GLASSY CARBONS Semi-Annual Progress Report for the Period June 1, 1971 to December 31, 1971 January 1972 ARPA Order Number: 1824 Program Code Number: 1D10 Contractor: The Regents of The University of Michigan Effective Date of Contract: 1 June 71 Contract Expiration Date: 31 May 72 Amount of Contract: $150,263 Contract Number: DAHC15-71-C-0283 Principle Investigator: Professor Edward E. Hucke Department of Materials & Metallurgical Engineering The University of Michigan Ann Arbor, Michigan 48104 (313) 764-3302

The views and conclusions contained in this document are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied, of the Advanced Research Projects Agency or the U.S. Government.

TABLE OF CONTEITS Summary I. Introduction i. M'aterials Preparation III. Structural Studies A. Solid Structure Microscopy and Diffraction Thermodynamics B. Pore Structure Helium Pycnometry Surface Area Mercury Porosimetry Electron Scanning Microscopy IV. Mechanical Property Evaluation Hardness Compressive Strength Ultimate Tensile Strength Modulus of Elasticity References Appendix ii

GLASSY CARBONS Sunmar The program has three major areas of endeavor, (1) Materials Preparation, (2) Structural Characterization, and (3) Property Evaluation and Correlation. The major accomplishment in the materials preparation program has been the successful preparation of at least some samples with cross sections up to 1.2 inches in diameter. Previous materials available have been limited to sections of about 1/8 inch, and in rare cases up to 1/4 inch. In addition, samples have been successfully made with a bulk density of from.5 grams per cm3 to about 1.25 gm/cm3. These materials have been prepared with a variable pore size ranging from 46 angstroms to 50 microns. These samples represent a significantly lower bulk density than previously available. Improvement in the yield of crack free material has also been made. Present efforts are toward more control over the polymerization of the furfural alcohol type resins employed together with better control over other processing variables such as heating rate. In structural characterization, techniques have been established and results obtained on at least some samples using optical microscopy, scanning electron microscopy, transmission electron microscopy, dark field electron microscopy, wide angle electron diffraction, and wide angle X-ray diffraction. iii

Insufficient samples have yet been examined to draw any general conclusions, however, it has been possible to reveal various elements of structure including pores, cracks, and "crystallites" on a size scale ranging from 20 angstroms to 100 microns. In several cases changes in this structure with final heat treating temperature have been observed. Additional structural characterization has been made by way of He density, pore size distribution (Hg porosimeter), and surface area by Knudsen flow and B.E.T. methods. Pore size has been correlated on several samples with structure obtained by microscopy. Several materials with substantial mechanical strength have been prepared having a N2-B.E.T. surface area of around 500 m2 per gram, a figure high enough to be of interest for chemical absorption applications. In the area of property evaluation preliminary data has been obtained for hardness, compressive strength, tensile strength, and sonic modulus. Suitable experimental methods have been established for each measurement and data gathered on several representative samples. Too few replicate samples have been run to establish any of these properties definitively. However, on a single sample basis tensile strengths of at least 10,000 psi and compressive strengths of 50,000 psi, should be obtainable on material with a bulk density of.8 gm/cm3. Sonic modulus has been varied over a range of 1.6-106 psi to 4.6-106 psi with further variation apparently possible. iv

A detailed literature study of the thermodynamic properties of carbon has shown the feasibility of determining alccur-atcly by an electrochemical cell technique the free energy and configuration entropy of glassy carbon. Preliminary measurements have been carried out, which when refined should yield directly an experimental determination of the degree of crystalline order in glassy carbon as well as more accurate thermodynamic data for the various carbon-oxygen equilibria. v

GLASSY CARBONS I. Introduction In recent years the availability of a group of highly disordered carbons in nonparticulate form has gained the attention of materials engineers. These materials have been called vitreous or glassy carbon, due mostly to their physical resemblance to black glass. In this report they will be called glassy carbon, meaning only that they are bulk carbons possessing little if any graphitic structure. The excellent combination of properties reported leads to the belief that such materials should be capable of solving many important problems. In particular, the very low density coupled with isotropic high strength and hardness at both low and extreme temperatures, and excellent corrosion resistance have lead to many proposed applications1. A full exploitation of the potential applications awaits development of more detailed information on properties obtainable. While all of the commercially available materials have rather striking simularities in reported properties there are significant variations reported in density, strength, conductivity, and stiffness, warranting the conclusion that differences in structure exist and might be further varied. All of these carbons are apparently produced by the controlled thermal decomposition of crosslinked polymers and therefore their -1

structures might be expected to be subject to the kind of polymer used as starting material, the evolution of the lo\w temperature crosslinked structure, as well as the details of the higher temperature decomposition. An excellent review of the chemical steps taking place during decomposition has been published2, but there is as yet relatively little information relating these changes to properties of the residual materials. The commercially available materials can be obtained only after the pyrolysis is substantially complete, typically after heating to temperatures of 1000~C to 3000~C. One purpose of this study is to follow the evolution of structure from polymerization through decomposition with appropriate correlation to the properties developed. The structural studies being carried out were chosen to yield information about structural details over a wide size range. Despite several X-ray studies3'4 and one low temperature specific heat measurements there has yet been no clear cut determination of the short range (one to ten nearest neighbors) bonding in glassy carbons. While this study will not produce an X-ray radial distribution function, it has provided controlled material for a cooperative study by neutron diffraction, where the same material is being examined by other.methods. In addition, this program is endeavoring to establish the short range disorder by an electrochemical cell measurement which allows calculation of the configurational entropy of glassy carbon relative to graphite. -2

Atomic arrangements on the somewhat larger scale of from 20A to 1000A are being conducted with dark and bright field transmission electron microscopy as well as electron diffraction and wide angle X-ray diffraction. Other structural details on a size ranging from 100A to the macroscopic are being developed using surface replica electron microscopy and scanning electron microscopy. From the previous studies it is apparent that the properties developed not only depend on the short range bonding of the carbon present, but on the amount, size, and distribution of the void space included in the structure. A considerable extent of void, either connected or isolated has been demonstrated at several size levels in glassy carbon materials. At the 4 to 10A level molecular sieve studies6 have revealed an interesting structure, while small angle X-ray diffraction7 8 surface absorption, and electron microscopy have revealed a void structure on a larger scale (20A to 1 micron). In this study small angle X-ray diffraction, scanning electron microscopy, helium pycnometry, mercury porosimetry, and surface adsorption are being selectively employed to study the pore structure. The foregoing details should influence the mechanical properties of the resulting carbon. A representative profile of typical properties is being measured to relate to the structures produced with particular attention to be focused on obtaining as wide as possible variation in properties. For -3

this purpose, hardness, compressive strength, tensile strength, and modulus of elasticity have been chosen. II. Materials Preparation In order to have control over sample processing throughout the various stages, carbon materials have been prepared in this laboratory. Representative commercial materials are being used for comparison purposes even though they are only available after nearly complete processing. The initial phase of the study surveyed a wide range of organic precursors and catalyst systems. At present concentration has been placed on furfural alcohol and a related commercial resin, Durez 16470*. The catalyst chosen was paratoluene sulfonic acid (PTSA). Over 90 different batches of material have been prepared yielding about 500 separate samples ranging in size from about 10 to 500 cm3. A wide range of procedures for catalyzing the mixes, including variation of catalyst level, the initial setting time and temperature, together with variation in the subsequent pyrolysis schedule has been employed. In many instances it was impossible to obtain samples without extensive cracking. At present an optimum procedure has not been established even though the yield of crack free samples has increased steadily. The successful survival of samples is strongly influenced by factors related to the low *Hooper Chemical Company, North Tonowanda, New York. -4

temperature treatment of the polymer and to very uniform heating rates in the pyrolysis step. Modifications of the furnace and furnace control system have been particularly helpful in increasing the yield of good samples. Heating rates in flowing N2 from 6 to 43~C/hr have been used up to 1000~C, with rates of about 200~C/hr employed for the range of 1000~C to 2000~C. The temperatures reported with a sample number refer to the highest processing temperature. In the low temperature range (less than 1000C) no holding time is employed, while in the high temperature cases a holding time of 1 hour is used. At present only a limited number of samples have been produced at temperatures higher than 2000~C. Sample cracking has in many instances been correlatable with the presence of small bubbles introduced in the mixing and casting operation. Further efforts are required to develop procedures to completely avoid this problem. In addition, particles of an unknown origin have been found in the original Durez resin. These particles appear to cause cracking during processing and a non-uniformity in structure in the final sample. They may be clearly seen as areas about 1 micron in size in Figure 1. These particles are apparently the same as those reported by Schmitt9. Additional work on the occurance and elimination of these particles is planned. Thus far it has been possible to produce sound samples of up to 1 1/8" thickness with a considerable range in structur~ -5

W44 4i j::::"~~~~~~~~~~~~~~~~~~~~~~~~~J::~ Figure 1. S.E.M. of Fracture Surface of 317-7 (700~C) showing large particles.:i: —6.::n:: i.~l'i:i!.'4 —::~~~,A;::*j~**::::* i:44:i 4..:i:i:: 44 4~li4.: /44ii /4' 44 " 4,, 4.'4 iii: 4 / i::4i4 v 7::i~i~:-~':: 4:''::::::: 4/4,,f ""!~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~; 4,~~~~~~~~~~~~~4~~~ Figure 1 S E M of Fracture Surface of~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~: 317-7e (70W showingo largce particles. -6

and densities (see later discussion). Further extension of sample size should be obtainable with more refined preparation techniques. III. Structural Studies While the eventual goal of this study is to correlate structure at various stages of the process with properties, the results thus far are largely drawn from samples taken at only two processing stages. The first stage corresponds to a maximum temperature exposure in the range 650-1000~C, while the second stage is that formed at about 2000~C. Further exploration of the stages of formation will proceed after appropriate techniques of sample preparation, structural analysis, and property measurement are fully developed. The techniques chosen for structural examination fall into two broad categories with respect to the information yielded. The first yields (predominantly) information about the state of of the solid making up the structure, while the second deals mainly with the void structure. In the first category this study is employing bright field and dark field transmission electron microscopy, electron diffraction, wide angle X-ray diffraction (pinhole and diffractometer methods) and scanning and electron microscopy. The last two methods also yield information on pore structure. Additional methods used to establish the pore structure are small angle X-ray scattering, helium pycnometry, mercury -7

porosimetry, and surface adsorption. An attempt is also being made to gain structural information through a precise measurement of the thermodynamics of the equilibrium C = C graphite glassy Since the effort required to carry out the above techniques varies substantially, a complete set of data will be collected on only a limited number of samples after screening with more routine tests. Apparent and real (He) density, wide angle X-ray diffraction, scanning electron microscopy (SEM), hardness, compressive strength, tensile strength, and sonic modulus are being run on representative samples of each batch in order to select appropriate samples for more intensive study. At this stage at least several samples have been examined by each of the methods mentioned, but no single sample has been examined by all. A. Solid Structure Microscopy and Diffraction Intensive structural analysis has been completed on two Durez based glassy carbons, 311-19 baked at 2000~C and 311-19 baked at 750~C, the only difference being the final baking temperature. Samples for direct transmission bright and dark field electron microscopy and electron diffraction consisted of small (0.1-1.0 diameters) particles prepared by filing clean fractured surfaces of the samples. The particles were -8

deposited directly on 400 mesh copper microscope grids. The edges of these particles were found to be tlhin enough for direct transmission of the electron beam in a JEM-6A electron microscope. In addition, two stage Pt shadowed replicas of fractured surfaces were prepared. The scanning electron micrographs were obtained from fractured surfaces that had been coated with a chromium-conductive layer. Wide angle X-ray diffraction pinhhole patterns were obtained using a l.0rmm thick sample and nickel filtered copper radiation at 35KV and 15ma for 1 hour. Diffractometer traces of the (002) and (100) peaks were obtained using the same radiation at a scan rate of 0.50/ 20 min. between 20 values of 10~ and 65~. Observation of a large number of sample particles using transmission electron microscopy revealed particle sizes ranging between 0.1 and 1.0 microns in diameter. Within each of the two samples the particles appeared to be uniform and to contain the same structural features. A typical micrograph of the sample baked at 2000~C, Figure 2a, shows the particles to be irregular in shape and to consist of irregularly shaped platelets having 0 diameters on the order of 150-500A. The platelets show a dis0 tinct granular texture, about 20-40A. The platelets appear to be uniform in thickness and based on the ease of penetrability 0 of the electron beam may be about 50-300A thick. A typical micrograph of the sample baked at 750~C, Figure 2b, shows similar features except that the platelets and granulation were not as 0 distinct anc well defined. Platelet size is smaller, 150-350A -9

1000/A (a) (b) Figure 2. Bright field transmission electron micrographs of glassy carbon particles. Micrograph (a) sample 311-19, 2000~C and (b) sample 311-19, 750~C. -10

0 in diameter, while the ill-defined granulation may be 20-30A in diameter although in some cases it is completely absent. Electron microscopy of replicas of the brittle fractured surfaces were not as successful in revealing structural features due to the large deformation that occured in the replica when it was removed from the samples' surface. However, occasionally in the 2000~C sample large circular features, 2 microns in diameter were noticed and found to have a different type of structure. These particles are thought to be related to particles found in the original resin. Figure 3 shows the surface to be textured somewhat similar to the platelets observed in the previous micrographs. This sample has been Pt shadowed and so the smaller granulation may be due to the granulation of the Pt and not the sample. A dark field micrograph, Figure 4, obtained from a small portion of the (002) diffraction ring show diffracting 0 regions 20-40A in diameter. No diffracting regions were observed when using the (100) or (110) rings. Selected area electron diffraction patterns taken through the edges of the particles consisted of diffuse halos corresponding to the (110), (100) and (002) planes although the (002) halo was missing for the 750~C sample, probably due to sample orientation with respect to the electron beam. The d-spacings are summarized in Table 1. The diffracting halos were sharper for the 2000~C material than the 750~C material -11

~ Gdwps DO000:'61-1T aq. uT PaJe.afns poanqDa: ~ _ 40 JO DT-[~.[ pIDMopPTS kd q'eqs OMq 0o suoTqPDTJTU6pui SnOTa:eA q. qd'eJ. -oaJDTUI uoq-aDG P1TT _ u-6Ta'~ enbT ~~ VOOOS jQ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~', ~~~~~~~~~~~~~~~~~~~ V.:x~..,~::~'"" ~ ~~~ N ~~~~~~~ ~ "%"~;!~"~'~'~:'~:?~:' ~':":'i ~':............. tA4 *i~~~~~~~~~~~~~~~~~~~~~' r:?~i.~! 0.. " ~'".~~r ~;,,*:,i -;."..,:'*4'~'.;',~,..;,~",, ~'%'~'"~'~ ~~~~~~~~~~~~~~~~~~~~~~.... _*~~~~~~~~~'71r~ ~ ~ ~ ~~ ~~.:~:r,,"X',r —:'~:,,.,'~;......,g~ r~ "l'-:" P~~~~?,,,:~~'t,:i~:.:''~'""......~: ~:.,//', ~ VOO0/_ j ~,~~~~~~~~~~~~~..?..~~~ i~~~~~~ ~ ~~ ~ ~ ~~',,,'~/'i..,~i~... -~fr~ ~X"Qt*: $ %;.:*denW E~~,,,,r ~~::Ix ~,x~~ o~'jEQ "Irl

Figure 4. Dark field electron micrograph obtained from the (002) diffraction ring. The small white dots gre the diffracting regions having 20-40 A diameters. (a) (b) Figure 5. WAXD patterns by the pinhole method. (a) 311-19, 2000~C sample, (b) 311-19, 750~C sample. -13

suggesting that larger or more ordered regions are present in the 2000~C material. The d-spacings and the sharpness of the diffraction pattern did not change with time in the electron beam suggesting no electron beam damage. The size of the selected area for diffraction ranged between 0.4 square microns to 1.2 square microns showing no change in the appearance of the pattern. Some variation in the sharpness of the halos was observed from particle to particle within each sample. The diffraction pattern for the one exceptional case in the 750~C material shown in Figure 6 consisted of spots instead of diffuse halos and showed more diffracting rings suggesting a more crystalline structure. Similar rather well defined diffracting regions have been occasionally encountered by Whittacrel~ in samples of glassy carbon. Wide angle X-ray diffraction pinhole patterns, Figure 5, agree generally with the electron diffraction patterns. The 2000~C material pattern is sharper. In addition, the (002) spacings are smaller for the 2000~C material, see Table 1. Line broadening of uncorrected diffractometer traces for the 750~C material reveal a "crystalline" dimension in the c-direction of 0 Lc = 14.1A from the (002) peak and in the a-direction of La = 18.9A from the (100) peak. See Table 1. The scanning electron micrographs agree well with the transmission electron micrographs. A surface roughness of order of.1 to 1l is observed which is better defined in the 2000~C material. (Figures 7, 8, & 9.) The low magnification micrographs -14

TABLE 1 Technique 311-19,2000~C 311-19, 750~C Bright Field E.M. 0o Platelet diameter (A) 150-500 150-350 0o Granulation diameter (A) 30-40 20-30 Dark Field E.M. o Granulation diameter (A) 20-40 Scanning E.M. o Platelet diameter (A) 300-700 300X-ray Line Broadening o (Uncorrected) La (A) 18.9 o Lc (A) 14.1 20 of (002) 23.0~ 20 of (100) 43.6~ o d-(002) (A) 3.86 d-(100) (A) 2.08 X-ray Pinhole o (Uncorrected) d-(002) (A) 3.56 3.79 o d-(100) (A) 2.17 2.19 Electron Diffraction d-(002) (A) 3.45 None d-(100) (A) 2.09 2.07 o d-(110) (A) 1.21 1.21 -15

Figure 6. Electron diffraction pattern of sample 311-19, 750~C, showing spots in diffraction rings. -16

(a) (b) Figure 7. Scanning electron micrographs (a) 311-19, 2000~C, (b) 311-19, 750~C. -17

:-.::~ ~ ~ ~ ~~~~i*#"'::........4*::!~.. Cin cnna:,:.::-~~i~~ii,:i~l~~i:::-cii:i ii~::~ii::iri:!:iiii~i:ii~ i~:i:,::::~:..::::!:!?,,,:ii:c:?:'.:::!ii::,,::i,~:.....:::~:':i i:iii; iriii' i:-:!!i::,i~? ~liiiii:!!~iia:! i!:i! iii!?!??:::!iiiir i i?:::.~':',,~~i i:i,:::::i::~~:::i~:: ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~':i"?:iii::i:.?:::.::....::;~:~~~~~~~~~~~~~~~~~~~~:...'iiii:....:: iii::.....!: Under iiting (a) (iia1:, 1-19,i:.0.0.. ( 3 (-) Figure 8. Scanning electron micrographs. (a) 311-19., 2000~C., (b) 311-19., 750~C. ~~~~~18

(a) Figure 9. Scanning electron micrographs (a) 311-19, 2000~C, (b) 311-19, 750~C. -19Figure 9. Scanning electron micrographs (a) 31119, 2000IW, (b) 311 19, 750W0. -19

reveal a different type of fracture surface showing larger steps in the 750~C material. Trhe major part of both samples appears to consist of 0 0 150-500A diameter platelets containing a 20-40A granular microstructure. These features were sharper, as expected, for the 2000~C material than for the 750~C material suggesting a better developed, but still far from perfect structure. Electron diffraction and X-ray diffraction patterns are sharper for the higher baking temperatures supporting this observation. The 0 granular microstructure 20-40A in diameter was observed in both the bright and dark field micrographs possibly suggesting the materials consist of small "crystallites" of this size. This structure could be seen in the dark field studies only in the 2000~C sample. The X-ray line broadening studies yield a structural parameter which may correspond to the granular "crystallites". The interpretation of L and L strictly as c a "crystallite" size is, however, open to question.11 Although most of the material consisted of the granulated platelets occasionally other structures were observed with electron microscopy. Electron diffraction also suggests that the materials are not completely homogeneous. Further studies will be conducted to investigate these inhomogenieties. Thermodynamics Although there are diverse methods for gaining informati about the structure of solids on various size scales, each has its drawbacks in sensitivity, interpretation, and in ease of -20

measurement when applied to carbon which is not well crystallized. As a result a thermodynamic method has been proposed. The details of the feasibility of this method have been presented elsewhere12. The thermodynamic differences between two of the crystalline forms of carbon, i.e., graphite and diamond, are relatively well known. While less well studied, there are some data available showing differences in heat of combustion13 and specific heat between graphites of different origins and between graphite and ill-crystallized carbons such as glassy carbon. 5,12 The heat of combustion data allows only an interpretation in terms of the average difference in binding energy between the two carbons. The temperature dependence of the low temperature specific heat can be interpreted in terms of structure, but not unambiguously.5 Since the reaction C C graphite glassy involves no change in composition, the enthalpy (AH) and entropy (AS) represent a two parameter measure of the structural difference of a glassy carbon relative to graphite. An accurate determination of the Gibbs free energy for the reaction over a range in temperature allows the separate evaluation of both AH and AS. The specific heat for crystalline graphite has been measured and at least one study of specific heat of glassy carbon has been made.5 With the specific heat data, a separation of both the enthalpy and entropy into vi-21

brational and configuration components can be made. This procedure thus yields a four parameter measure of the average difference between a given glassy carbon structure and that of graphite. The purely configurational enthalpy and entropy have a very straight-forward meaning as numerical measures of the difference in bonding energy due to disordering, and the degree of disordering respectively. Differences in the vibrational parts result from the effect of differences in average bond strength on the atomic vibrations which determine the specific heat. In principle it is possible to calculate the configurational entropy for various simple structural models and compare for consistency with the measured value. This procedure can rule out some models, but never establish the existance of a given model. However, whatever the real structure, the data becomes a numerical measure of the disorder relative to graphite of a given carbon. In order to be successful, the above plan must be capable of measuring AG as a function of temperature with high precision. Electrochemical cells using solid oxide electrolytes, as well as fused salt cells appear to offer the possibility of such a measurement. 12 These measurements should not be influenced by minor impurities and also will yield as a by-product more accurate and direct data on the equilibria CO(g) + 1/2 0 (g) = C2 (g) (1) -22

C aphite + C0 ( ) - 2CO( ) (') graphite + 0 (g) = CO (g) (3) and graphite + 1/2 02(g) = CO(g) (4) In spite of the importance of these reactions, such direct free energy measurements have not been carried out. All of the presently available thermodynamic information on carbon-oxygen equilibria is based solely on calorimetry combined with statistically calculated entropies for the gases. The accuracy of the above data depends on the ash content and crystallinity of the graphite and should be determined by direct measurement. A critical evaluation of the above factors and available data is presented as Appendix I. At present the cell assembly shown in Figure 10 and an associated gas purification and metering train have been constructed and operated in calibration runs. The cell experiments offer excellent opportunity to check for consistency and accuracy by determinations on well characterized materials as well as independent measurements possible with different cell confirurations. Such calibration runs have been completed on three well known oxide systems with excellent results. The niobiumoxygen system determination is in progress. Table 2 presents the results together with Figures 11, 12 and 13. The cell employed was Pt/ir (g)// Zr0.85Ca0.1501.85// Metal(s), Metal oxides/Pt Purified argon -23

Pt-10% Rh wire Pt wire To Potentiometer Pt wire 0 L. 4, p ii th vt-~c~-i I -^/-l i i-^- pt wire 1. Mullite tube 2. Solid electrolyte tube 3. Graphite electrode 4. Solid electrolyte pellet inside the tube _j ^ —-— 2) 5. S.E. pellet outside the tube 6. Glassy carbon electrode 7. Thermocouple bead e1) 8. Alumina sheath 9. Glass-ceramic joint 18t^- 10. O-ring 11. Aluminum coupling 12. Gas inlet 13. Gas outlet 14. Wax seal 15. Spring hook 16. Spring 17. Glass-ceramic joint Figure 10. 18. Vycor glass 19. Platinum ring -24

TABLE 2 Standard Free Energy Change Standard Correlation Reaction AG~ = A + B T Deviation Coefficient cal/mole cal/mole-~K cal/mole A B cal/mole cal/mole 1i ~Ni(s) + ~02(g) = NiO(s) -53,988 18.9 +50 0.9999 U, Co(s) + O2 (g) = CoO(s) -56,665 18.0 +65 0.9997 Fe(s) + ~-0 (g) = "FeO"(s) -62,646 15.1 +135 0.9987

RUN No. 1 Ni-NiO EQUILIBRIUM Pt/Air(g)/Zro85Cao15185 /Ni(S), NiO(S)/Pt 0185.85 850 ---— Charette and Flengas --- Present work... -Alcock and Belford 750, E LL 650 550I I I 600 700 800 900 1000 TEMPERATURE, ~C Figure 11. -26

RUN No. 2 CO-CoO EQUILIBRIUM Pt/Air(g)/Zr02-CaO/CO(S), CoO(S)/Pt - Kiukkola and Wagner 900 _ -o- Present Study ---- Coughlin (Calculated from'x^o ~calorimetric data) x —x Wicks and Block (Calculated from calorimetric data) 800 >, xx 700 600 I 600 700 800 900 1000 TEMPERATURE, ~C Figure 12. -27

RUN No. 3 Fe-"FeO" EQUILIBRIUM Pt/Alr(g)/ZrO2- CaO/Fe(S), "FeO"(S)/Pt 1100 Kiukkola and Wagner Charette and Flengas.... _Steele and Alcock Rapp Levitskii et. al 1000 1000~ *Present work E L. 900 W LJ 800 700 I 600 700 800 900 1000 TEMPERATURE, ~C Figure 13. -28

In addition, the first determination of the equilibrium for reaction has been made using controlled PCo /PCO in the cell, Pt/Air(g)// Zr85Ca0.1501.85// CO(g), C2 (g)/Pt The data are presented in Figure 14 and are well within the scatter band of the calorimetrically determined values. Thus even these preliminary runs indicate that it will be possible to significantly narrow the scatter band which now is only about 600 cal/mole out of 47,000 cal/mole. Further refinement in technique should yield data of even higher accuracy. In addition to the free energy data to be measured at the University of Michigan, several investigators at other laboratories have volunteered precision measurements of low temperature specific heat and heat of combustion on selected samples. B. Pore Structure The details of the pore structure are being investigated with a variety of techniques. Thus far no useful results have been obtained with small angle X-ray scattering due to a problem with interfacing the diffraction equipment with the counting electronics. After resolution of this problem, this technique will be employed on a limited number of samples to 0 measure size of pore structure in the range 20-100A where previous studies7'8 have presented interesting results on similar materials. -29

RUN No.4 CO-02-C02 EQUILIBRIUM Pt/Air(g)/ZrO2-CaO/CO(S), C02(S)/Pt Pco PCO 1100- ---— Scatter Band of Calorimetric Data ---. Present Study 1000' E 900 800Ll l > 700 800 900 1000 TEMPERATURE, ~C Figure 14. -30

Helium Pycnometry As a routine characterization of the pore volume open to He, the real density is being measured in a commercial pycnometer made by Micrometrics Inc. Corp. of Norcross, Georgia. For porous samples, data have been gathered in gross form as well as on 200 mesh powder. Thus far, samples have been produced with He densities on powdered material ranging from 1.1 gm/cm3 to 1.8 gm/cm3 which is a slightly wider range than has been previously reported (1.28 to 1.55 gm/cm3)1. An experiment to open the closed pore structure of commercial glass carbon powder by partial oxidation (7.7 wt% loss) in air failed to yield the increased density often shown with other carbon materials. There was no significant increase in density indicating that oxidation failed to open channels between the pores. Efforts are being made to produce a still wider variation in density as well as efforts to improve the accuracy of the measurements. In addition to real density, the apparent density of the bulk samples is determined geometrically. Samples have been produced with apparent densities ranging from about.5 gm/ cm3 to 1.2 gm/cm3. Surface Area Surface area measurements have been made using three techniques. Those derived from Knudsen flow permeability and Hg porosimetry have been dropped due to lack of confidence in the results and the necessity for assuming a pore model in the latter case, While the data obtained by these methods is in -31

the same general range as that from B.E.T. gas adsorption, the sensitivity required to discriminate between samples appears to be inadequate. Data on seven rather typical samples has been obtained thus far. Table 3 shows these results along with the helium densities. As expected, the lower temperature samples (700~C) showed a much higher surface area of about 500 m2/gm compared to 20-50 m2/gm for the 2000~C samples. This formation of an open pore structure in the low temperature range has been noted before1 and leads to interesting possibilities for application in surface adsorption processing6. The values shown are about double those reported for similar carbons1. However, the high temperature values are also substantially higher than previously reported. Knudsen flow derived surface area data are given where available for comparison. They are generally lower and show much less variation from sample to sample. The density data shown were determined by the Micrometrics Instrument Corp. laboratory but also were generally in agreement with those carried out at the University of Michigan. They are higher than expected, with a definitely higher trend for the lower temperature samples. This result definitely shows the existence of a fine structure permeable to He in the low temperature samples. -32

TABLE 3 Surface Area Specific Surface He Density Knudsen Flow Area Sample gm/cm3 (m2/gm) (m2/gm) 311-32, 2000~C 1.41 3.0 26.4 317-9, 700~C 1.83 506.0 319-9, 2000~C 1.70 12.5 59.9 317-12, 700~C 1.80 9.1 510.0 317-12, 2000~C 1.72 109.0 318-22, 700~C 1.79 459.0 318-22, 2000~C 1.51 49.6 -33

Mercury Porosimetry The mercury intrusion method of porosimetry is used for the characterization of continuous porosity14'15. In this investigation an Aminco* 60,000 psi mercury porosimeter is being utilized to determine cumulative pore size distribution, interconnected pore volume, density values and median pore diameter. Typical cumulative pore volume-pore diameter curves are shown in Figure 15. These data indicate that a wide range of pore size (100A to 45pi) and pore volumes can be produced in glassy carbon materials. The curves also indicate that a rather sharp pore size distribution exists in these materials. Specific pore characteristics for several samples are indicated in Table 4. In most cases there is reasonable agreement between the real densities determined with He and Hg intrusion, indicating that at 60,000 psi almost all the pores are Hg filled. 0 This pressure should fill an approximately 30A pore diameter, 0 thus supporting the finding that below 30A the pore structure is not connected, at least in the higher temperature materials. Additional data will be gathered on lower temperature material. Electron Scanning Microscopy Electron scanning microscopy is employed on a fracture surface of each sample to obtain additional information on the 0 pore structure. For the coarser materials (>200A) the pore *American Instrument Company, Silver Springs, Maryland. -34

TYPICAL MERCURY POROSIMETRY DATA FOR GLASSY CARBONS 1.2 - o. 305-12 b. 305-6 _.. —-,~ c. 311-22 u d. 312-48 1.0- e. 315-1 b W f. 312-49 PE c 0 O.8n> 0O.4 D e 0.2 100 10 0.1 0.01 0.001 PORE DIAMETER, M Figure 15.

TABLE 4 He-p ra 2 p MPD 4 IPV 5 real gPreal a app Sample (gm/cc) (g/cc) (g/cc) (p) (cc/g) 305-12, 2000~C 1.55 1.562.557 4.19 1.1560 305-6, 2000~C 1.802.636 2.54 1.0151 311-22, 2000~C 1.00.847.484.154.8809 312-48, 2000~C 1.53 1.392.861.121.4425 315-1, 2000~C 1.41 1.356.968 45.9.2944 312-49, 2000~C 1.34 1.404 1.031.011.2579 312-45, 2000~C 1.26 1.298 1.211.0046.0554 1Real density as determined by He pycnometry 2Real density as determined by Hg intrusion to 60,000 psi 3Apparent density as determined by Hg 4Median pore diameter from Hg porosimetry 5Intrusion pore volume from Hg porosimetry -36

structure is easily discernible with this method. However, for the finer materials, the size of the pores is equal to or O finer than the 100A resolving power of the instrument, rendering the determination of quantitative information impossible. Also the S.E.M. pictures show a surface roughness of the same order of size as the pore structure which makes pore size assessment difficult. In at least one case it has been possible to gain a good check on the Hg intrusion pore size by quantitative microscopy methods on S.E.M. pictures. Figure 16 shows the structure of one of the coarser structures which can most simply be described as a concatenation of anastomoses. IV. Mechanical Property Evaluation Development of testing methodology and sample preparation for characterization of the mechanical properties of glassy carbon materials is currently in progress. Mechanical properties being investigated are; hardness, compressive strength, ultimate tensile strength, and dynamic modulus of elasticity. Hardness Samples for hardness testing were prepared by wet grinding through 600 grit silicon carbide paper and then polishing with No. 1, 2 and 3 alumina (5, 3 and.05p particle size) on felt wheels. A Tukon* hardness tester equipped with Diamond Pyramid (DPH) and Knoop indenters and a Tinius Olsen** Brinell *Wilson, Mechanical Instrument Div., American Chain and Cable Company, Inc., New York, New York **Tinium Olsen Testing Model Company, Philadelphia, Pa. -37

Figure 16. S.E.M. of sample 317-39 (20000C) showing relatively coarse pore structure. -38

Tester equipped with a 1/16" ball were used for this work. A standard load cycle was utilized for the Tukon tester. With the Brinell tester the load was applied manually for a period of 30 seconds and then released. Loads ranging from 500 grams to 40 kilograms were used. A large variety of samples produced under various experimental processing conditions was tested. In addition, commercially available products, Atomergie V25, Atomergie V10, Beckwith D-50, Tokai 1000~C, Tokai 2000~C, and Lockheed-2000~C material were also tested. Initially Knoop hardness determinations were made with a 500 gram load. In most cases the indentation was nonsymmetrical or appeared as a line. In some cases the indentations disappeared completely after the load was released. These observations suggested that the materials studied exhibited a definite anelastic behavior. Other investigators have also observed this type of behavior.16 To obtain a measurable indentation with a 500 gram load it was necessary to apply a thin collodion film to the test samples prior to testing. The coating solution consisted of: 12 cc ethyl alcohol 15 cc ethyl ether 2 cc collodion Each sample was immersed in a coating solution for 5 seconds, drained and blown dry. Average Knoop hardness (500 gm load) values are shown in Table 5. -39

TABLE 5 Summary of Knoop Hardness Data Sample Designation KHN(500 g) 310-1, 1000~C 41 310-3, 1000~C 37 310-16, 1000~C 30 312-8A, 2000~C 42 312-13A, 2000~C 83 312-33, Heavy coating 51 312-33, Light coating 77 312-48, 665~C 27 312-48, 700~C 52 312-49, 700~C 103 312-50, 700~C 76 315-14A 69 315-46 150 317-5, 2000~C 50 317-6A, 2000~C 31 317-13, 2000~C 48 317-24, 2000~C 73 317-25, 2000~C 47 317-26*, 2000~C 19* 317-32, 2000~C 69 Atomergie V-25 140 Atomergie V-10 145 Beckwith D-50 246 Tokai, 1000~C 251 Tokai, 2000~C 186 Lockheed, 2000~C 180 *Impression disappeared - even when sample was coated by immersion for up to 1~ minutes. -40

No cracks or cracking sounds were observed for samples tested with tile 500 gram load. Several samples were also tested with loads of 1 and 2 kilogJrams and exhibited no observa. lo cracking phenomena. A few samples were measured on different models of hardness machines which have quite different loading cycles. The data from the faster loading cycle machine were higher by a value of about 10 to 20 in KHN. Meyer hardness values were obtained for several samples using the brinell hardness tester. Hardness values were obtained as a function of load for each sample tested. Maximum Meyer hardness values attained before the onset of cracking are shown below in Table 6. TABLE 6 Summary of Meyer Hardness Data Sample Designation H Load 315-46 199 17 kg* 317-24 122 7 kg 317-26 19 28 kg Atomergie V25, 60 162 18 kg *No cracking with loads to 36 kg. Plots of Meyer hardness versus load exhibited much scatter and the circular indentations were difficult to measure. At some -41

loads only partial indentations were observed. At higher loads cracking sounds were heard and often concentric cracks were observed in the indentation. Vickers hardness values were also measured on a group of samples as a function of load. The results are given in Table 7. Most indentations were symmetrical and easy to measure on coated samples. On one sample, 315-46, no indentations could be found on the coated sample, but could be located as an "X" shaped mark on the uncoated sample. In general there was little variation in hardness reading with load over the range of 500 grams up to the load causing the sample to break. There was a tendency to yield a value from 10 to 20 DPH lower for loads less than 1 Kg. The DPH was consistantly higher than the KHN at 500 gm and the relative order ranking was different. In view of the weird anelastic recovery displayed, more reliable data will probably result from higher loads. However, there is reason to question all of the data available to date on glassy carbon since in no case are the values reported consistent with the scratch hardness displayed by the material. The DPH values (300 maximum) are far lower than indicated by the ability to scratch glass (Mohs' > 7). There was a wide difference in hardness (DPH 68 to 312), but an even more pronounced difference of toughness was noted. At loads as small as 1 Kg some samples developed small cracks -42

TABLE 7 Summary of Vickers Hardness (DPH) Data LOAD (Kg) FOR Cracks within Cracks Specimen Sample DPH Indentation at Corners Failure Atomergie V25 174 3 40 Atomergie V10 171 5 40 Beckwith D-50 244 1 20 Tokai, 1000~C 233 1 20 Tokai, 2000~C 312 3 20 Lockheed, 2000~C 222 3 10 315-46, 2000~C* 240 - No failure to 50 312-46, 6800C 100 3 5 10 312-46, 2000~C 106 3 -20 312-45, 680~C 138 3 5 20 312-45A, 2000~C 152 3 - 10 312-33, 2000~C 68 3 3 20 *Sample uncoated. -43

or "scuff" marks within the impression. Several of the samples also developed cracks radiating from the indentation corners at somewhat higher loads. At still higher loads, all but one sample, 315-46, cracked apart. The load to such failure varied from 10 to over 50 Kg. This behavior is encouraging since it indicates qualitatively a range of crack toughness and that at least some carbons may fail gradually rather than catastrophically. Thus far, the Vickers hardness test appears most suitable for routine testing of the high-density glassy carbon materials; however, additional investigation of the effects of coating, anelastic recovery of the indentation, and cracking must be undertaken. It is believed a ball test may be more suitable for testing the less dense carbon materials. A ball type hardness testing jig for adaptation to an Instron* testing machine is presently being produced. This apparatus will allow study of the anelastic relaxation also. Compressive Strength A series of compressive strength samples was produced by machining pieces from unbaked stock. The samples were then pyrolyzed, ultrasonically cleaned and tested. Samples 6mm in diameter and 12mm in length were tested in an Instron testing machine. In most tests, a crosshead speed of 0.05 cm/min. was used. Cushion pads, blotter paper or teflon tape, were placed *Instron Corporation, Canton, Mass. -44

between the test samples and the contact blocks of the testing jig to aid in distributing the load. Average values for compressive strength data are shown in Table 8. Compressive strength values ranged from 5,100 to 48,000 psi. Several samples were tested at increased crosshead speeds. Of the samples tested, two materials showed an increase in strength and two exhibited no increase in compressive strength. More work is necessary to determine the effect on increased strain rate on compressive strength. Several samples were load cycled and exhibited relaxation and recovery behavior. Additional work is needed to characterize this effect. Additional samples are being run to evaluate variables such as specimen surface finish, size, and shape. The results shown are lower than those quoted for commercial material, however, the densities were all in the range of 1 gm/cm3 or lower compared to about 1.55 for the commercial material. Also these data must be regarded as preliminary since specimen preparation and testing procedures are not yet optimized. Tensile Strength The Diametral-Compression Test17 was used to determine the ultimate tensile strength (UTS). In this test, a right circular cylinder specimen is compressed diametrically between two flat platens. Under proper conditions induced tensile stresses cause the cylinder to fracture along the dimetral -45

TABLE 8 Average Compressive Strength Values of Glassy Carbon Materials Sample Designation Apparent Density gm/cm3 psi 317-5, 2000~C.969 33,100 317-18, 2000~C.847 5,100 317-23, 2000~C.834 7,600 317-33, 2000~C 1.017 40,200 317-37, 2000~C.922 40,600 317-38, 2000~C.940 37,600 317-39, 2000~C.889 30,700 317-40, 2000~C.883 27,100 317-41, 2000~C.934 10,000 317-41A, 2000~C.898 7,400 317-41B, 2000~C 1.118 27,000 317-43, 2000~C.895 15,000 317-44, 2000~C*.903 33,300 317-45, 2000~C*.866 32,300 48,300** 317-46, 2000~C* 1.027 28,200 46,400** 317-47, 2000~C 1.025 27,500 *Teflon tape cushion pad; other data blotter cushion pad. **Head speed: 5 cm/min. -46

plane joining the lines of contact of the specimen and the platens. Initially a series of samples with constant diameters and with diameter to thickness ratios of 1:1, 2:1, and 4:1 were produced and tested. Because of limited amounts of material on hand the preliminary strength data were obtained from samples having diameters of 12mm, 10mm and 6 mm, respectively. An Instron testing machine with a crosshead speed of.05 cm/min. was used to carry out the test. Cushion pads, blotter pad or teflon tape, were positioned between the sample and contact blocks of the testing jig. Initial trials to determine sample thickness were carried out with sample 317-37 (2000~C). Average UTS values obtained are shown in Table 9. TABLE 9 Ultimate Tensile Strength Data for Sample 317-37(2000~C) Size UTS(psi) Diameter Thickness D/t 6,910 10mm x 2.5mm 4:1 5,140 10mm x 5.0mm 2:1 4,100 10mm x 10.0mm 1:1 On the basis of these preliminary data, samples were produced from available stock materials with a D/t ratio of 4:1. Average values of samples tested thus far are shown in Table 10. -47

TABLE 10 Sunmmary of Preliminary Ultimate Tensile Strength Data Sample Apparent UTS Diameter Thickness Designation Density (si) (mm) (mm) 317-5, 2000~C.969 7,500 12 3 317-18, 2000~C.847 3,300 12 3 317-23, 2000~C.834 1,900 12 3 317-33, 2000~C 1.017 5,600 10 2.5 317-37, 2000~C.922 6,900 10 2.5 317-38, 2000~C.940 4,182 10 2.5 317-39, 2000~C.889 5,530 10 2.5 317-40, 2000~C.883 4,500 10 2.5 317-41, 2000~C.934 2,350 10 2.5 317-41A, 2000~C.898 1,910 10 2.5 317-41B, 2000~C 1.118 3,900 10 2.5 317-43, 2000~C.895 2,470 10 2.5 317-44, 2000~C.903 5,100 6 1.5 317-45, 2000~C.866 5,200 6 1.5 317-46, 2000~C 1.027 6,800 6 1.5 317-47, 2000~C 1.025 5,300 6 1.5 -48

The ultimate tensile strength values varied from 1,900 to 7,500 psi. Because of the wide range of processing variables utilized in producing each type of material it is not yet possible to correlate the many variables. An additional factor which also must be determined is the effect of sample size on UTS when the diametral compression test is used. A series of samples is currently being produced which will provide information as to effect of diameter on the UTS. The effects of pad material on strength data were also observed. Of the two different pad materials tried, the blotter paper pad (%.5mm thick) produced the best results. Improved strengths and smoother stress-strain curves were obtained when the blotter pad was used. On many, but not all samples, failure of the sample was not sudden in that the load dropped and then picked up again before additional "partial failures" occured. This behavior is thought to be associated with small cracks propagating prior to complete failure. Modulus of Elasticity Preliminary measurements of the dynamic (sonic) modulus of elasticity of several glassy carbon materials have been made with the Michigan Intermediate Frequency Electromagnetic Resonator (MIFER). This instrument can also be used for measurement of internal friction characteristics. Preliminary dynamic modulus data are shown in Table 11. Initial measurements indicate that the dynamic modulus of -49

TABLE 11 Summary of Preliminary Sonic Modulus of Elasticity Data Baking Temp. Apparent Density Modulus of Sample C gm/cm3 Elasticity 318-38 70 1.20 1.14 x 104 318-39 70 1.20 1.12 x 104 318-40 95 1.20 1.26 x 104 318-35 95 1.03 2.03 x 104 318-36 95 1.03 3.29 x 104 318-37 95 1.08 2.08 x 104 318-26 680.915 4.62 x 106 317-48 700.820 1.85 x 106 317-48 2000.885 1.66 x 106 -50

elasticity is in the range of one to four million pounds per square inch after pyrolysis. Modifications in sample geometry and refinements in instrumentation to improve the precision of the measurements are currently in progress. Quantitative internal friction data have not yet been obtained. As expected, the modulus for the precursor after setting at low temperatures is much lower, generally in the range of 10,000 to 30,000 psi. An effort will be made to follow the degree of crosslinking at low temperatures with modulus measurements. Also the modulus derived from mechanical testing will be compared to the sonic modulus. At this time no definite conclusions about the effect of structure on mechanical properties can be drawn since structural evaluations have not yet been completed on samples where property data is available. However, it is readily apparent that by altering processing conditions such as catalization extent, baking times and temperatures, a wide range of mechanical properties can be achieved. This should allow tailoring of properties to fit specific needs when the important variables are finally sorted. -51

REFERENCES 1. S. Yamada, DCIC Report 68-2, Defense Ceramic Information Center (1968). 2. E. Fitzer, K. Mueller, and W. Schaefer, Chemistry and Physics of Carbon, 7, 237, Marcel Dekker, Inc., New York (1971) 3. T. Noda and M. Inagaki, Bull. Chem. Soc. Japan, 37, 1534 (1964). 4. R. W. Lindberg, Master's Thesis, Stanford University (1970). 5. Y. Takahashi and E. F. Westrum, Jr., J. Chem. Thermodynamics, 2, 847 (1970). 6. J. L. Schmitt, Jr., Thesis at Pennsylvania State University, College of Earth and Mineral Sciences Experiment Station (1970). 7. W. S. Rothwell, J. Appl. Phys., 39, 1840 (1968). 8. R. Perret and W. Ruland, Tenth Biennial Conference on Carbon, Bethlehem, Pennsylvania, 146 (1971). 9. C. R. Schmitt, AEC Research and Development Report Y-1738, Oak Ridge Y-12 Plant, Oak Ridge, Tenn. (1970). 10. Whittacre, American Ceramic Society, 24th Pacific Coast Regional Meeting, Los Angeles, California (1971). 11. S. Ergun, Tenth Biennial Conference on Carbon, Bethlehem, Pennsylvania, 164 (1971). 12. E. E. Hucke and S. K. Das, Materials Research Council Summer Study paper, (1971). 13. J. B. Lewis, R. Murdoch, A. N. Moul, Nature, 221, 1137 (1969). 14. F. A. Dullien and F. A. Batra, I & E Chem., 62, 25 (1970). 15. H. M. Rootare, Advanced Experimental Techniques in Powder Metallurgy, 5, 225, Plenum Press, New York (1970). 16. F. C. Cowlard and J. C. Lewis, J. Mat. Sci, 2, 507 (1967). 17. A. Rudnick, A. R. Hunter and F. C. Holden, Mat. Stds., 3, (1963). -52

18. F. C. Richardson and J. II. E. Jeffes, J. Iron and Steel Institute, 160, 261 (1948). 19. M. D. Thomson, The Total and Free Energies of Formation of the Oxides of Thirty-Two Elements, Electrochemical Soc. Inc., New York (1942). 20. 0. Kubaschewski, E. L. Evans, and C. B. Alcock, Metallurgical Thermochemistry, Fourth Edition, Pergamon Press (1967). 21. R. G. Ward, An Introduction to the Physical Chemistry of Iron and Steel Making, Edward Arnold Ltd., London 1962). 22. D. R. Stull, JANAF Thermochemical Tables, Dow Chemical Co., Midland, Michigan (1965). 23. E. J. Prosen, R. S. Jessup, and F. D. Rossini, J. Res. Nat. Bur. Std., 33, 447 (1944). 24. F. D. Rossini, J. Res. Nat. Bur. Std., 22, 407 (1939). 25. D. R. Stull and H. Prophet, JANAF Thermochemical Tables, 26. K. Schwerdtfeger and E. T. Turkdogan, Techniques of Metal Research, Volume IV, Part 1, 321, Interscience Publishers (1970).27. F. D. Rossini, D. D. Wagman, and W. H. Evans, Selected Values of Chemical Thermodynamic Properties, Nat. Bur. Std. Circular No. 500, Ser. III (1952). 28. F. D. Rossini, Selected Values of the Properties of Hydrocarbons, Nat. Bur. Std., Circular No. C461 (1947). 29. J. P. Coughlin, U.S. Bur. Mines, Bull. No. 542 (1954). 30. J. F. Elliott and M. Gleiser, Thermochemistry of Steelmaking, Addison Wesley Publ. Co., Cambridge, Mass. (1960). 31. C. E. Wicks and F. E. Block, U.S. Bur. Mines, Bull. No. 605 (1963). 32. P. Hawtin, J. B. Lewis, N. Moul, and R. H. Phillips, Phil. Trans. Roy. Soc. Ser. A, 261, 67 (1966). -53

APPENDIX Critical Evaluation of the Thermodynamic Data of Carbon Oxides In one of their classical publications, Richardson and Jeffes18 compiled and graphically represented the standard Gibb's free energy of formation of all oxides pertinent to the iron and steel making. However, they assigned only an order of magnitude value to the accuracy of free energy data by classifying them as A (~1 Kcal), B (~3 Kcal), C (+10 Kcal), and D (>+10 Kcal). The standard free energy values of carbon oxides were taken from the compilation of Thompson19, who had not stated the accuracy of his calculated values. Then Kubaschewski and Evans20 and Ward21 again compiled free energy data from the same source and quoted the same order of magnitude value of the accuracy. Therefore, all the later compilations derived from Thompson's19 calculations give only the order of magnitude of error and their accuracy is anything less than 1 Kilocalorie, but the exact number has never been stated. In JANAF Thermochemical Tables, first published in 1965, Stull et al.22 compiled the thermodynamic data of carbon oxides. The AH~f 29815 value for CO2 (Reaction 3) was taken from Prosen, et al.'s23paper and was corrected for the change in the molecular weight of C02 from 44.010 to 44.011. This correction would change Prosen et al.'s23value by a factor of 1.0000227. The AH~f 298815 value for CO (Reaction 4) was computed by using AH~ 0 values of reactions 1 and 3 taken from the works of f,298. 15 -54

Rossini2'and Prosen, et al.23 respectively. Table 12 lists the AHOf 29815 values from the original sources,23'4 according to the present calculations, and from original and recent JANAF Thermochemical Tables.22,25 The values of AH~ f,298.15 (according to the present calculations) match reasonably well with those listed in the original JANAF Thermochemical Tables,22 but the accuracy limits attached to the values are in complete disagreement. The tabulation of Schwerdtfeger and Turkdogan26 is based on the original JANAF Thermochemical Tables22 and therefore they list highly exaggerated values of the uncertainty limits. However, the accuracy limits reported in recently published (June, 1971) edition of JANAF Thermochemical Tables25 have been corrected and they match extremely well with those obtained in the present calculations. Rossini27 made an elaborate compilation of chemical thermodynamic properties which for carbon oxides are essentially the same as made by Rossini et al.28 under American Petroleum Institute Research Project No. 44. Later on Coughlin29, Elliott and Gleiser30, Wicks and Block31 reported the same values. They used AH0f 298.15 values for reaction 1 and 3 with an accuracy of ~10 cal/mole (which, according to the present calculations, should be ~39.5, and +10.8 cal/mole, respectively) and calculated the values of AG~ of reactions 3 and 4 at f,T various temperatures. The listed accuracy of AG~f 2 values f,298.15 of reactions 3 and 4 are 30 and 20 cal/mole, respectively. Using equation 5, and the most accurate values of AS0 f298 1 -55

TABLE 12: Critical Evaluation of Thermodynamic Data of Carbon Oxides Reaction Of,298.15 Af,298.15AG f,100 cal/mole cal/mole cal/mcro Source Value Source Value Present Caic'a-ion 23 -94,051.8~10.8 * -94,260.6~55.8 * -94,053.9~10.8 29 -94,260.0~30.0 C(gr.)+02(g)=CO2 (g) -94,628.6:122.S 22 -94,054.0~30.0 22 -94,265.2 25 -94,054.0~11.0 25 -94,265.2 24 -67,636.1~28.7 Ii *'67,637.6~28.7 CO(g)+~02(g)=C2 (g) * -61,480.8~73.7 -46,769.6~118.7 22 -67,636.5 25 -67,638.0~30.0 24 -26,393.8~30.8 * -32,779.8~130 * -26,416.3~39.5 29 -32,808.0~20.0 C(gr. )+O02 (g)=CO(g) -47,859.0~219.5 22 -26,416.5~620 22 -32,786.0 25 -26,417.0+40.0 25 -32,786.0 24 41,242.3~58.4 C (gr.)+CO2 (g)=2CO (g) * 41,221.3~68.2 * 28,701.0+203 -1,089.4+338.2 *Present Calculation

of reactions 1 and 3 (-20.650 ~ 0.045 and 0.693 ~ 0.045 cal/mole~K, respectively) the calculated accuracy of AG~f,298.15 values of reactions 3 and 4 are +55.8 and ~129.5 cal/mole, respectively. The accuracy of AS0~ used in the present calculations incorporates f,T errors in the experimental heat capacity values as well as their numerical integration. Error in (AH~ - AHf 28 ) values fT f,298.15 are taken to be T times error in AS0~ values. f,T T 6 ^ AGO 6 AHO AC Ifp dTI+T6 ASO (5) IAGfTJ = f,298.15 { A p fT 298.15 Table 12 shows AG~ values of reactions 1-4 at 298.15 and f,T 1000~K as obtained in the present calculations, listed by Coughlin29, and original and recent JANAF Thermochemical Tables.22'25 The thermochemical calculations and the extended discussion of their probable accuracy depend heavily on accuracy of the heat of combustion of graphite (reaction 3) as reported by Prosen et al.23in 1944. When an accuracy of +10.8 cal/mole has been reported, one has to critically consider the effects of impurity content, impurity composition, pretreatment, and the kind of graphite. Experimentally it is almost impossible to determine the weight of ash present in the sample after combustion, and in calculating the results it is necessary to assume that any residue present in the crucible after combustion has the same composition as in the original graphite. This method of correcting for the weight of ash is not necessarily sound because the assumption is made that the impurities are -57

originally present in the graphite in the same form as they are found in the ash, namely, as oxides.32 It is more correct to assume that the impurities are present as carbides which oxidize during the combustion process. Thus, not only will the weight of ash differ from the original impurity content, but in addition, a significant quantity of heat will also be evolved by the oxidation of carbides. The effect of the impurity content on the heat liberated per gram of graphite burned can be demonstrated by considering the behavior of a graphite which has an ash content of 300 ppm; that is 1 gram of graphite produces 0.3 mg of ash. During the thermal pretreatment silicon and calcium present in the graphite are converted into carbides which in turn, during the combustion, undergo exothermic reactions: (i) SiC + 202 = Si02 + C02 (-AH~ = 269.1 Kcal/mole or 9578 cal/g Si); (ii) CaC2 + 5/202 = CaO + 2C02(-AH0 = 325.0 Kcal/mole f or 8109 cal/g Ca). If the concentrations of silicon and calcium impurities are both x mg/g carbon, then on combustion by stoichiometry, 2.14x mg of silica and 1.40x mg of calcium oxide will be produced for every gram of carbon. The weight of ash will be 3.54x mg. Equating this with the assumed figure of 0.3 mg gives x as 0.085. Thus the silicon and calcium content of the original graphite is 0.085 mg/g and consequently the SiC and CaC2 contents are 0.121 and 0.136 mg/g, respectively. The amount of carbon which burns as graphite is depleted by the amounts combined as carbides and -58

these are 0.036 mg/g as SiC and 0.051 mg/g as CaC2, leaving 0.999743 g/g of "free" carbon. The heat of combustion value 23 (94,051.8 cal/mole) reported by Prosen et al. can be written as: heat of combustion/g ash free graphite = observed heat release/g sample 1 - ash content/g sample Thus, the observed heat release = 7830.47(1-0.0003) = 7828.12 cal/g. Therefore, 0.121 mg of SiC and 0.136 mg of CaC2 would liberate 0.6g and 0.81 cal, respectively, and the heat liberated by the combustion of 0.999743 grm of "free" carbon in graphite is 7826.62 cal. This yields a value of 7828.63 cal/grm or 94029.7 cal/mole for the true heat of combustion of graphite. It can be seen that the apparent heat of combustion measured by Prosen et al.23is 22.1 cal/mole (94,051.8 - 94029.7) higher than the true value for a graphite assumed to contain equal Si and Ca impurity and totaling 300 ppm of ash. If the graphite sample was not prepared by heating to at least 2700~C in an inert atmosphere, the heat of combustion obtained was significantly higher (40 cals/mole).32 This result which was fully corrected for ash characteristics, shows the significant effect on heat of combustion due to difference in degree of graphitization. However, the value of heat combustion reported by Prosen et al. 23is an average of 17 measurements on Buckingham natural graphite and various artificial samples, with varying degrees of graphitization and ash characterisitcs. -59