x2)RO° cell. We find a = 0.94 R in good agreement with the expected value of 2 x 2.06 R17 but a slight in plane expansion of the host graphite layer is evident. This expansion is consistent with those 18 8 8' 8 plane diffraction pattern, we calculate on the basis of Eq 2 and the difference in structure observed for KC and CsC While KC does make a contribution to the (2x2)R0° in factors that its contribution would be reduced by a factor of 17 for the (100) and (110) reflections compared to the corresponding contributions from KCsC“. These two reflections also contain carbon layer contributions. Therefore, we are confident that KCsC does indeed exhibit a (2x2)RO° in plane superlattice. From the width of the (210) 16 reflection, again taking account of the instrumental width, we determine an in plane correlation length of 100 R. The intercalant layers in KCsC16 are clearly commensurate with the graphite layers and exhibit long range order in the plane. In order to determine the stacking sequence of intercalant layers in KCsC“, we have examined (hkt) reflections throughout the reciprocal space along constant levels (h, k, 9. = const.) and constant rows (h = const., k = const., 2 ). Typical examples of constant level (9 £ 0) and constant row scans are shown in Fig. 7 and Fig. 8 respectively. Note from Fig. 7 that reflections associated with k £ 0 appear. These reflections result from the fact that HOPG and its intercalated derivatives yield a two-dimensional powder diffraction pattern for scattering configurations with ELLE It is also evident from Fig. 7 that the (h kl = l) reflections are relatively sharp and indicate the same a-axis correlation range (~iaoR) as deduced from the (hkO) pattern of Fig. 6. In contrast to the sharp reflections of Figs. 6 and 7 the (119.) constant row pattern of Fig. 8 yields reflections which are extremely broad and pseudo shifted from their expected interger positions. This pseudoshift results from the overlap of closely spaced broad bands superimposed on a continuum background. The pattern of Fig. 8 is a classic 19 From the widths of example of that expected for a layered system with stacking faults. the observed reflections we estimate the c axis size of an unfaulted region to be «.35 R. The observed reflections from Figs. 5, 6, 7 and 8 and from a multitude of similar scans are summarized in Fig. 9. The reflections are indexed according to a cell which has a (2x2)RO° in plane superlattice and a c-axis repeat distance of ZdH = 22.54 R. The reflections observed are consistent with what could be expected from a polysynthetic crystal which is divided into domains corresponding to the different possible stacking sequences, AaA aAyAis, and those generated by permutation of QBY6 . Here A refers to the c—layer and crew to the intercalant layers where for example ow = K; 8,6 : Cs. Out of the 6 possible permutations of «,8,y,6 only 3 are distinct21 and energetically these stacking sequences are the same. The square of structural factor S2 (h,k,l.) of such a crystal can be readily calculated. Neglecting the Debye-Waller factors, it is given by n 2 £hd1K l * , a i— I(h,k,2.) + 2Re(Fc(h,k,L)FI(h,k,a)e 2dH )] $2(h,k,l) - F2(h,k,l) +-— E [F c 3 o8y6 where iid /d (10) , _ h-k h - k Fc(h.k.k) YFCCOSC—g-3)cos- ncosi H[1+(-1)£][1+e 1K H] (11) 2 IO and h+k+1 keiin/Z [1+(-1) llfK+(-l) h+kei£H/2 f ], 3876 cs F ' [I+(~1)h+2][fx+(-l) fcs], 3768 (12) [1+(_1)k+£][fx+(-1)heiZH/2 fCS], aéSY are the structure factors of the carbon and intercalant sublattices, respectively. The summation is over the 3 distinct permutations of days. The final expression for $2 is quite elaborate, but clearly predicts that for all (hkO) other than (0,0,1. = 2n+l)(n = integer) there will be diffraction. Furthermore for (h = 2n, k = 2m,2 = 2p) (n,m,p integers) the intensity will strong and relatively weak for (h = 2n+l, k, I = 2p). This is consistent with the observation that for (1,1,!) and (l,0,£) scans, the 2 = 2p diffractions and for (1,0,!) scans, 1 : 2p + l diffractions are, seemingly absent but do appear in various level scans (Fig. 9). In fact, this polysynthetic character is similar to that of KC8 and RbC8 and, as has been pointed out,21 could be a general property of the graphite intercalation compounds. There are other possible stacking sequences, e.g. AolApATAei, which are not energetically very different from AeLApA YAS, especially when only the next and third neighbor inter- actions are considered.13 Thus, it is not surprising that a c-axis stacking fault occurs approximately every one and one-half repeat distances. Nevertheless, the c-axis stacking faults do not disrupt the long-range layer heterostructure nature of KCsC“. Cflcmding Remarks The compound KCsC16 is not a unique example of a ternary GIC heterostructure. Indeed we have already presented preliminary evidence that the analogous compound RszC16 can also be synthesized] Moreover Stump and coworkers have synthesized molecular ternary GIC heterostructures20 but these compounds are unlikely to possess the high degree of in plane order (i.e. epitaxy) characteristic of their alkali counterparts. We ll expect that KCsClG is merely the first synthesized member of a large class of alkali ternary GIC heterostructures with considerably longer c-axis repeat distances and/or fractional stage22 behavior. For instance one can envision the sequential intercalation of stage 4 CsC“ with K to yield a stage 2 like compound KCsC“ with a c-axis stacking sequence...CKCCCsCCKCCCs... Perhaps the most intriguing thing about KCsC16 and the class of potential new compounds which it represents are the novel physical properties and phenomena which they may be expected to exhibit. For example while the intercalant stacking sequences of KC8 and CsC8 are QBYG and QBY respectively, that of KCsC16 is <18de . It would be very interesting to explore the c-axis order disorder stacking transition at elevated temperatures to determine what trives KCsC16 into a KC8 like arrangement as opposed to a CsC8 like arrangement. It would also be of considerable interest to probe the electronic eigenstates of KCsCl6 which is, with respect to the c-axis, a novel one dimensional Kronig-Penny23 like system. Acknowledgements We gratefully acknowledge useful discussions with P. Vora, T. Kaplan and S. Mahanti. Thanks are also due to A. W. Moore for providing the HOPG used in this study. This work was supported by the NSF under grants DMR 80—10086 and DMR 82-11550. The x-ray equipment was provided, in part, by the ONR under grant N 00014-80-C-0610. References l. 0‘ V! o s 10. ll. 12. 13. 14. 15. 16. l7. 18. A. HErold, Proc. of the Intl. Conf. on the physics of intercalation compounds, ed by L. Pietroneso and E. Tosatti (Springer-Verlag, New York 1981) p. 7. L. E. DeLong, V. Yeh and P. C. Ecklund, Solid State Commun. fl, 1105 (1982). P. Vora, B. R. York and S. A. Solin, Syn. Metals, in press. M. E. Misenheimer, P. Chow and H. Zabel, MRS Symposia Proc. Q, ed. by M. S. Dresselhaus et al., North Holland, 1983, p. 283. B. R. York, S. K. Hark and S. A. Solin, Phys. Rev. Lett. 29, 357 (1983). P. Lagrange, M. El Makrini, D. Guerard and A. Herold, Physica 22B, 073 (1980). B. R. York, 5. K. Hark and S. A. Solin, Syn. Metal, in press. M. B. Panish, Science a, 916 (1980). S. A. Solin, Adv. in Chem. Phy. Q, 055 (1982). A. Herold, Bull. Soc. Chim. Fr. 999 (1955). International Tables for X-Ray Crystallography, Kynoch Press (Birmingham, England 1962) Vol III, p. 150. L. E. Campbell, L. G. Montet, and G. J. Perlow, Phys. Rev. §1_5, 3318 (1977). A. Guinier, X-Ray Diffraction (W. H. Freeman, San Francisco, 1963). H. P. Klug and L. E. Alexander, X-Ray diffraction procedures for polycrystalline and amorphous materials (John Wiley 6: Sons, New York 1974) p. 103. B. R. York, 5. K. Hark and S. A. Solin, Bull. Am. Phys. Soc., in press. B. R. York, 5. K. Hark and S. A. Solin, unpublished. R.W.G. Wyckoff, Crystal Structures, Vol. 1 (Oxford University Press, London, 1962) p. 26. D.E. Nixon and G. S. Perry, J. Phys. C 2, 1732 (1969); D. Guerard, C. Zeller and A. Hérold, C. R. Acad. Sci. Paris C E, 037 (1976). I9. 20. 21. 22. 23. J. B. Hastings, W. B. Ellenson and J. E. Fischer, Phys. Rev. Lett., 02, 1552 (1979). E. Stump, Physica B+C 105, 15 (1981). A. Herold, D. Billand, D. Guerard, P. Lagrange and M. El. Makrini, Physica 105B, 253 (1981). C. D. Fuerst, J.- E. Fischer, J. D. Axe, J. B. Hastings and D. B. McWhan, Phys. Rev. Lett. 10, 357 (1983). G. H. Wannier, "Elements of Solid State Theory" (Cambridge University Press, Cambridge, England 1960) p. 136. Figure Captions Fig. 1 (002) diffraction patterns of CsCza immersed in liquid K at 70°C after a) 2 hours, b) 2 days and c) 12 days of immersion and quenching. The peaks belonging to the binary compounds are labeled according to nM(00 2) where n is the stage and M = K or Cs. Those identified as belonging to the heterostructure is labelled H(OO 2). Fig. 2 Plot of the q values versus the order, 2 , of diffraction peaks for stage 1 and 2 K are stage 1 and 2 Cs graphite intercalation compounds. The slope of the straight line gives the d spacing for the corresponding compound. Fig. 3 Plot of the order, 2 , and \jh2 + k2 + hk of the diffraction peaks verses the corresponding q values for the (002) and (hkO) scans of the heterostructure graphite intercalation compound. The slopes of the plots give respectively dH = 11.27 R and a=034R. Fig. 0 High resolution scan (instrumental resolution = 0.0005 R- 1 as indicated) of the (000) peak of the heterostructure. The dotted trace is the experimental data and the solid line is a Gaussian fit using N = 31 (see Eqs. 3 and 4 of text). Fig. 5 Reproduction of the (00 2) scan of the heterostructure after 12 days of immersion in liquid K. The background of the glass envelope and peaks due to KC8 have been removed. The dotted trace is the experimental result. The solid line is a theoretical fit to data using N = 31 and the (000) peak as the intensity normalizing peak. Fig. 6 (hkO) scan of the in plane structure of the heterostructure. The peaks are indexed according to a (2x2)ROo in plane unit cell. Fig. 7 (hkl) scan of the heterostructure. The peaks are indexed according to the (2x2)RO° in plane unit cell with a height c = 2d”. Fig. 8 (1 l 2)'scan of the heterostructure. Fig. 9 Summary of all the heterostructure diffraction peaks observed. (0) represents reflections observed while scanning along a constant level, (h, k, 2 = const), and (0) along a constant row (h = const, k = const, 1). INTENSITY(arb. units) 2.18 . 2.22 2.26 mil“) s a. 8.: L . M1 6888 .. S m 6. 8.: A m 2 .2. - 2 r 2 68.8. 1 :83.“ r 88:: r 1 E888 i 88.: .. . Aogvxfl A§vm0F .3 no.3.» 3 . ahgv: L 1 A085" .83. I; 885 .w 68g.) i. 000 A mxxstN A I L L AflOOvaF l 382%. 8888 883: L i 88.8. . .. .. .o .. ..« «3M. lessoooefinobdoooofkladwfl «Jpn-.5. .1 88.... «mate. 3......3...» .mrn... . e N 8888 s .. 88.: .. .. 38:5 . r 1 :85: 88%.? $8.: $880. A ) .. ) - 2.88.1 i m :8: m w _ _ _ p r - . _ _ . Ll, . 5. o. s o. o. o 1 1 1 2 1 8.8 F x 90893289 IE: 20 25 30 35 4O 29 (DEGREES) 15 10 qu‘t') lNTENSlTY(arb. unitS) O INTENSITY (ARBITRAR .p. a Y UNIT m S) 3 . 2 L l 8 001) ' r P I . (002) T r I 1 (003) (004) (.2) b c=11.27 A a - 4.94 A _ _ _ _ m N a: o... J: ($338930) 93 CI SI 08 98 08 98 07 INTENSITY (ARBITRARY UNITS) 5 K110) ova ”sat. C(IOO) + I(ZOO) g , , . ..C(.‘|.1.0)+l(220) g.» 1(310) Synthetic Metals. 7 (1983) 25 - 31 ID or ALKALI TERNARY HETEROSTRUCTURE GICs: A NEW CLASS OF GRAPHITE INTERCALATION COMPOUNDS B. R. YORK, S. K. HARK and S. A. SOLIN Department of Physics and Astronomy, Michigan State University, East Lansing, MI 48824-1116 (U.S.A.) Summary We have prepared a stable, two-phase compound containing KCsC”, and KCg by exposing CsC” to liquid potassium at 70 °C. X-ray diffraction studies confirm that the KCsC”, is an ideal heterostructure GIC with the stacking sequence ...CKCCsCKCCs. ..,an in-plane structure that is (2 X 2)RO°, a c-axis correlation range of ~350 A and an in-plane correlation range of ~140 A. The basal spacing of the heterostructure c-axis unit is d“ = 11.27 A. Thus, to within experimental error, (1,, = dxc, + dosc, = 5.34 A + 5.94 A. We present evidence that a similar heterostructure compound, RbKCm, with d“ = 11.00 A has also been prepared. These new compounds are physical realizations of an ideal heterostructure because their layers are atomically flat, epitaxial, exhibit long-range correlations in both the c-axis and a-axis directions, and exhibit no interdiffusion at the layer interfaces. 1 . Introduction During the past decade an enormous effort in manpower, and enormous sums of money have been expended to fabricate ideal man-made layered heterostructures using sophisticated molecular beam epitaxy (MBE) tech- niques [1]. While much progress has been made, an MBE derived ideal heterostructure has not yet, to the best of our knowledge, been synthesized. For our purposes, an “ideal” layered heterostructure is one in which the adjacent layers at the layer-layer interface are epitaxial, exhibit no inter- diffusion, are flat on an atomic scale, and also exhibit long-range correlations in both the a-axis and c-axis directions. In a recent letter [2] we reported the successful synthesis of an ideal layered heterostructure [3], the ternary graphite intercalation compound KCsC,g, which possesses a stacking sequence ...CKCCsCKCCs..., and which is one member of a new class of ternary GICs which can be stoichiometrically represented by the notation (Am C,x,,),(Bm’C,-x,,'), [2, 3]. Here A and B are different alkali metals, C is carbon, ((1’) represents the number of times a bracketed unit repeats in the minimum size c-axis stacking array which 0379-6779/83/83.00 © Elsevier Sequoia/Printed in The Netherlands 26 defines the stacking sequence, n/m(n'/m') is the stage designation within the bracketed unit and l,’s(1.‘s') represents the areal number density of intercalate atoms relative to carbon atoms. In this paper we shall present (a) the first detailed data which confirm that the in-plane structure of KCsC", is such that each metal layer forms a (2 X 2)R0’ superlattice [4] with respect to the carbon layers, and (b) show preliminary evidence ofthe synthesis of another ternary GIC heterostructure, KRI)C'6. 2. Experimental Samples of KCsC“, were prepared in Pyrex with highly oriented pyrol- ytic graphite (HOPG) (zSX 10X 0.5 mm3) by a sequential intercalation technique [2]. A well-ordered, pure [4] stage-2 binary GIC, CsC2, was rapidly transferred in a glove box (<0.5 ppm 02) to a Pyrex tube containing pure potassium (< 50 ppm impurities). The tube was evacuated, sealed off in the usual manner, and heated to 70 °C at which temperature it was inverted so as to immerse the CsC24 in liquid K. Periodically, the sample tube was placed in a centrifuge maintained at 70 °C and the liquid potassium was spun off. After quenching to room temperature in air, high resolution (0.003 A") X-ray diffraction patterns were acquired using apparatus which is described elsewhere [2]. The KRbC”, samples were prepared from RbC24 in a similar manner. Note, however, that the K-Cs heterostructure developed more rapidly (z 12 days) than did the K-Rb heterostructure. 3. Results and discussion The (001) X-ray diffraction pattern of the CsC“ sample taken after 12 days of immersion in liquid potassium is shown in Fig. 1. As can be seen from that Figure, two sets of sharp reflections occur, one (labelled H) which corresponds to the heterostructure KCsC", and one (labelled 1K) which .- i7. . H004)_ INTENSITY IARBITRA‘RY UNITS) '0 I a 6 8 T s- .. -= 8 - - 52 ~ 6 = g gas 1 #5 §§ 8 §§ ' " I! x 1 SE 8 8 I x ‘ 1" " r x o ..,.-:- s '. ‘--I 1‘ _.- ,_.,,_ .. - ' ,..- .. . a l 1 o s 10 15 20 25 ab :35 4b 29 (DEGREES) Fig. 1: (001) Diffraction pattern of stage-2 CsC-34 after 12 days of immersion in liquid potassxum. The reflections labelled H correspond to the heterostructure KCsC.g. Those labelled 11“ COUC‘SPOnd to stage-1 KCg. The diffraction patterns in the Figure and sub- sequent Figures were acquired with Mo Ka radiation. 27 corresponds to stage 1 KC,,. The intensities of the two sets of reflections shown in Fig. 1 are quantitatively consistent with the chemical reaction CSC24 + 21“ _. KCSClb + KC8 (1) which governs the initial period (t S 20 days) of the immersion intercalation reaction. From a plot of the wave vector versus the order. I. of the H reflections. (solid circles of Fig. 2), the basal spacing, d" = 11.27 A, can be deduced. Within experimental error this basal spacing is equal to the sum of the basal spacings of KCg and Cng, 5.35 A [5] and 5.94 A [6], respectively. Thus. the heterostructural nature of the sample is, in part, confirmed. In further support of that heterostructure form, we have calculated the structure factor for the H reflections assuming the ...CKCCsCKCCs... stacking arrangement. This calculation is described elsewhere [2] and yields results which are in excellent agreement with experiment. We find from the above cited calcula- tion a c-axis stacking sequence correlation range of ~350 A which evidences a structure with long—range c-axis stacking order. Careful examination of (hkl) reflections [7] throughout reciprocal space along (hh,I-constant) and (h = constant, k = constant, I) scans indicates that the heterostructure adopts an 04376 stacking sequence where or, 7 = K; {3, 5 = Cs, similar to that of KCg but with a high probability of stacking faults (i.e., 0113613, 015137, etc.) in the intercalate layer. V —N a-4.94A " a ro— -9 N _ x o 0 + c-11.27A .1 z -1 arm" s-l N; #00 F -‘V —N 1 l 1 l 1 1 1 o 0 2 4 6 8 6-1 q (A) Fig. 2. A plot of the order, I, vs. wave vector, q, of the (001) heterostructure diffraction pattern of KCsC“, (0, right hand ordinate) and of a function of h and I: (appropriate to a hexagonal (2 x 2)R0° in-plane structure) vs. q (I, left hand ordinate). Each of the solid lines shown represents a linear least-squares fit to the data. ’13 8 : a z A V o - D 9 " >- v 3 A c __ O .- _ v - A 0 V. ,. o 1 c e. 8 ° )- ‘ "3. . U - '- 2: s "‘o".- "0&3 l‘ ' 8 g m w%?/"" T', ° 8 - Z i "‘19.! ,. ' w .W T¥H' ‘V ' .- 5 «Ajmw" k . . A . T ,.m ;«mmsa1; ‘n'. '2' ' “as! r“ We 1 L 1 l 1 1 l 1 5 10 15 2o 25 3° 35 4° 29 (DEGREES) Fig. 3. In-plane (hkO) diffraction pattern of the KCsC“, heterostructure compound indexed with reference to the (2 x 2)R0° superlattice. Reflections associated with the graphite layer are labelled with a ”C" in the Figure. The diffuse scattering background is due to the glass envelope. The in-plane (hkO) diffraction pattern of KCsC”, is shown in Fig. 3 and has been indexed according to the reflections of the (2 x 2)R0° super- lattice [4]. A plot of the q value of each reflection vs. (122 +122 + hie)" 2 is shown in Fig. 2 (solid squares). The SIOpe of this plot should yield the in- plane lattice parameter (a) of the (2 X 2)R0° cell. We find a = 4.94 A, in excellent agreement with the expected value 2X 2.47 A [8], but never- theless providing evidence of a slight in-plane expansion of the host graphite A ('3 O 0 §: "5 N _E, .' > K < K t g A N S 8 v é A i: "a 8 r7: 8' 3.0 a 8 3 v . . 9. x g v 8 x , -'1- g " 18 x N a 'e V V r. .ii- ~ =. «if as i“ I 1 1 1 1 1 l 1 1 5 10 15 2O 25 3O 35 4O 29 (DEGREES) Fig. 4. (001) Diffraction pattern of stage-2 RbCN after 4 h immersion in liquid potassium. Reflections labelled H are most likely due to a KRbCIg heterostructure, while those labelled 1K and 2K arise from stage-l KCg and stage-2 K034. 29 layers. From the width of the (210) reflection (corrected for instrumental broadening) we determine an in-plane correlation length of 140 A. Thus, the intercalate layers in KCsC", are clearly epitaxial to the graphite layers and also exhibit long-range in-plane order. As noted in Section 2 above, we have attempted to synthesize KRme by the same preparation methods as were used for KCst Figure 4 shows the (001) diffraction pattern of RbC24 after immersion in liquid potassium for 4 h at 70 °C. The pattern contains three sets of reflections, labelled l-l, 1K, and 2K, and representing a heterostructure, stage 1 KCB, and stage 2 KC24. By indexing the H reflections in the manner discussed for Fig. 2, we would deduce a heterostructure basal spacing of d“ = 11.02 A, which is in excellent agreement with the sum of the basal spacings for KCg and Rng, which are 5.35 A and 5.65 A, respectively. Only even orders of the (001) H pattern in Fig. 4 are detected, however. These reflections could also be indexed on a c-axis basal spacing of 11.02 A/2 = 5.51 A, a value which would A V O 8 A I (‘0 o . 8 X N A (D 1: :-- g Q a KCsC”s > C O ( 9 t: . x 1— . " a . . E ”Wow-9’3““, ..~‘o‘"y-I.‘ M sun-m V s > A e . a U) 5 . 3‘2 ’- N Z ‘3 8 a 55 8 KRwa X p . I: -.,;" yaoI-eaa—r -..‘..p..u' A”. ' \m..~\.-- L l l l 12 13 14 15 16 17 20 (deg) Fig. 5. Expanded (001) diffraction patterns of KCsC”, and what is likely KRbC”, in the region of the heterostructure (004) reflection. See captions of Figs. 1 and 4 for a descrip- tion of the labels used in this Figure. 30 infer that the compound examined was a solid solution [9]. l\'_,Rhl “C, with x = 0.5. Several facts argue against the solid solution interpretation: these are discussed below. (i) Zabel and coworkers [10] have studied the K,,Rb1_.,,C,B system [11] and have proposed a method by which x can be determined from the relative intensities of the (001) and (002) reflections. These would correspond to the (002) and (004) reflections of the heterostructure (double sized) cell. The ratio 1,1(004,/IH(002, which we observe, however, is ~3.5/‘0.7 = 5, whereas for KO‘SRbMCB, label and coworkers [10] report [002/1001 “s- 2. Even given the uncertainties attendant on measuring the intensities of low angle reflections from (3le [2], the discrepancy of a factor of 2.5 indicated above is well outside the bounds of experimental error and is evidence against the solid solution interpretation of the patterns of Fig. 4. (ii) For the KRbC,6 heterostructure the odd order peaks (001 = 2n + 1) should be much weaker relative to, say, (004) than is the corresponding case for KCsC”, because the atomic scattering factor for Rb is much closer to that of K than is the scattering factor for Cs. In the limit in which Rb assumed the scattering factor of K, (i.e., the sample became KCB) the odd order peaks would have identically zero intensity when indexed on the heterostructure double cell. (iii) The K-Cs and K-Rb ternary immersion intercalation systems behave similarly during the intercalation process. This can be seen from Fig. 5 in which the diffraction patterns of the two systems are compared in the region of the H(004) reflection. The slight difference in the position of the 2K(003) reflections for the two compounds corresponds to slightly different KC“ basal spacings, 8.70 A and 8.76 A for the upper and lower traces, respectively. 4. Conclusions We have confirmed unambiguously that one can prepare an ideal layered heterostructure, KCsC”, by immersion intercalation of CsC24 in liquid potassium. We have also presented evidence that a similar heterostruc- ture, KRbClg,C3n be prepared using the same synthesis techniques. However, there is the possibility that the rubidium-based ternary GIC is a solid solu- tion K0.5Rb0_5C8 and further measurements will be necessary to definitely assess the structural character of the Rb-K-C immersion intercalation system. Acknowledgments We are grateful to R. D. Spence, Parul Vora, T. Kaplan, and S. Mahanti for useful discussions. Thanks are due to A. W. Moore for providing the HOPG used in this study. The research reported here was supported by the 31 National Science Foundation under grants D.\lR80-10486 and 031118211554. The X-ray equipment used in this research was provuled hy the Office of Naval Research under grant N00014-80-C-0610. References 1 M. B. Panish,Science, 208 (1980) 916. 2 See 8. R. York, S. K. Hark and S. A. Solin, Phys. Rev. Lett., 50 (1983) 1470 and references therein. A more complex hydrogen-potassium ternary heterostructure has been synthesized previously. See P. Lagrange and A. Hérold, C. R. Acad. Sci., Ser. C, 278 (1974) 701. S. A. Solin, Adv. Chem. Phys, 49 (1982) 455. D. F. Nixon and G. S. Parry,J. Phys. D, I (1968) 291. G. S. Parry, Mater. Sci. Eng, 31 (1977) 99 and references therein. S. K. Hark, B. R. York and S. A. Solin, to be published. R. W. G. Wyckoff, Crystal Structures. Vol. 1, Oxford Univ. Press, London, 1962, p. 26. 9 A. Herold, in L. Pietronero and E. Tosatti (eds.), Proc. Int. Conf. on the Physics of Intercalation Compounds, Trieste, 1981, Springer, New York, 1981, p. 7, and refer- ences therein. 10 M. E. Misenheimer, P. Chow and H. Zabel, in press. 11 These ternary solid solutions were first synthesized by D. Billaud and A. Herold, Bull. Soc. Chim. Fr., (1971) 103. OD $405013— Synthetic Metals, 7 (1983) 257 - 262 [Q on K) X-RAY DIFFRAC’I‘ION STUDIES OF POTASSIUM—AMMONIA ALKALI-MOLECULAR TERNARY GRAPHITE INTERCALATION COMPOUNDS S. K. HARK, B. R. YORK and S. A. SOLIN Department of Physics and Astronomy, Michigan State University, East Lansing, MI 48824-1116 (USA) Summary We report high resolution X-ray diffraction and gravimetric studies of potassium-ammonia ternary GICs. Both the weight uptake and the sandwich thickness of the intercalated layers vary with the vapor pressure of NH3 to which the binary potassium—graphite compounds are exposed. A model is presented to account for these observations and to relate them to the degree of change exchange exhibited by the potassium ions. Introduction Research on graphite intercalation compounds (GICs) has been mainly concentrated on the binary systems formed from a single intercalated species and the graphite host. Much less attention has been devoted to the. ternary GICs which contain two distinct intercalant species A and B and which can be represented as AxBx'Cn. The intercalants can be a combination of two donor species, a donor and acceptor, or two acceptor species. Hérold [1] has distinguished between the “solution” temaries in which the ratio x/(x +x') can vary between wide limits without significant structural changes, and the “compound” temaries in which mechanical, chemical or electronic con- straints impose more or less narrow limits on the ratio x/(x + x'). Recently [2, 3] a new type of ternary, KCst, which does not belong to either one of Hérold’s classifications has been discovered. In this compound alternate layers of K and Cs were intercalated into HOPG to form a heterostructure. We have prepared and studied a system of “compound” temaries in which K and NH3 constitute the intercalant species. Their preparation and X- ray diffraction studies will be presented in this report. In an accompanying report [4] Raman scattering studies will be discussed. Stage 1 and Stage 2 alkali metal—ammonia ternary GIC systems which were prepared from the action of metal-ammonia solutions on graphite powder have been studied as early as 1954 by Riidorff and coworkers [5]. They prepared a stage I compound with the composition K(NH3)2JC,;5 and a stage-2 compound K(NH3)2_8C2&7. The respective c-axis repeat distances 1,. 0379-6779/83]S3.00 ©Elsevier Sequoia/Printed in The Netherlands 258 were determined to be 6.5 A and 9.9 A. We found that the composition and value of [c are not unique, but for the K-NH3 temaries can vary over a wide range depending on the conditions under which the samples were prepared. Materials preparation and experimental methods The ternary GICs were prepared from a successive intercalation process using highly oriented pyrolytic graphite host material. Samples of different stages of K-GICs were first prepared from HOPG using the standard two-bulb technique [6] and the stages were confirmed by (001) X-ray diffraction scans. The well staged K-GICs were then transferred, in a glove box, into another glass tube which was subsequently evacuated to < 10‘5 Torr. Anhydrous NH3 which was precleaned with sodium [7] was distilled into the sample tube which was then sealed off. Sufficient NH3 was transferred to ensure that the K-GICs were exposed to a saturated vapour throughout an NH3 tem- perature range 77 K < T< 300 K. Ammonia intercalated rapidly at room temperature, as indicated by the expansion of the sample thickness. For all our samples, there was always some initial liquid condensation on the sample surface. There were indications that the liquid might play a role in assisting the intercalation process. T emaries prepared from stage-2 and higher stage K-GICs were unexfoliated, their mosaic spreads are comparable with those for typical binary K-GICs. In the case of NH3 reacting with KCB, the sample A N (a) O O v A '- O 8 A F (D A A y— ; V .. (’3 z .. o o A A A o in to K D 8 :5 o o o > " " 8 8 3 - P u. ' s - .. c —____)' :n A A A I: a E 8 (b) < 9 > t a (O 5 *- 8 g o O v I A N A 8 "’ " .0 A 00 0E0 .08 " 0 v o 8 0’ 01v v V v! 0 V v o O N .- LN N N ; " N N )——"—_A‘&_J\__A A A 'I A 4' l 1 1 s 0 1 2 3 4 5 6 7 ~-i q (A ) Fig. 1. (001) X-ray diffraction of ternary compounds formed from the reactions of KCN and NH3 at a vapour pressure (a) 10 atm and (b) lowered to 0.1 Torr. Reflections of stage n (n = 1,2) K-NH] ternary GIC were indexed as n(001). Reflections of stage m(m = 1, 2) binary K-GIC were indexed as mK(001). The pattern shown in this and the following Figure were acquired using Mo Ka radiation. 259 had a tendency to exfoliate and expel potassium which dissolved in the liquid NH3 to produce a light blue K—NH3 solution. The stages of the temaries prepared were determined from (001) .\'-ray diffraction using the Mo Kaline. In-plane (hkO) scans were used to determine the in plane order of the K-NH3 ternary. Gravimetric measurements reported here were made using a two step procedure [8]. The graphite specimen (typical mass 60 mg) was weighed before and after intercalation of potassium (accuracy 50 pg). It was then loaded in a glove box into an evacuable cylinder fitted with a McBain bal- ance. The McBain balance was used to measure the mass uptake of NH3. Results and discussion When NH3 at a vapor pressure of ~ 10 atm corresponding to room tem- perature was allowed to react with KC“, a pure stage 1 ternary compound, K(NH3), 024, resulted, as determined by the X-ray (001) scan, Fig. 1(a). In the cases of KC“ and K043 reacting with NH3, a phase separated mixture of stages 1 and 2 (Fig. 2(a)) and stages 2 and 3 (Fig. 2(h)), respectively, were obtained. The instrument limited widths of the diffraction peaks indicate that the temaries are pure stage [9] compounds which exhibit long range E (a) o 0 V O A A p, v o 0 A 9 3 8 N '- 1 9 ~ 8 s - s . A .‘i’ 3A2 383 8 :—~; 3 - '0‘ 00 N O‘no O z o 00 00 o D o Evv - v08 V :1 -m “:0‘ o, ). mp—u—Jh——-_— ‘_—A L < E 8 (b) g o < S v (V O f- : CS 0 2'5 2 g o t‘.‘ n 9 z N ¢ ~ '5 8 o o 9 y, n O A w 8 5: '- Ccae") 33:53: ’5:— "' 8 00208 00000 000 o v eeze~ 99999 oOooe retina-'7 manner ESE; 3 A4 ———-—-—-1 1 t 2 3 4 5 5 7 ‘-1 (HA) Fig. 2. (001) X-ray diffraction of ternary compounds formed from the reactions of NH3 (vapour pressure, 10 atm) and (a) KC3o and (h) KC“. Reflections of stage n(n = l, 2, 3) K—NHJ ternary GIC were indexed as n(OOl). 2260 order (> 200 .\) in the c-axis direction. The c-axis repeat distances, It, of the stage-1 K—Xll3 ternary GICs, and of the stage 2 and 3 compounds prepared in a similar manner by reacting NH3 at 10 atm with KC”, KC36 and KC“, were determined to be 6.63, 9.94, and 13.30 f\, respectively. Thus, the inter- calation of NH3 into the KCW, compounds involves a decrease in stage. Similar decreasing stage changes have been observed by Beguin and coworkers [10] and have recently been confirmed by us [11] in the case of furan, and tctrahydrofuran interacting with KCnn. Such stage changes at room temper- ature are easily interpreted using the Daumas-Herold model [12]. Our stage-1 temaries gave 1,. = 6.63 A rather than the 6.56 A quoted by Riidorf (for a sample with the composition K(NH3)2,,C,2_5). Furthermore, stage 1 ternary samples prepared in our laboratory from the reaction of KCs and NH3, although exfoliated, gave Ic = 6.54 A, which agrees with the value reported by Riidorf who pumped off NH3 from his samples after intercala- tion. Using a method suggested by Setton [13], the value ofx in the formula K(NH3), C). can be estimated from the intercalation volume available for NH3 molecules if y and the volume of the K ion are known. For a sample with a measured value of y = 23.86 we calculate that for the saturated stage 1 compound (NH3 vapor pressure of 2: 10 atm surrounding the sample) the weight uptake of NH3 should yield x in the range 4.48 < X < 5.64 where the lower and upper limits correspond to a potassium radius of 2.29 A (atomic radius deduced from K metal) and 1.33 A (ionic radius deduced from potas- sium halides), respectively. The measured value of x deduced from a McBain balance was 4.45, corresponding to a stage-1 sample with composition K(NH3)4,45C23_36. As the temperature of the excess liquid NH3 in the sample tube is re- duced, thereby reducing the NH3 vapor pressure, so also is the measured value of 2:. When the excess liquid NH3 was frozen at liquid nitrogen temperature, the pure stage 1 ternary NH3-K GIC phase separated into a three phase system containing pure stage 2 NH3—K GIC and small amounts of pure KC8 and K024 as indicated by the (001) X-ray diffraction pattern of Fig. 1(b). The stage-2 ternary component has a c-axis repeat distance of [c = 9.65 A, which not surprisingly is different from the corresponding value reported by Rudorff [5]. If we again apply Setton’s procedure [13] to calculate the expected value of x for the stage-2 NH3-K GIC (NH3 vapor pressure < 0.1 Torr for excess ammonia at 77 K), we find, using the same extreme values of the K radius noted above, that 1.23 < X < 2.39, while the measured value of x is 1.66 yielding now a stage-2 sample of composition K(NH3)L66C23,86. Concurrent with the changes in weight uptake with NH3 vapor pressure we have observed that the value of 1,. for stages 1 and 2 K—NH3 GICs can be continuously tuned by adjusting the NH3 vapor pressure over the sample [14]. Thus the differences in the value of 1,. between our samples and those of R'Lidorff relate to the fact that they have different compositions. Presum- ably, NH3 (which is known to solvate K), when intercalated into KCW, com- petes with the graphite layer to receive the donated charge. The amount of intercalated NH3 thus affects the degree of charge transfer to the graphite 261 layer. This, in‘tum, changes the electrostatic interaction between layers and causes the system to adjust its equilibrium value of 1c. For the stage-1 com- pound K(NH3)4,46C23.86, the K is only weakly ionized (i.e., has a large radius) and the sandwich thickness of the intercalate layer (which is equal to [c for the stage-1 compound) is considerably larger than the sandwich thickness of the stage 2 K(NH3),_66C23.36 compound in which the K is more strongly ion- ized (i.e., has a smaller radius). We have also examined the in-plane structure of the stage-1 ternary NH3-K GIC using X-ray (hkO) scans. Only diffraction peaks corresponding to glass and to the graphite in-plane lattice were found. Both the intercalated K and NH3 are in a disordered state at the saturated vapour pressure of NH3 at room temperature. The disorder is consistent with the fact that the com- pound is nonstoichiometric and, in the case of the stage 1 ternary there is only one K ion for every 24 C atoms in the layer. If the compound formed an in-plane superlattice at all, at a certain NH3 vapour pressure and sample temperature, that superlattice would be considerably larger than the (2 X 2) R0° structure of the binary KCS [15]. Conclusion We have shown that NH3—K ternary GICs exhibit unique properties not found in other binary or ternary GICs. The compounds which Riidorff studied in 1954 were specific NH3-K ternary GICs prepared under a specific set of conditions. By contrast, there is a whole range of possible states of NH3-K ternary GICs which have the same stage but varying composition. Acknowledgements Thanks are due to P. Vora and especially to T. J. Pinnavaia and R. D. Spence for useful discussions. We also thank A. W. Moore for supplying the graphite used in this work. This research was supported by the NSF under contracts DMR80-10486 and DMR82-11554. The X-ray equipment was provided by the Office of Naval Research under contract N00014-80-C-0613. References H A. Hérold, in F. A. Lévy (ed.), lnlercalated Layered Materials, Vol. 6, Reidel, Dord- recht, 1979, p. 323. B. R. York, S. K. Hark and S. A. Solin,Phys. Rev. Lett., 50 (1983) 1470. B. R. York, S. K. Hark and S. A. Solin,Synth. Met, 7 (1983) 25- P. Vora, B. R. York and S. A. Solin,Synlh. Met. 7 (1983) 355. W. Riidorff and E. Schulze, Angew. Chem., 66 (1954) 305. W. Rudorff, Adv. Inorg. Chem. ltadioclicm., I (1959) 323. 6 A. Hérold, Mater. Sci. Eng, 31 (1979) 1. 0“le0 262 7 J. V. Aerivos and J. Azebu, J. .\lagn. Reson.. 4 (1971) 1. 8 J. W. .\leBain and A. .\1. Bakr, J. Am. Chem. Soc., 48 (1926) 690. 9 S. A. Solin, Adv. Chem. Phys, ~19 (1982) 455. 0 F. Be’guin and R. Setton, Carbon, 13 (1975) 293. F. Béguin, L. Gatineau and R. Setton, Proc. 6th London Int. Carbon and Graphite Con/Z, 1983. p. 97. 11 S. K. Hark, B. R. York and S. A. Solin, Bull. Am. Phys. Soc., 28 (1983) 348. 12 N. Daumas and A. Hérold, Bull. Soc. Chim. Ft, 5(1971)1598. 13 R. Setton, F. Béguin, J. Jegoudez and C. Maziéres, Rev. Chim. Mince, 19 (1982) 360. H S. K. Hark, B. R. York and S. A. Solin, to be published. 15 S. A. Solin, Adv. Chem. Phys, 49 (1982)-155, and reference therein. 1 Synthetic Metals, 7 (1983) 355 - 360 RAMAN SCATTERING STUDY OF ALKALl-MOI.ECULAR 'I‘ERNARY GRAPHITE INTERCALATION COMPOUNDS P. VORA, B. R. YORK and S. A. SOLlN Department of Physics and Astronomy, .\lichigan State University, East Lansing, MI 48824-1116 (U.S.A.) Summary We have studied the room temperature Raman spectra of several stages of ternary GICs prepared by sequential intercalation of potassium—binary GICs with either furan, THF or NH3. A Raman study, which detects the shift of the graphite intralayer vibrational frequency due to intercalation, provides a useful tool for the probe of charge transfer of these novel mate- rials. For all three stage 1 compounds, we have observed an intralayer graphite vibrational frequency at ~1606 cm'l (upshifted from the pure graphite peak at 1582 cm"), which is Fano broadened and has a Raman profile similar to that exhibited by LiC6 and EuCg. These results are com- pared and contrasted with those for the alkali binary GICs. 1. Introduction The study of graphite intercalation compounds (GICs) continues to be an exciting field of interest [1, 2]. These interests stem primarily from the possibility of studying 2D systems and their corresponding phase transitions. Another aspect, which has been the subject of considerable debate, has been the nature and quantitative value of the charge transfer to the host lattice [3]. Most work to date on GICs has been on binary systems where there is an intercalation of only one species in the graphite host. For a ternary GIC, however, there exists the possibility of an even richer variety of phases and also of intercalation of competing (donor Lersus acceptor) species. This, in turn, may give a better understanding of 2D phase transitions and the nature of the interaction between the intercalant and the host [3]. In this paper, we present the Raman spectra for three different graphite ternary systems; K-furan, K-THF and K-NH, GICs. X-Ray studies [4 - 8] on these materials have shown that these temaries exhibit different stages and also undergo interesting staging phase transitions. Our Raman results 0370-6779/83/83.00 © Elsevier Sequoia/Printed in The Netherlands 356 supplement the X-ray data and also give further insight into the above mentioned aspects of intercalant-host bonding. We compare and contrast our results with corresponding binary potassium GICs. 2. Sample preparation and experimental procedure The molecular ternaries were prepared by exposing the potassium binary GICs (previously made by using the standard two-bulb technique) [9] to the molecular species involved (i.e., furan, THF and NH3). A long glass tube with a break seal initially isolated the molecular liquid from the graphite-potassium system. After preparation of the alkali binary, the seal was broken and the sample exposed to the molecular liquid. Intercalation took place through the molecular vapor or through direct contact with the liquid. The tube was then further sealed to isolate the ternary GIC with just the molecular liquid to avoid interactions of the bare potassium with the liquids (furan and THF react with potassium over a period of time). Several samples were made in this manner with starting binaries of stages 1, 2 and 3. With NH3, a ternary was also made with an initial potassium stage 4 binary. Table 1 indicates the stages of the resulting temaries which resulted from intercalation of the various initial binaries. An argon ion laser provided the incident beam (k = 4880 A) for the scattering studies. The power of the beam was measured at ~30 mW near the sample. A Jarell-Ash double grating monochromator was used for the analysis of the scattered light. Alignment of the sample consisted of focussing a line image of the beam onto the GIC surface. The scattered light was collected and focussed on the spectrometer slit in the back scattering 90° geometry arrangement. The incident beam was polarized in the scattering plane to obtain a maximum coupling of the incident radiation to the sample. TABLE 1 Stages of the ternary GICs produced with different initial potassium binary GICs Initial K binary Stage of resulting ternary after intercalation" Furan THF NH3 KCg -- - 1 KCN 2 1 1 ** KC36 2 2 1 + 2 KCag — — 2 + 3 *Verified by X-rays. *‘A different stage 1 from the one made by KCB. 337 3. Results and discussion The Raman spectra of graphite (or HOPG) has been known for some time [10, 11]. There are two Raman active bands corresponding to a low frequency shear mode at ~12 cm " and a prominent intralayer mode at 1582 cm”. The changes of the 1582 cm " mode due to the intercalation of potassium in different stages have also been well studied [12]. Recently, the vibrational modes of the intercalant have also been observed directly [13, 14]. Figure 1(a) shows a spectrum for the KC,, stage 1 GIC. The 1582 cm ‘ band has been considerably Fano broadened [12] due to an electronic con- tinuum arising from the strong coupling between the potassium donor and the host [15]. There also exists a low frequency 563 cm” band arising from both disorder induced scattering and the (2 X 2)R0° ordering of the potassi- um in the plane [16]. However, the spectrum of a stage 1 ternary GIC, K- THF in this case, is noticeably different and is shown in Fig. 1(b). The sample was prepared by the intercalation of KC}, with THF vapor, as indi- cated in Table 1, and also characterized by X-rays to be a pure stage. Note from Fig. 1(b) that the intralayer mode has been shifted to 1606 cm”I and is considerably narrower. The spectrum in fact is more like that of stage 2 KC“ or stage 1 Equ [17]. Indeed, the color of the sample here is a deep blue, as in KC24, rather than the gold expected of KCB. The narrowness of the peak (AD’~ 22 cm“) inplies a weaker coupling between the 1582 cm“1 mode and the continuum than in the KCB case. Interestingly, this peak is even narrower than in KC34 (AU~ 29 cm"). The disappearance of the low frequency 563 cm’1 band is consistent with the X-ray observation [7] of disorder in the plane. A similar comparison has been made of the spectrum for the ternary (K-furan) C24. Here, the resulting ternary prepared from KC24 (see Table 1) is also stage 2. Again, the carbon intralayer mode of the ternary is sharper (A17~ 23 cm“) than that in KC24, as shown in Fig. 2. Since one would not expect any appreciable change in the graphite phonon modes in the two cases, the width seems to be related to the charge transfer between the host and intercalant. We suggest that the furan molecule, being an acceptor, reduces the charge transfer to the host from the potassium donor. There is no appreciable shift in frequency between the two compounds. Studies performed on the (K-NI—I3)C24 system are shown in Fig. 3. Here, again, the ternary was prepared via the NH3 intercalation of KC34. At room temperature, where the vapor pressure of NH3 surrounding the sample is “:10 atm (there is an excess amount of NH3 liquid in the sample tube), the ternary is a pure stage 1 [8]. It has been shown from the X-ray results [8] that as the vapor pressure of the NH3 is reduced by maintaining the excess NH3 at liquid N; temperatures, for which the vapor pressure sur- rounding the sample is <0.1 Torr, the sample evolves into a three phase system containing KC”, KCM and pure stage-2 K(NH3)._06C23_,,6. These changes are clearly reflected in the Raman spectra as well. The appearance of 358 Kca I ,' t“ i «- A \M/ ““an . a. / IMM.MW .5: - // g “M/ 2’. (a) 2 - S l K(THFi c l i‘ '2 x 24 3 _, g Q KC” 3 5‘ K(Furan)‘ 02‘ —_ 3 E. .. - " z e S .1, Q d l\ s _ ‘1 ,2 \l1 (b) \4v- 1 I I n l L 1:: 1500 1000 500 ‘ 0 1700 1650 1600 1550 Raman mm (001”) . Raman shift fem") Fig. 1. The Raman spectra of an (a) stage 1 KC], and (b) stage 1 ternary K(THF),KC24. The abcissa in this Figure and the Figures which follow is linear in wavelength rather than wavenumber. Fig; 2. Comparisons of the widths ~of the graphite intralayer mode for samples KC24 (Av = 29 cm“) and K(furan)xC24 (AV = 23 cm"). (a) (suun Meiiiqrv) Aigsuaiui l l l 1500 1000 500 " 0 Raman shitt (cm.‘) Fig. 3. The room-temperature Raman spectra of (a)stage 1 K(NH3)4,33C23_36, excess NH3 at room temperature, and (b) stage-2 K(Nll3)|.66C23.86 + KC3 + KC“, excess NH3 at 77 K. 359 the 563 cm"i peak is an indication of the presence of KCB which. With KC24, tends to reside on the sample surface. The superposition of the intralayer mode at slightly different wavenumber shifts for KC:, and K(NH3),,66C23,86 has resulted in the broadening of the ~ 1580 cm“1 peak. Similar staging phase transitions may occur in the furan and TIIF systems as a function of inter- calant vapor pressure as well. Other higher stage phase transitions in the K-NH3 ternary GIC system, as seen by X-ray [18], have also been evident in the corresponding Raman spectra. Figure 4 shows the change to a stage 2 + 3 (K-NH3) ternary from a stage 4 KC48 sample. Similar results were also seen for the THF and NH3 samples [19]. The relative intensities of the members of the doublets shown are consistent with the stage 4 and stage 2 + 3 designations. It can be clearly seen that the results presented here indicate a richness in variety of possibilities for a better understanding of GIC systems. The Raman scattering results have also thrown light on understanding the width of the graphite intralayer mode. The Raman results are consistent with, and supplementary to, the X-ray results. KC48 “mm/{WW 5 8 3 $2. (a) < J I In 2 “ E K(NH3)X C48 ‘2 _. ‘( S. E V0 '1 Viv/W \‘Wk’kd (b) l L 1‘ 1700 1650 1600 1550 Raman shitt (cm '1) Fig. 4. Raman spectra showing the transformation of a stage 4 KC-is binary GIC ‘0 a stage 2 + 3 K(Nl13)xC48 ternary GIC. 360 Acknowledgments A. We are grateful to S. K. Hark for useful discussions. Thanks are due to W. Moore for providing the graphite host material used in this study. This work was supported by the National Science Foundation under grants number DMR80-10486 and DMR82-11554. References 14 15 16 17 18 19 S. A. Solin, Adv. Chem. Phys, 49 (1982)-155. M. S. Dresselhaus and G. Dresselhaus, Adv. in Phys, 30 (1981) 139. M. S. Dresselhaus and G. Dresselhaus, Top. Appl. Phys., 51 (1982) 3. F. Béguin, J. Jegoudez, C. Maziéres and R. Setton, to be published. F. Béguin, L. Gratineau and R. Setton, to be published. W. Rudorff and E. Schulze, Angew. Chem., 66 (1954) 305', W. Riidorff,.4dv. Inorg. Chem. Radiochem., 1 (1959) 323. S. K. Hark, B. R. York and S. A. Solin, Bull. Am. Phys. Soc., 28 (1983) 348. S. K. Hark, B. R. York and S. A. Solin, Synth. Met, 7 (1983) 257. A. Hérold, Mater. Sci. Eng., 31 (1979) 1. R. J. Nemanich, G. Lucovsky and S. A. Solin, in M. Balkanski (ed.), Proc. Int. Conf. on Lattice Dynamics, Flammarion, Paris, 1978. R. J. Nemanich, G. Lucovsky and S. A. Solin, Solid State Commun., 23 (1977) 117. F. Tuinstra and J. L. Koenig, J. Chem. Phys, 53 (1970) 1126. S. A. Solin and N. Caswell, J. Roman Spectrosc., 10 (1981) 129. N. Wada, M. V. Klein and H. Zalul, in L. Pietronero and E. Tosatti (eds.), Physics of Intercalation Compounds. Springer, Berlin, 1981. P. C. Eklund, J. Giergiel and P. Boolchand, in L. Pietronero and E. Tosatti (eds.), Physics of Intercalation Compounds, Springer, Berlin, 1981. H. Miyazaki and C. Horie, J. de Phys.. 42 (1981) C6-335. D. M. Hwang, S. A. Solin and D. Guérard, in L. Pietronero and E. Tosatti (eds.), Physics of Intercalation Compounds, Springer, Berlin. 1931- D. M. Hwang and D. Guérard, Solid State Commun., 40 (1981) 759. S. K. Hark, B. R. York and S. A. Solin, to be published. P. Vora, B. York and S. A. Solin, to be published. Printed in Great Britain. @ Solid State Communications, Vol.50,No.7, pp.59S-599, l98é 0038-l098/85 $3.00 + .00 Pergamon Press Ltd. TUNEABLE SANDWICH THICKNESS IN POTASSIUM-AHMONlA GRAPHITE INTERCALATION COMPOUNDS 5.x. Hark: B.R. York, 5.0. Mahanti, and S.A. Solin Department of Physics and Astronomy, Michigan State University, East Lansing, MI 48824-1115. USA Received 27 February 1986 by J. Tauc He have studied the composition dependence of the sandwich thickness. d in stages 1 and 2 (n - 1,2) K(NH3)xC24 (01> 1 for xK of interest ) + re) and dA RA d 0 ds 1 * ‘d—+ E. + YxK(d—- ' 1)] x1 x2(1 - xx x )2 1_ K K g - K K1 (4) a? ““11 "k’kA‘i 'k—‘i K Hawralak and Subbaswamy4 found that for K C . the electrostatic contribution to d is qfiice small and we also find it to be neg igible. Thus we have fit ds vs. PNH of Fig. 3 using the Langmuire form df x1(PNH3) given above and Eq. (4) with only two adjustgble parameters which are the coefficients of x in the numer- ator and denominator of the firgt tenn. The results of that fit are shown in Fig. 3 and constitute very good agreement with experiment. The small deviation between theory and experi- ment evidenced in the expanded plot of dS vs. PNH of Fig. 3 shows that the latter exhihits a s ightly larger slope in the plateau region. This discrepancy cannot be removed by includ- ing the electrostatic tenn of Eq. (4) and may be due to nonlinear effects in Eq. (3) and/or to a variation in the force constants with f. From the fit of Eq. (4) (1st term) to Our data, we find directly a ratio for the ammonia and potassium force constants of kA/kK - .86 indicating that the NH3 molecule is softer than the K2' ion as expected. Knowing kA/k and the fact that x = 1/24 we can determine the rela- tive elastic contributions of the size differ- ence term, dA/dK - kA/kK. and charge transfer term YXK(dKo/dK+ - 1) in Eq. (4) provided that we can establish a value of r. The proper value of Y I 4.8 and the data of Fig. 3 (inset) were determined as follows: For a given PNH we know In(001)[Fl9‘ 2], dS [Fig. 3] and the3 11.14 ,, gravimetrically measured 1H3 uptake x. POTASSIUM‘MDTUbIA CRAPhITE INTERCALATION COMPOUNDS . S97 Calculated In depend on X", dS and r. Also. x I 01x1 + 02x2, where On is the fraction of stage n and 91 + 92 . 1. For fixed Y, on and X" (n - 1.2) can be fOund from the simultane005 equations for l and on Trial Y's were tested for compatibiliiy with Fig. 2. The best fit was obtained with Y- 4.8 and Fig. 3 (inset) for xn vs. PNH is based on that value. As independent c0 firmation of this choice of x we note that the maximum values of x and x deduced from Y a 4.8 are in very goo agree- ment with the saturation weight uptakes derived using the well established ”volumg available” criteria of Setton and coworkers. 5 From the above cited parameters and Eq. (4) we find that although the size difference term dominates (70% contribution), the charge trans- fer term accounts for fully 30% of the sandwicn expansion. These results have the following physical origin: As NH is added to the layer interspace which contai s K+ ions. there are V three major effects: First, some sandwich ex— pansion results from the size difference of ammonia (A) relative to ionized potassium (K‘) i.e. dA/dK+ > 1. Second. as ammonia is inges- ted. the charge originally donated to the car- bon layers is extracted back into the K—NH3 layer and delocalized. Much of the delocalized charge resides near K+ ions leading to an in- crease in their effective radii. The increased radii of the potassium ions generates additional elastic energy which leads to an increase in dS . Third, NH entrance is concurrent with a dilu- 1 tion of tde K-C layer stoichiometry from KC12 to KC 4 evidencing a stage 2 * stage 1 tran- sitioh at constant K composition. As a final point we consider the d2 vs. x2 data of Fig. 3. Clearly d is keyed more to the variation of d with x1 than to x2 since x2 is essentially residue-like and pressure insen- sitive at low PNH . The keying of d2 to 01 is not surprising in view of the expected high mObility of NH between laterally conti900us stage I and stage 2 regigns as would exist in the Daumas-Herold model1 of the layer struc- ture of GIC's. Thus. the expanded stage 1 regions impose both a dynamic and static strain on the layers of the stage 2 regions which accordingly acquire a separation 0 only slightly less than ds . S2 1 Acknowledgement--He gratefully acknowledge use- ful interactions with T.J. Pinnavaia. J.L. Dye, R. Setton, K.R. Subbaswamy, and especially P. 598 POTASSIUM-AMMONIA GRAPHITE INTERCALATION COMPOUNDS V01. 50, No. -' 6.623 -6973 I v ‘— lfl- : C - 'O-eecoo'°""oe' ,. -6.83I Fig. 3. The dependence of sandwich thickness, ds . on PNH displayed on a normal (13ft-hand drdinate) and expanded (right-hand ordinate) scale. Experi- mental points including one at the origin of the normal scale. are in- dicated by solid dots (0) and open circles (0) for n - 1 and n a 2, re- spectively. The solid line is a fit to the data using Eq. 4 with the following parameters: 8 . o, dKT s 5.35 A, dKo - 7.27 K. dA . 5.47 A. kA/kK - 0.86, r - 4.8, xK . 1/24. (Inset) The dependence of x for n - 1 (o) and n a 2 (0) on PNH . nThe solid line is a fit to the data using the Langmuire equation (see text) x1 . 4.49 PNH /(.0447 + P NH )' Unless otherwise indicated. error Bars are smaller than the dots. Vora. who assisted with some data acquisition and made major contributions to sample prepar- ation. Thanks are also due to A.N. Moore for REFERENCES 1. S.A. Solin, Advances in Chemical Physics 32, 455 (1982). 2. S.A. Safran and D.R. Hamann, Physical 7. Review B 22. 606 (1980). 3. S.E. Millman and G. Kirczenow, Physical Review B gg. 2310 (1982). 4. P. Hawrylak and K.R. Subbaswamy, Physical Review (in press). See also J.R. Dahn, D.C. Dahn, and R.R. Haering, Solid State Communications 42. 179 (1982). S. A. Metrot and J72. Fischer, Synthetic Metals 3, 201 (1981). 6. K.C. H00. H.A. Kamitakahara, D.P. Divincenzo, D.S. Robinson, H. Mertwoy, providing the HOPG used in this study. This work was supported by the NSF under grants DMR 82-11554 and DMR 81-17297. 9. J.W. McKiben, and J.E. Fischer, Physical Review Letters 59. 182 (1983). A. Herold, Proceedings of the International Conference on the Physics of Intercalation Com ounds, edited by L. Pietronero and E. losgtti lSpringer-Verlag, New York, 1981), p. . M.E. Misenheimer, P. Chow. and H. Zabel, Materials Research Society Proceedings 20, edited by M.S. Dresselhaus, et al. (North Holland, New York.1983). p. 283. H. Rudorff and E. Schultze, Angew. Chem. §§. 305 (1954). V01. 50, No. 7 POTASSIUM-MIMONIA GRAPHITE INTERCALATION COlfPOUNDS 599 10. S.K. Hark, B.R. York. and S.A. Solin, Syn- 14. J.R. McBain and A.M. Baker, Journal thetic Metals, in press. American Chemical Society fig. 690 (1926). 11. B.R. York. S.K. Hark. S.D. Mahanti. and 15. F. Beguin, R. Setton. L. Facchini, A.P. S.A. Solin, to be published. Legrand, G. Merle, and C. Mai, Synthetic 12. T.L. Hill, Statistical Mechanics) Princi- Metals ;. 161 (1980). ples and Selected AppTications (ficGraw- 16. N. Daumas and A. Herold, Bulletin Society Hill, New York, 1956). Chim. Fr. S, 1598 (1971). 13. T. Enoki, M. Sand, and H. Inokuchi, Jour- nal of Chemical Physics lg, 2017 (1983). NOVEL PROPERTIES OF ALKALI‘METAL AMMONIA GRAPHITE INTERCALATION COMPOUNDS S.A. SOLIN,’ Y.B. FAN,‘ AND B.R. YORK’ 1'Department of Physics and Astronomy, Michigan State University, East Lansing, MI. The ternary alkali-metal ammonia graphite intercalation compounds (GIC's) n(NH3)xC12n where M - K, Rb, Cs, and n - 1,2,... have been shown to have unusual staging transitions when the binary 010 is exposed to ammonia.‘ These staging transitions have been studied using (001) x-ray diffraction methods‘ and Raman scattering techniques. In this paper we will report the inplane (hkO) patterns of the saturated potassium ammonia GIC K(NH3)u 38CN which is a pure stage-1 compound with long-range c-axis correlations. We will show that the K- NH3 intercalants actually form a two-dimensional (2-D) liquid in the graphite galleries. This liquid is the 2-D analog of the well- known' 3- D K- ammonia solution which has been heavily stueied via-a-vis the metal-insulator transitiion. The K(NH3)4.3BC 24 samples studied here were prepared by exposing K02” to NH3 vapor at a pressure of ~10 atm. The Kczu was prepared from highly oriented pyrolytic graphite (HOPG) using the usual two-bulb method.” It was transferred in a glove box (£_0.5 ppm 02, H20) to a pyrex tube into which dry NH3 was condensed at liquid nitrogen temperatures. The end of the pyrex tube was epoxied to a thin-walled (< .005") aluminum can which contained the sample and was capable of simultaneously withstanding high NH3 pressure, yet providing high transmittance of the incident and diffracted photons. The sample tube was sealed under vacuum at - 10 5 torr.and the NH was allowed to 3 warm to room temperature, thus acquiring an equilibrium vapor pressure of - 9 atm. and intercalating the K02” specimen. Note that since no potassium was expelled during the insertion of NH3, the potassium layer stoichiometry in the resultant stage-1 compound is Kczu corresponding to a reduced K density in comparison to the KC12 layer stoichiometry of the K02” binary GIC. The diffraction data reported here were acquired with incident MoKa radiation from a Rigaku 12 kw rotating anode source coupled to a vertically bent graphite monochromator and a computer-controlled Huber 4-clrcle diffractometer. A NaI scintillation detector was also used. In Ficg. 1 .is shown the corrected inplane diffraction pattern (3.13) for K(NH3)u 38 C2” acquired at room temperature. The solid line of Fig. 1 results from applying several correction factors to the as recorded data. Let lexp(q) be the observed (hko) diffraction pattern. This contains Bragg reflections associated with the powder pattern of the aluminum sample can and the ordered carbon layers. A reference (hkO) pattern acquired from a sample free region of the aluminum can was recorded. appropriately scaled to and subsequently subtracted from Iexp(q)' The Bragg peaks associated with the carbon layers were removed (for clarity) from the resulting pattern to yield Iéxp(q) whichis the observed diffuse scattering. The pattern Iéxpfq) was further corrected’ for absorption, and the Lorentz polarization factor to yield I'(3) as follows: I'(q) . 1|:éxp(q)[('l‘lcose)exp(-u'I'/cose)J.1 x 226')]-1 (1) [(1 + coszze' cos226)/(1 + cos Here T is the sample thickness, u is the effective absorption coefficient, 6 is the diffraction angle, and 6' is the graphite monochrometer diffraction angle for the (OCR) MoKo reflection. The corrected diffuse scattering function I'(q) which is shown in Fig. 1 was scaled to oscillate about the incoherent scattering contribution 2 1mm) - 2 rm + i(m) <2) uc (shown as a dashed line in Fig. 1) at high q and from this scaling the ordinate scale of Fig. 1 (in electron units) was established. In Eq. (2), fm is the atomic scattering factor of the m'tn atom, uc + unit of composition which in our case is K(NH?" 38 and i(m) is the Compton modified scattering from the K, N, H, and C atoms. ; i 9 5» =2 g 12.0 g m E E 3,00 ; 3 ~ 1 ‘3 1 E 4.0 (E 1 '1 0.0 1.0 2fo 310 {o 51.0 efo f on 4:0 3}) Bio 150 am) i r«) Fig. 1. The diffuse inplane (qic) Fig. 2. The pair distribution func- scattering I'(q) from K(NH3)u 38C2A tion 2wrp(r) deduced by applying Eqs. (solid line) and the incoherent 2 and 3 (see text) to the data of Fig. contribution (dashed line) as des- 1 with u - 5.0 mm-1, c - 0.7 A, and cribed in the text. T - 1.0 mm. We have deduced the two-dimensional pair distribution function 2wrp(r) of the K-NH3 liquid using a Bessel function transform of I'(q). namely qm 2 -c2q2 2wrp(r) - anpo + r g [I'(q) - IInc(q)] (q/fe) Jo(q,r)e dq (3) Here po is the average areal density, fe is the q-dependent scattering factor per electron, and a is a damping factor that minimizes cut-off errors associated with the finite range, qm, of the integral of Eq. (3). The pair correlation function that results from the application of Eq. (3) to the data of Fig. 1 is shown in Fig. 2. As expected, Zwr(p(r) - 00) + 0 at large values .of q. Moreover, the cut-off error alluded to above is <15 electrons as 105A deduced from g 2nrp(r)dr. From the known sizes of the K+ ion and the NH3 molecule we can deduce the origin of some of the peaks in the pair correlation function. The peak at -»3.5A.or equivalently the broad peak in I'(q) centered at q - 2.10A-1 is associated with the K-NH3 and NH3—NH3 correlations of a liquid that is dominated by clusters composed of K ions to which four NH3 molecules are bound.‘ This four-fold coordination is the natural 2-D analog of the 6-fold octahedral coordination of the 3-D K-NH3 solutions and is compatible with the composition of our specimen if there are 0.33 "spacer" NH molecules per 3 K(NH3)u cluster as indicated by preliminary NMR studies.’ The peak in 2nrp(r) at.r~ - 7.5A1(or equivalently the peak at q - 0.85A-F in I'(q)) corresponds to K-K correlations and its position and width indicate that there is no tendency for the K+ ions to occupy carbon hexagon center sites. Thus, the K-NH3 intercalant does indeed acquire a liquid structure between the carbon layers. ACKNOWLEDGEMENTS We acknowledge useful discussions with S.D. Hahanti, K.R. Qian, and D. Stump. Thanks are due to A.W. Moore for providing HOPG. This research was supported by the NSF under grant DMB82-1155u. REFERENCES 1. S.K. Hark, B.R. York, S.D. Mahanti, and S.A. Solin, Solid State Comm. 22, ' 595 (198A). 2. P. Vora, B.R. York, and S.A. Solin, Syn. Met. 1, 355 (1983). 3 J.C. Thompson, Rev. Mod. Phys. £2, 70“ (1968). u A. Herold, Bull. Soc. Chim. Fr. 999 (1955). 5. See e.g. H.P. Klug and L.E. Alexander, X-Ray Diffraction Procedures, 2nd 6 7 Ed. (Wiley, New York, 197”). p. 8A9. . K.R. Qian, D. Stump, and S.A. Solin, to be published. . H. Resing, B.R. York, S.A. Solin, to be published. OPTICAL REFLECTASCE STUDIES OF C24K and C24K(NH3)4.3 0.x. HOFFMAN*, A.M. RA0*, G.L. DOLL', P.C. skLUND*, B.R.*}oak**, and S.A. SOLIN’ , University of Kentucky, Lexington, xv 40506; Michigan State University, East Lansing, MI 48824. INTRODUCTION Rudorff and Schulze [1] in 1954 reported the synthesis of the potassium-ammonia GIC's. These ternary compounds are now being studied extensively because of recent observations of staging transitions [2] and tuneable intercalate layer thickness [3] associated with variable concentrations of NH3 in the intercalate layer. The stage n ternary GIC's CzanK(NH3)x (n-l,2,3) can be prepared by the reaction of the stage n+1 binary GIC's ngnK with the vapors of NH3 [2]. In this paper we report the results of optical reflectivity studies on the NH3-saturated stage 1 ternary and Cz4K in the photon energy range 0.4-6 eV. The data were analyzed via Kramers-Kronig analyses to separate intra- and inter-band contributions to the dielectric function. By comparisdn of the optical results obtained for the binary and ternary GIC's we were able to determine the change in charge transfer between the intercalate and carbon layers induced by the NH3-intercalation. Since the carbon layers in C24K(NH3)4,3 are separated by 6.63 A we anticipate only weak C-C interlayer coupling. The data are therefore also discussed in terms of the quasi-2D model of Blinowski et al [4] applied previously to acceptor GIC's [lo-6] . EXPERIMENTAL DETAILS The ammonia-saturated ternary GIC's were prepared from highly-oriented- pyrolytic-graphite by NH3-intercalation of stage 2 Czax [2]. The 024K samples were transferred to cylindrical quartz tubes suitable for optical measurements and evacuated. An excess amount of anhydrous NH3 was then distilled into the tube and kept frozen until the tube was sealed off. The stage of the samples were characterized using Raman spectroscopy [7]. The penetration depth for the Raman scattering probe is the same as that for reflectance measurements. Near-normal incidence reflectance measurements were made using single-beam reflectometers described elsewhere [6]. RESULTS AND DISCUSSION The Spectra of C24K and 024K(NH3)4.3 are shown in Fig. 1. The insets to the figure are the respective Raman spectra of the samples. In all the figures we adopt the convention of labeling the C24K data as (a) and the C24K(NH3)4,3 data as (b). The Raman data are in good agreement with that published previously [8]. The dashed lines in Fig. 1 represent the data extensions used to perform the Kramers-Kronig analyses. The data deviate from the dashed lines due to either absorption in the NH3 vapors or the ampoule walls. Both spectra exhibit metallic character with prominent Drude edges near -l.8 eV associated with intraband absorption; higher energy features are identified with interband transitions [4-6]. Our C24K spectrum is in good agreement with a previously published spectrum [9] in the range 0.6 to 2.2 eV, however this spectrum was not analyzed to determine a plasma frequency. Kramers-Kronig transforms of our data were carried out to determine the real and imaginary parts of the dielectric function. Results for the imaginary part and the energy loss function are shown in Fig. 2 and Fig. 3, respectively. The dashed lines in Fig. 2 indicate the interband contributions. The separation of inter- and intra-band contributions are accomplished by fitting the low energy data to the Drude form for intraband absorption [6] and the results are shown in Fig. 4 where the theory is indicated by the solid lines. The Drude parameters obtained from this analysis are: (a) C24K--‘fimp-3.9 eV, mpr-27 and (b) C24K(NH3)4.3 -fimp-2.7 eV, upr-ZZ. The plasma frequency mPis related to the carrier concentration N and the effective mass m via wp'léwNezlmll/Z (1) and r is the carrier lifetime. Accordingly, the ratio of the square of the plasma frequencies for C24K and 024K(NH3)4.3 is given by In,./u,012 - <£,n,n°>/<£°non,>, (2) where the additional subscripts o and a refer to the C24K and C24K(NB3)4.3 parameters, respectively. we have also incorporated into (2) the relation N-fn, where n is the K concentration and f is the fractional mobile charge donated per K atom. In both compounds we assume the K 43 band is above the Fermi level (Hp). Thus fo-l and, to a good approximation, mo-ma. Using (2) we therefore arrive at a value fa-O.73, or 0.73 electrons donated to the graphitic pi-bands per K atom. The remaining 0.27 electrons must then reside in lower 'lying NH3-derived states. Our value for fa can be compared to 0.83 determined recently by NMR studies [10]. In the remaining space we briefly discuss our results in terms of the 2D formalism developed by Blinowski et a1 [4]. Using f-.73 and a value for 70-2.4 eV in the expression [4] 2,-2.33 yo [HUN2 (3) we arrive at EF-0.97 eV. Furthermore, in the 2D limit [4], amp-[8.69 2,11/2, which yields‘fim =2.9 eV. This value compares favorably with the observed value of 2.7 eV. This preliminary analysis seems to indicate that the C24K(NB3)4.3 compounds are quasi-2D and reasonably described by the theory of Blinowski et a1 [4]. Further study and analysis will be necessary to establish the connection. ACKNOWLEDGEMENTS This research was funded, in part, by grants from DOE DE-FGOS-84ER45151 (D.R.), American Cyanamid (U.K.), and NSF DMR 82-11554 (M.S.U.). REFERENCES 1. W. Rudorff and E. Schulze, Adv. Inorg. Chem. Radiochem. 1 (1959) 323. 2. S. K. Hark, B. R. York, and S.A. Solin, Synth. Metals 7 (1983) 255. 3. S. K. Hark, B. R. York, S.D. Mehanti and S. A. Solin, Solid St. Comman. _5__0, (1984) 595. 4. J. Blinowski, N.R. Ran, C. Rigaux, J.P. Vieren, R. LeToullec, G. Furdin, A. Herold, and J. Melin, Journal de Physique 41 (1980) 47. S. P.C. Eklund, D.S. Smith and V.R.K. Murthy, Syn. Met. 2, 111 (1981). 6. R.E. Heinz, G.L. Doll, P. Charron and P.C. Eklund, Mat. Res. Soc. Symp. Proc. Vol. 20, 87 (1983), M.S. Dresselhaus, G. Dresselhaus, J.E. Fischer and M.J. Moran, eds. Elsevier, New York. 7. M.S. Dresselhaus and G. Dresselhaus, in Topics in Appl. Phys. 21; Light Scattering in Solids, M. Cardona and G. Guntherodt, eds., Springer-Verlag, Berlin (1982). 8. P. Vora, B. R. York and S.A. Solin, Synth. Met. 7 (1983) 355. 9. M. Zanini and J. E. Fischer, Mat. Sci. Eng. 31 (1977) 169. 10. H. A. Resing, B .R. York and S. A. Solin, to be published. ID‘ I T I I I 3 ‘3 (a) - LC) 3 (b) 500 1600 I700 " -8 Q) l 2 A. (cm)" C3 *5 .6 £2 9* Q) 0: .4 .2 O 2 4 6 Photon Energy ,‘hw (eV) Fig. 1. Reflectance of (a) C24K, and (b) C24K(NH3)4.3. The dashed lines indicate extensions used in the analysis. 2 r. I I I I T I g I f. g; —(0) A '- u _-_ (b) n “ ii ’19: a- g: : I; i w I ' 2 I'- 1 1:1- 1 3 1 " "’ : f 3 1 I g g ‘ t '-. ‘ I i. a. 1' 1“ I ...\\\‘ O J i L Li i'" Photon Energy , ‘hw (eV) Fig. 3. The electron energy loss function for (a) CZ4K and (b) C24K(NB3). Dots were used in the energy ranges where extensions to the data were used. Photon Energy . hm (eV) Fig. 2. Imaginary dielectric function of (a) C24K and (b) 024K(NH3)4.3. The dashed lines indicate the inter-band contribution. 2.0IIIIT °°°°° ° Experiment — Theory J l o 075 Co 1.5 (m2, (M) Fig. 4. The solid lines were calculated using the Drude expression for the dielectric function with parameters found in the text for (a) C24R and (b) C24K(NR3)4.3.