Carbon 40 (2002) 23072315

Increased chemisorption onto activated carbon after ball-milling

N.J. Welham a , *, V. Berbenni b , P.G. Chapman c
a Department of Electronic Materials Engineering and Department of Applied Mathematics, Research School of Physical Sciences and Engineering, Australian National University, Canberra, ACT 0200, Australia b ` di Pavia, Viale Taramelli 16, 27100 Pavia, Italy CSGI Dipartimento di Chimica Fisica dell Universita c School of Applied Chemistry, Curtin University of Technology, Kent Street, Bentley, WA 6102, Australia

Received 14 July 2001; accepted 12 March 2002

Abstract Activated carbon has been milled for up to 1000 h in a laboratory-scale tumbling ball mill under a vacuum. Thermogravimetric (TG) analysis of the powder in argon showed an increasing mass loss with milling time indicating the presence of chemisorbed gas. TG Fourier transform infrared spectrometry showed the gas was a mixture of water, CO 2 and an unidentied gas (probably oxygen). BET surface area measurements showed a decreasing surface area with milling time, however, this was shown to be massively in error for the longer milling times due to the presence of the chemisorbed gas. The area occupied by the chemisorbed gas increased from 40 to 80% of the true surface area which was almost constant at 1258627 m 2 g 2 1 for all three powders. These results show that extremely large errors may be made when using BET analysis to determine the surface area of powders, especially those where the surface activity is substantially increased during processing. 2002 Elsevier Science Ltd. All rights reserved.
Keywords: A. Activated carbon; B. Grinding; C. Thermal analysis (DTA and TGA), BET surface area; D. Adsorption properties

1. Introduction Mechanical processing of powders has been studied extensively in recent years with the growing interest in nanocrystalline materials and their unusual properties. The changes in the crystallinity of graphite during milling has been examined on several occasions [18] and the general conclusion is that graphite passes through a nanocrystalline phase prior to amorphisation. Although a wide range of mechanochemical or mechanically activated reactions involving carbon have been reported [921], there has been surprisingly little study on the subsequent physical properties of the carbon itself. This is especially surprising given the results from a range of milled materials indicating enhanced reactivity after extended milling [2230] and the global industrial importance of carbon. Graphite has been shown to be substantially more
* Corresponding author. Murdoch University, Department of Mineral Science, Perth, WA 6150, Australia. Tel.: 1 61-8-93602831; fax: 1 61-8-9360-6343. E-mail address: nicholas.welham@murdoch.edu.au (N.J. Welham).

reactive towards oxygen after milling [7,8,31]. Increased intercalation of lithium ions has also been reported for milled graphite [7,32]. Several publications [7,8,3133] have measured the BET surface area of milled carbons and all report that a maximum occurs for graphite. However, thermogravimetric analysis in argon of graphite milled in vacuum for up to 1000 h showed that there was a substantial fraction of gas (up to 0.5 g gas per g graphite) adsorbed onto the surface [31]. This result brings into severe doubt the values of the surface area measured as nitrogen would only adsorb onto gas-free surfaces. Therefore, the claim that milling produces nanoporous carbon [33] based on nitrogen adsorption is clearly open to doubt without conrmatory evidence. This paper examines the effect of extended ball milling on the adsorption properties of an activated carbon.

2. Experimental The activated carbon (AC) was derived from charred and steam activated coconut shell (James Cumming, MD4050C0). Thermogravimetric analysis (TGA) in oxy-

0008-6223 / 02 / $ see front matter 2002 Elsevier Science Ltd. All rights reserved. PII: S0008-6223( 02 )00123-9

2308

N. J. Welham et al. / Carbon 40 (2002) 2307 2315

gen showed that it was 5% moisture, 90% was burned away leaving 5% ash. Samples (7.00 g) were ball milled at room temperature under vacuum |10 2 2 Pa for 100 and 1000 h in a laboratory tumbling ball mill [34]. After the allotted time the mills were opened, the powders were scraped out and placed into loosely capped containers to allow equilibration with air. TGA was performed using a Shimadzu TGA-50 at a heating rate of 20 8C min 2 1 up to 1200 8C where the sample was held for 1 h all under a 100 ml min 2 1 ow of dried argon. The mass loss at 1200 8C was modelled as a function of time and the tted curve extrapolated to estimate the total mass loss. TGA Fourier transform infrared spectrometry (FT-IR) measurements were made using a DuPont 951 TGA system at a heating rate of 20 8C min 2 1 up to about 860 8C, the maximum temperature of this instrument. The samples were purged with dried nitrogen owing at 100 ml min 2 1 for 3 h prior to heating to remove free water and reduce the CO 2 introduced into the system during loading. The nitrogen ow continued throughout the heating and swept any evolved gas into a gas cell tted within a Nicolet 730 FTIR instrument, the gas lines and cell were heated to 200 8C throughout the entire run to prevent condensation of water within the system. A lag of approximately 1 min was evident between the mass loss and the FT-IR detecting the evolved gas, this was due to the volume of the transfer system between the furnace and the gas cell. This lag has not been corrected for in the diagrams presented below. The IR spectra of the evolved gases were recorded in absorbance mode at a resolution of 8 cm 2 1 using 16 coadded interferograms per measurement. Raman spectra were obtained using a Renishaw Model 2000 instrument over the range 8002000 cm 2 1 at a resolution of 1.87 cm 2 1 . No preparation of powders was made with the as-milled powders simply imaged under the microscope, this method was found to be representative with replicate sample areas showing identical Raman response. The source was a near infrared (NIR) laser operating at 780 nm at a power of 10 mW. The data obtained were tted by mixed GaussianLorentzian peaks. Due to the extremely high infra-red absorbance of the carbon FT-IR measurements were made using an MTEC Photoacoustic Model 200 accessory mounted in a Bruker IFS 66 FTIR. The photoacoustic cell was purged using high-purity helium immediately prior to measurement and the FT-IR system was continuously purged with dried CO 2 -free air. The data was collected by coadding 256 scans at a resolution of 4 cm 2 1 over the range of 400 4000 cm 2 1 . The samples were backgrounded against C198 carbon lamp black. The surface area of the powders was determined by multipoint nitrogen adsorption using a Micromeritics Gemini system. Samples were prepared by heating up to 200 8C under vacuum for 24 h prior to measurement. The internal (micropore) area and external surface area were

determined by the Harkins and Jura method [35] and their sum was identical to the area determined separately by the BET method [36].

3. Results

3.1. Raman spectroscopy

Raman spectra of the milled and heated carbons are shown in Fig. 1. There is little change in the spectra regardless of milling time or thermal treatment indicating that the initial structure of the carbon was essentially unchanged by milling and showed no sign of graphitisation during thermal treatment. Fitting of the data was found to require four Lorentzian peaks, as shown for the as-milled 1000 h powder in Fig. 1. These peaks were at 1158621, 130468, 1502612 and 21 21 159067 cm ( 62s ). The peaks at 1304 cm and 1590 21 cm are the well known graphite D- and G-peaks which 2 relate to graphitic sp -bonding. The lower of these is not observed in single crystal graphite as it is associated with structural disordering and its position is reported to change with excitation wavelength [37]. Combining reported peak 21 positions (1355 cm when using the 514.5 nm Ar-ion 21 laser line and 1330 cm with the 632.8 nm HeNe line [37]) with the position here indicates a linear decrease in the peak position with increasing excitation wavelength: D-peak position 5 1452.7 2 0.1915l r 2 5 0.9987 The peak at 1158 cm 2 1 can be attributed to sp 3 -bonded amorphous diamond [3840], the peak at 1502 cm 2 1 does not seem to have been reported previously, but may be due to fullerene-like structures in the carbon [41,42].

Fig. 1. Raman spectra of the unmilled and milled activated carbon powders unheated and heated for 1 h at 1100 8C in argon. A t for the unheated 1000 h milled powder is shown.

N. J. Welham et al. / Carbon 40 (2002) 2307 2315

2309

FT-IR of the as-milled powders showed the presence of CO 2 and H 2 O, no other IR active phases were apparently present. This is reasonable given that the TGFT-IR only detected water and CO 2 up to 800 8C, where CO also became evident, this FT-IR result conrms that the CO was due to reaction between chemisorbed CO 2 and the carbon surface. The CO 2 peak for the 1000 h milled powder was slightly more intense that the other powders implying that it had a greater content of CO 2 which agrees with the TGA results. After heating to 1200 8C, the powders were found to resorb both water and CO 2 on exposure to atmosphere, during a subsequent reheating to 1200 8C water and CO 2 were both evolved. The trace was similar to a sample which remained in nitrogen between heatings except for | 1% water evolution below | 150 8C and | 2% more CO 2 up to 860 8C. The sudden release of CO 2 at | 760 8C was also evident in this powder indicating that CO 2 was retained on the surface up to 860 8C, the initial heating lead to a lowering of the bond energy and its release at a lower temperature during the second heating run.

3.2. TG and TG FT-IR

The mass loss of the activated carbons during TGA are shown in Fig. 2. The initial rapid mass loss of | 78% up to 150 8C was due to removal of adsorbed water. It is clear that the mass loss is somewhat greater for the milled powders, especially the 1000 h milled powder, than the unmilled powder. After the water loss, the mass loss continued during heating up to 1200 8C with the milled powders showing greater mass loss than the unmilled. The 1000 h milled sample showing a substantially greater mass loss than the 100 h powder. Holding the samples at

1200 8C showed a slow decay in the rate of mass loss with time, this was modelled using an exponential t and the estimated mass loss at innite time is shown by the squares in Fig. 2. The loss of water during the initial heating stage was conrmed during pre-purging of the TGFT-IR system where both unmilled and milled powders showed similar mass loss (| 6.2% of the initial mass) prior to heating. Despite this water loss, there still remained a small amount of water, (| 1%), which was removed only during heating up to | 150 8C. This implies that water was present in two different forms, free water and adsorbed, the former removed by purging with dry gas and the latter only removed by heating. The derivative of the TGA trace (DJG) up to 860 8C is plotted in Fig. 3 with the concomitant evolution of CO 2 . Up to | 200 8C the mass loss does not follow the same trend as the CO 2 evolution, this is not unexpected as it is known that water is the major mass loss during this temperature range. The water loss was somewhat more extensive for the 1000 h milled powder than for the shorter milling times, indicating that the 1000 h sample must have contained more adsorbed water than the other two powders after equilibration in air. The reason for this is uncertain, but may be due to a larger number of surface sites where physisorption could occur. For all samples evolution of CO 2 is clearly evident above | 100 8C. The CO 2 evolution and DTG plots follow similar trends for all three milling times, this most obvious for the 1000 h milled powder where a sudden mass loss event was mirrored by an increase in CO 2 evolution. Integration of the CO 2 evolution curve enabled comparison of the CO 2 evolution with the total mass loss during heating. These values, presented in Table 1, are normalised with respect to the unmilled powder. The increase in mass loss and CO 2 evolved is similar for

Fig. 2. TGA traces of activated carbon milled for 0, 100 and 1000 h heated at 20 8C min 2 1 to 1200 8C where the samples were held for 1 h. The squares are the estimated mass loss were the heating to be continued indenitely.

Fig. 3. DTG traces and evolution of carbon dioxide as a function of temperature.

2310

N. J. Welham et al. / Carbon 40 (2002) 2307 2315

Table 1 Ratio of mass loss and total CO 2 evolution of the powders heated to 860 8C Milling time (h) 0 100 1000 Mass 1.00 1.24 2.40 CO 2 1.00 1.30 1.75

the 100 h powder indicating that the mass loss increase was mainly due to CO 2 . For the 1000 h milled powder the mass loss increase is almost double the increase in CO 2 evolved, clearly, there is another gas evolving. FT-IR did not detect any other gas until around 800 8C where CO became evident. This is most probably due to reaction between the CO 2 and carbon which is thermodynamically favourable at . 700 8C and not due to direct evolution of CO from the surface of the solid. A chemical reaction between O 2 and an active carbon surface has been shown to occur [43] forming CO and CO 2 . In this case CO 2 is certainly released while CO may have formed but dissociated into C and CO 2 at low temperatures. It is certain that other gases were evolved during heating, the most likely candidate is oxygen since this is not detectable by IR. It could be suspected that the mass loss may have partially been due to adventitious oxygen in the nitrogen stream causing some oxidation to CO 2 . A sample of the 1000 h milled powder which was heated, allowed to cool and then heated again, all with the nitrogen owing and the results from this are shown in Fig. 4. Had adventitious oxygen been responsible for the mass loss then similar mass losses would have been expected for both runs, clearly, this is not the case. The mass loss during the rst heating run shows a peak for water, centred at | 100 8C. After this, the mass loss rate increased slightly

until | 670 8C when a rapid mass loss event was evident. These events are mirrored in the CO 2 trace with no CO 2 evolving until | 120 8C then an increased evolution until a sudden release of CO 2 occurred. This rapid evolution of CO 2 was found to be endothermic, had it been due to oxidation of carbon then the peak would have been exothermic (for C 1 O 2 5 CO 2 DH 5 2 395 kJ at 700 8C). The second heating run produced a close correlation between the mass loss rate and CO 2 evolution. Unlike the rst run, there was no evidence of water during the initial heating, this is not unexpected as the nitrogen owing throughout was dried. The start of CO 2 evolution was | 200 8C with an associated mass loss, this continued until | 720 8C where a sudden mass loss and concomitant increase in CO 2 evolution were evident. Again, the CO 2 evolution was endothermic discounting oxidation. Integration of the mass loss and CO 2 evolution showed that the mass loss in the second run was 32% that in the rst run and the CO 2 evolved was 34% that of the rst run. This clearly, demonstrates that the mass loss in the second run was entirely due to CO 2 released from the sample and not due to oxidation of the surface. The mass loss was similar in both argon and nitrogen atmospheres indicating that the adsorbed gas could not have been nitrogen. Thus, it would seem that the difference in mass loss between the samples can only be due to difference in the gas content of the powders, the longer milled powders containing more gas than the shorter milled powders. The increasing mass loss with temperature would imply that there was a range of bond strengths between the gas and carbon. The presence of peaks for CO 2 evolution would suggest that there was a larger fraction of a particular bond strength than most. During the second heating, there is a mass loss with an associated peak for CO 2 evolution at | 750 8C which was completely absent after the rst run. It has been shown [44] that partial recovery of damage to graphite induced during milling is recovered by heating. During this process, strong CO 2 bonding sites are partially annealed which therefore lead to a reduction in the bonding between the CO 2 and the underlying carbon. The lowering of the bond strength would allow breaking of the bond at a lower temperature during a subsequent heating thereby releasing gas. The absence of oxygen was also conrmed by the identical mass loss up to 1100 8C of a sample run using a owing atmosphere of 5% hydrogen in argon, beyond 1100 8C the sample run in hydrogen lost mass more rapidly, presumably this was due to reaction between the hydrogen and carbon.

3.3. BET surface area measurements

Fig. 4. DTG traces and evolution of carbon dioxide as a function of temperature for 1000 h milled AC, sample heated in nitrogen, allowed to cool to room temperature then heated again.

Surface area measurements on the as-milled powders are shown in Fig. 5. It seems clear that the external surface area only decreases slightly after the initial 100 h of

N. J. Welham et al. / Carbon 40 (2002) 2307 2315

2311

Fig. 5. Internal (micropore) and external surface areas of as-milled powders, the BET surface area is the sum of these two. The chemisorbed area was calculated according to the method in the text.

milling. The internal, or micropore, area decreased slightly during the rst 100 h of milling and then fourfold between 100 and 1000 h milling. Fig. 5 shows that the initial BET surface area of the AC (external 1 internal surface areas) was very high (770 m 2 g 2 1 ), as is typical of AC, with the majority in the form of internal micropores. Scanning electron microscopic (SEM) images of the milled AC are shown in Fig. 6 and from examination of the unmilled (but gently crushed) starting material (a) it is clear that the carbon is somewhat brittle with conchoidal fracture evident in the micrograph. There are a number of smaller particles which were formed during crushing and show similar fracture marks to the large particles. Pores can easily be seen on the surface of most particles (b) and this ready access into the interior of the particles would explain the high micropore area observed. A decrease in the BET surface area (651 m 2 g 2 1 ) was evident after 100 h of milling with the pore and external surface areas decreasing approximately equally during this period. The SEM images, Fig. 6c and d, show that there were very few particles larger than 5 mm and the majority | 1 mm. A closer examination of the particles (Fig. 6d) shows that the larger particles still have relatively smooth faces and distinct edges, although there are signs that corners are becoming rounded. It is somewhat surprising that the external surface area measured decreased with milling time, especially when the particle size undergoes the substantial size reduction evident in Fig. 6. Therefore, during the initial 100 h of milling, size reduction by brittle fracturing occurred in the larger particles. The smaller particles are somewhat rounded implying that abrasion of

corners and edges is also occurring leading to the very small particles evident in Fig. 6d. It is reasonable to expect particles to break along lines of weakness, such as pores or cracks, breakage through pores would simply convert pore area to external area, thus leading to an overall increase in area due to the new surface exposed. However, the undoubted increase in the external surface area was not evident from BET. Further milling to 1000 h lead to a massive decrease in the pore area and a BET surface area of 246 m 2 g 2 1 suggesting that milling was reducing the pore density, possibly by compaction during impact. After milling for 1000 h there is a radical change in the morphology of the carbon with . 20 mm structures evident in Fig. 6e. Clearly, the longest milling time leads to agglomeration of the smaller particles into these large particles. A more detailed view of the surface of one of these clusters Fig. 6f shows that it consists of smaller and substantially more rounded particles than those present after 100 h. Evidently, the formation of the smallest particles ( , 100 nm) was greatly enhanced during this period implying that brittle fracture was minor compared with abrasion of edges and corners. The apparent increase in particle size was due to the formation of aggregates of smaller particles. Again, the increased external surface area due to particle size reduction was not observed in the BET. Clearly, the BET technique does not appear to provide a true value for the surface area of these powders with decreases in the external surface area apparent where increases are expected. The BET technique relies on the formation of a monolayer of gas on the surface of a solid [36] at low pressures. The general technique for measure-

2312

N. J. Welham et al. / Carbon 40 (2002) 2307 2315

Fig. 6. Micrographs of (a, b) unmilled; (c, d) 100 h and (e, f) 1000 h milled activated carbon.

ment is to remove the adsorbed gas (usually by a combination of heating and evacuation) and then to adsorb nitrogen onto the gasless surface. However, if there is gas on the surface that is not removed during the pretreatment step a lower BET surface area results. If it is assumed that the mass loss above 200 8C was entirely due to the presence of adsorbed CO 2 (preparation for the BET measurement was

made by heating to 200 8C), then from the mass loss above this temperature a value for the chemisorbed surface area can be estimated. Taking the area of a carbon dioxide molecule to be 0.21 nm 2 [45,46], the absorbed gas areas were derived from the mass loss shown as squares in Fig. 2 and are also shown in Fig. 5. It is clear that the sum of the BET and adsorbed areas

N. J. Welham et al. / Carbon 40 (2002) 2307 2315

2313

are similar and average at 1258627 m 2 g 2 1 ( 62s ). From these results it appears that milling did not change the surface area of the powder, it simply changed the ratio of physisorbed gas to chemisorbed gas. This is not unreasonable given that the fracturing and strain caused during milling exposes fresh, highly active surfaces which are more likely to form bonds on exposure to the atmosphere. Indeed, the enhanced reactivity of some materials is such that on opening a mill in air spontaneous combustion has occurred. The apparent rewelding of particles which is evident in Fig. 6e may be, partially at least, due to binding of the carbon by metal abraded from the stainless steel balls and mill used. Indeed, heating of the 1000 h sample up to 1100 8C in argon showed the presence of one (or more) iron carbide phases, previous work [17,18,31] has indicated that iron and graphite react during milling in a similar mill. After heating in air, hmatite was the only iron phase present. An estimate of the iron content in the powders was made by ashing the carbon, deducting the mass of ash in the unmilled powder and assuming the residue was hmatite (69.94 wt.% Fe). The mass % of iron in the powder is shown in Table 2, clearly, the longer milling times contain more iron, as would be expected. Leaching of the 1000 h milled carbon in hydrochloric acid resulted in a mass loss of 1.9%, substantially less than the estimated iron content. This suggests that the iron in the carbon was either in an unleachable form (probably carbide) or was occluded within a carbon shell which prevented access by the acid. Work on 1000 h milled graphite indicated a mass loss of 35.7% suggesting that the iron was present in a different phase to that formed during the milling of activated carbon [31]. A similar decrease then increase in particle size has been observed in a system involving a metal and an inert powder [47]. It is interesting to note that the BET surface area decreased as the iron content increased, however, the results above indicate that iron was playing no signicant role in the surface area of the system. Of the 1000 h as-milled powder, 41.9% was adsorbed gas (extrapolated from Fig. 2), 11.4 wt.% was iron and 4.4% was ash (both from Table 2) indicating that 42.3% of the powder mass was carbon. This gives a gas loading of 0.99 g g 2 1 of carbon in the system, which at NTP is | 0.54 l CO 2 g 2 1 .

The high absorbency carbon formed by milling may prove to have applications in areas where activated carbon is currently used, such as lters. The apparently strong chemisorption obtained after milling may also allow long term storage of gas. However, from the heat cycling experiments, it was obvious that there was little chemisorption during the second exposure to air and the physisorption remained about constant. Thus, it would seem that the capacity for chemisorption is removed by annealing and remilling must be undertaken to reactivate the carbon. The extremely long milling times used here are clearly not commercially viable. The mill used is extremely inefcient and other devices, such as vibratory or planetary mills, which have a greater rate of energy transfer [4850] could lead to the same gas absorption behaviour for a shorter milling time. Modelling of tumbling and vibratory mills indicates that vibratory mills may be 300 times faster [51,52] making the process potentially more viable than in the tumbling mill used here [34]. It seems clear that the BET isotherm does not apply to milled carbon, whether it has applicability to other milled systems may well depend on the reactivity of the nal powder. Highly active powders are more likely to take gas up by absorption or reaction, although it seems probable that almost all milled powders will take up some gas from the atmosphere.

4. Conclusions Milling of activated carbon leads to a signicantly enhanced capacity for chemisorption of gas which increased with milling time. BET measurement of the surface area of milled carbons appear to be increasingly misleading as the milling time lengthened as a result of increased activation of particle surfaces and enhanced chemisorption. The adsorbed gas was evolved by heating in an inert atmosphere and below 800 8C consisted of CO 2 and an unidentied gas (probably oxygen). The presence of CO 2 on the surface of the as-milled carbons was conrmed by FT-IR. The mass of gas desorbed was used to determine the surface area of the powder which was not accounted for by BET. This area was found to increase with milling time, as expected given the increasing activity of the powders with milling time, and when added to the BET value gave a true surface area that was essentially the same for the three samples examined at 1258627 m 2 g 2 1 . The BET value decreased from 770 to 246 m 2 g 2 1 during milling with the majority of the decrease being due to the apparent loss of internal porosity. These results show that use of the BET technique for milled powders is potentially very misleading, especially when the powder becomes highly chemically active towards exposure to active gas atmospheres such as air or oxygen.

Table 2 Mass of ash and derived iron content obtained by heating the carbon for 1 h at 1100 8C in air Milling time (h) 0 100 1000 Mass of ash (mg g 2 1 ) 44 55 207 Fe content (wt.%) 0.0 (assumed) 0.8 11.4

2314

N. J. Welham et al. / Carbon 40 (2002) 2307 2315 milling of FeC (2075% Fe). Nanostructured Mater 1995;6:38992. Campbell SJ, Wang GM, Calka A, Kaczmarek WA. Ball milling of Fe-75C-25formation of Fe 3 C and Fe 7 C 3 . Mater Sci Eng A 1997;226:759. Welham NJ. Room temperature reduction of scheelite (CaWO 4 ). J Mater Res 1999;14:61927. Welham NJ. A novel route to micronised tungsten carbide. J Am Inst Chem Eng 2000;46:6871. Welham NJ. Mechanical enhancement of the carbothermic formation of TiB 2 . Metall Trans A 2000;31:2839. Tkacova K, Balaz P. Reactivity of mechanically activated chalcopyrite. Int J Miner Process 1996;4445:197208. Tkacova K, Sepelak V, Stevulova N, Boldyrev VV. Structurereactivity study of mechanically activated zinc ferrite. J Solid State Chem 1996;123:1008. Balaz P. Inuence of solid state properties on ferric chloride leaching of mechanically activated galena. Hydrometallurgy 1996;40:35968. Puclin T, Kaczmarek WA, Ninham BW. Mechanochemical processing of ZrSiO 4 . Mater Chem Phys 1995;40:7381. Welham NJ, Llewellyn DJ. Mechanical enhancement of the dissolution of ilmenite. Miner Eng 1998;11:82741. Welham NJ. Enhanced dissolution of tantalite / columbite following milling. Int J Miner Process 2000;61:14554. Balaz P, Balassaova M. Thermal decomposition of mechanically activated arsenopyrite. J Thermal Anal 1994;41:11017. Balaz P, Briancin J. Reactivity of mechanically activated pyrite. Solid State Ionics 1993;635:296300. Welham NJ. Effect of extended grinding on the dissolution of a Ta / Nb concentrate. Can Metall Quart 2001;40:14354. Welham NJ, Williams JS. Extended milling of graphite and activated carbon. Carbon 1998;36:130915. Disma F, Aymard L, Dupont L, Tarascon JM. Effect of mechanical grinding on the lithium intercalation process in graphites and soft carbons. J Electrochem Soc 1996;143:395972. Chen Y, Gerald JF, Chadderton LT, Chaffron L. Nanoporous carbon produced by ball milling. Appl Phys Lett 1999;74:27824. Calka A, Radlinski AP. Universal high performance ball milling device and its application for mechanical alloying. Mater Sci Eng A 1991;134:13503. Harkins WD, Jura G. An adsorption method for the determination of the area of a solid without the assumption of a molecular area and the area occupied by N 2 molecules on the surface of solids. J Chem Phys 1943;11:4312. Brunauer S, Emmett PH, Teller E. Adsorption of gases in multimolecular layers. J Phys Chem 1938;60:30919. Endo M, Kim C, Karaki T, Tamaki T, Nishimura Y, Matthews MJ et al. Structural analysis of the B-doped mesophase pitch-based graphite bers by raman spectroscopy. Phys Rev B Condensed Matter 1998;58:89916. Merkulov VI, Lannin JS, Munro CH, Asher SA, Veerasamy VS, Milne WI. UV studies of tetrahedral bonding in diamondlike amorphous carbon. Physical Review Letters 1997;78:486972. Schwan J, Ulrich S, Batori V, Ehrhardt H, Silva SRP. Raman spectroscopy on amorphous carbon lms. J Appl Phys 1996;80:4407. Wang CZ, Ho KM. Structure, dynamics, and electronic

Acknowledgements The authors would like to thank Dr. Andrei Rode of the Laser Physics Department and Dr. Vincent Craig of the Department of Applied Mathematics, both of the Research School of Physical Sciences and Engineering at the Australian National University for their invaluable discussions on a range of matters pertaining to this paper.
[18]

[19] [20] [21]

References
[1] Niwase K, Tanaka T, Kakimoto Y, Ishihara KN, Shingu PH. Raman spectra of graphite and diamond mechanically milled with agate or stainless steel ball-mill. Mater Trans JIM 1995;36:2828. [2] Shen TD, Ge WQ, Wang KY, Quan MX, Wang JT, Wei WD et al. Structural disorder and phase transformation in graphite produced by ball milling. Nanostructured Mater 1996;7:393 9. [3] Tang J, Zhao W, Li L, Falster AU, Simmons WB, Zhou WL et al. Amorphization of graphite induced by mechanical milling and subsequent crystallization of the amorphous carbon upon heat treating. J Mater Res 1996;11:7338. [4] Zhou WL, Ikuhara Y, Zhao W, Tang J. A transmission electron microscopy study of amorphization of graphite by mechanical milling. Carbon 1995;33:117780. [5] Aladekomo JB, Bragg RH. Structural transformations induced in graphite by grinding: analysis of 002 X-ray diffraction line proles. Carbon 1990;28:897906. [6] Nakamizo M, Honda H, Inagaki M. Raman spectra of ground natural graphite. Carbon 1978;16:2813. [7] Ong TS, Yang H. Effect of atmosphere on the mechanical milling of natural graphite. Carbon 2000;38:207785. [8] Hermann H, Schubert T, Gruner W, Mattern N. Structure and chemical reactivity of ball-milled graphite. Nanostructured Mater 1997;8:21529. [9] Kasai E, Mae K, Saito F. Effect of mixed-grinding on reduction process of carbonaceous material and iron oxide composite. ISIJ Int 1995;35:144451. [10] Yang H, Nguyen G, McCormick PG. Mechanochemical reduction of CuO by graphite. Scripta Metall Mater 1995;32:6814. [11] Welham NJ. A parametric study of the mechanically activated carbothermic reduction of ilmenite. Miner Eng 1996;9:1189200. [12] Welham NJ. The effect of extended milling on minerals. CIM Bull 1997;90:648. [13] Welham NJ. Enhancement of the Becher process by ball milling. Proc Aust Inst Mining Metallurgy 1997;302:613. [14] Matteazzi P, Le Caer G. Room-temperature mechanosynthesis of carbides by grinding of elemental powders. J Am Ceram Soc 1991;74:138290. [15] Liu ZG, Guo JT, Ye LL, Li GS, Hu ZQ. Formation mechanism of TiC by mechanical alloying. Appl Phys Lett 1994;65:26668. [16] Wang GM, Millet P, Calka A, Campbell SJ. Mechanosynthesis of tungsten carbide. Mater Sci Forum 1995;179181:1838. [17] Wang GM, Campbell SJ, Calka A, Kaczmarek WA. Ball-

[22] [23]

[24]

[25] [26] [27] [28]

[29] [30] [31] [32]

[33]

[34]

[35]

[36] [37]

[38]

[39]

[40]

N. J. Welham et al. / Carbon 40 (2002) 2307 2315 properties of diamondlike amorphous carbon. Phys Rev Lett 1993;71:11847. Eklund PC, Holden JM, Jishi RA. Vibrational modes of carbon nanotubesspectroscopy and theory. Carbon 1995;33:95972. Ishigaki T, Suzuki S, Kataura H, Kratschmer WYA. Characterization of fullerenes and carbon nanoparticles generated with a laser-furnace technique. Appl Phys A Mater Sci Process 2000;70:1214. Laine NR, Vastola FJ, Walker Jr. PL. The importance of active surface area in the carbonoxygen reaction. J Phys Chem 1963;67:20304. Harker H, Horsley JB, Robson D. Active centres produced in graphite by powdering. Carbon 1971;9:19. McClellan AL, Harnsberger HF. Cross-sectional areas of molecules adsorbed on solid surfaces. J Colloid Interface Sci 1967;23:57799. Pittman CU, Jiang W, Yue ZR, Leon y Leon CA. Surface area and pore size distribution of microporous carbon bers

2315

[41]

[47] [48]

[42]

[49] [50]

[43]

[44] [45]

[51]

[52]

[46]

prepared by electrochemical oxidation. Carbon 1999;37:85 96. Welham NJ. Mechanical activation of the solid-state reaction between Al and TiO 2 . Mater Sci Eng A 1998;255:819. Heegn H. On the connection between ultrane grinding and mechanical activation of minerals. Aufbereitungs-Technik 1989;30:63542. Tkacova K. In: Mechanical activation of minerals, London: Elsevier, 1989, p. 179. Tkacova K, Heegn H, Stevulova N. Energy transfer and conversion during comminution and mechanical activation. Int J Miner Process 1993;40:1731. Maurice D, Courtney TH. Milling dynamics. 2. Dynamics of a Spex mill and a one-dimensional mill. Metall Mater Trans A 1996;27:19739. Maurice D, Courtney TH. Milling dynamics. 3. Integration of local and global modeling of mechanical alloying devices. Metall Mater Trans A 1996;27:19816.