- Email: [email protected]
Combustion synthesis of fullerenes
102
C O M B U S T I O N A N D F L A M E 88: 102-112 (1992)
Combustion Synthesis of Fullerenes J. T H O M A S M C K I N N O N * and W I L L I A M L. BELL TDA Research, Inc., Wheat Ridge, CO 80033
and R O B E R T M. B A R K L E Y Cooperative Institute f o r Research on Environmental Sciences, Boulder, CO 80309 We report the isolation of C6o and C7o from combustion soot that is produced in high-temperature, low-pressure premixed flat flames. A critical parameter for high fullerene yields in combustion appears to be a very high flame temperature. Equilibrium calculations indicate that low pressures are important, but the experimental evidence is not clear at this time. Combustion synthesis yields fullerenes with a C 70/C 60 ratio of about 40 %, as compared with the 12% reported for electric-arc-generated fullerenes. The overall yields from carbon are very low (ca. 0.03%) but the soot studied had been produced in flames that were in no way optimized for fullerene production.
INTRODUCTION In 1985, Harold Kroto and co-workers, while conducting experiments by vaporizing a carbon substrate with a pulsed Nd:YAG laser, discovered a new class of all-carbon molecules that have now come to be known as fullerenes, with C6o and C7o being the most abundant [1]. The first macroscopic synthesis method for fullerenes was reported by Kr~itschmer et al. using an apparatus in which carbon was vaporized in an electric arc in a helium environment below atmospheric pressure [2, 3]. The IR absorption of the carbon dust produced was fairly featureless with the exception of four sharp absorption bands at 1429, 1183, 577, and 528 cm - l . Theoretical chemists had predicted that C6o would have only four IR absorption bands due to its I n point group and had made calculations of their location (to within about 100 c m - I ) [4]. Kr~itschmer et al. recognized that their IR spectrum fit well the description of C6o. In 1987, Gerhardt et al. observed C6o ions in sooting flames of benzene and acetylene using a molecular beam/mass spectrometer technique [5].
* Present address: Department of Chemical Engineering and Petroleum Refining, Colorado School of Mines, Golden, CO 80401. To whom correspondence should be sent. 0010-2180/92/$5.00
Only the concentration of the flame ions was measured, not the total carbon cluster concentration. The flames were at 20 torr and equivalence ratios (normalized fuel-air ratio--~b) from 1.9 to 2.63. The authors concluded that carbon clusters were produced from the PAH of similar size. The C6o ions, and to a lesser extent C7o ions, were much more resistant to oxidation in the postflame zone, compared with the other high-mass species present. We have calculated the equilibrium yield of C6o from the pyrolysis and oxidation of a hydrocarbon. This has been presented elsewhere [6], but will be summarized here. The thermodynamic properties of C6o were estimated from the MNDO calculations of Newton and Stanton [7] and derived from the vibrational frequencies of Stanton and Newton [4] and statistical mechanical formulas [8]. The thermodynamic properties of the other species in the mix (PAH, olefins, acetylenes, oxygenates, etc.) were taken from standard sources [9, 10]. Figure 1 shows the mass fraction of C6o in equilibrium with hydrocarbon gases when the global C:H ratio is unity as a function of pressure and temperature; no oxygen is present in this calculation. Figure 1 demonstrates the window of thermodynamic stability of C6o predicted by the MNDO calculation between 2200 to 2600 K at a pressure of 1 arm. At lower pressures, the optimum temperature range is larger. Copyright © 1992 by The Combustion Institute Published by Elsevier Science Publishing Co., Inc.
COMBUSTION SYNTHESIS OF FULLERENES
103
0.70
0.01 atm 0
0.60
0.10 atm D
cO 0.50
0.50A atm
. m
1.00 atm
2.(X) atm
i~" 0.40 or) ¢/)
0.30
0 0
o.20 0.10
0.00
1600
1800
2000
220O
2400
Temperature (K) Below the optimum temperature, PAHs are the dominant product and above about 2600 K acetylene and polyacetylenes dominate. The presence of oxygen in the calculation suppresses the equilibrium C6o yields drastically. It should be noted that solid carbon (graphite) has been excluded from the product mix; its inclusion would totally eliminate PAH and C6o from the picture. The exclusion of graphite was based on kinetic grounds since PAH are produced in rich flames under conditions in which equilibrium would predict only solid carbon formation. EXPERIMENTAL
We have tested a number of different soot samples for the presence of fullerenes. The only sample found to contain fullerenes came from a set of premixed, low-pressure benzene/oxygen/ argon flames studied at MIT in a Ph.D. dissertation [11]. These flames varied in equivalence ratio from 4 ) = 2.2 to 0 = 2.5 and pressures from 20 to 40 torr. The argon mole fraction varied from zero to 45 %. This high-temperature, low-pressure soot sample (called HTLP soot) was collected over a number of years in the process of cleaning the combustion chamber used in the research on soot formation kinetics, so soot from
2600
Fig. 1. Computed global equilibrium yield of C6o from a 1:1 C/H element ratio mixture as a function of temperature and pressure.
all the different flames was mixed together. The dominant portion of sample, however, was from a flame at
104 sumption that the fullerenes should be more resistant to oxidation than the other components of soot since the fullerenes have no edge carbons or C - H bonds to attack. In the oxidation of graphite, it is the edge carbons that are attacked first by oxygen; basal-plane carbons are much more resistant [12]. Thus, by submitting a sample of soot to mild oxidizing conditions, we thought that the fullerenes could be separated from the background soot matrix. The oxidative isolation was attempted by holding the soot sample in a tube furnace at 400"C to allow the soot matrix to burn away, leaving the fullerenes behind. Our first oxidation experiments were done with soot samples in beakers in a muffle furnace. We learned later, in the report of Kr~itschmer et al. [3], that C6o will sublime at this temperature, so we changed the experimental configuration to a flow tube in a furnace at 400"C over which a low flow of air (--- 50 cm 3/min) was passed. The exhaust of the reactor was passed through a trap immersed in a dry ice/isopropyl alcohol bath to collect any fullerenes that may have been sublimed and convected with the air flow. The oxidation experiments typically took several days to consume a few grams of soot. A variant on this experiment was tried in which we attempted to isolate the fullerenes by sublimation alone. We used the same apparatus and procedures described for oxidation, but used a flowing stream of argon that had been passed over an oxygen trap (Oxyclear, Model DGP-250R1). This experiment was carried out for 1 week. In the solvent extraction method, soot was treated with toluene in an ultrasonic bath to remove the soluble fraction. Greist et al. have reported that solvent extraction in an ultrasonic bath is more reproducible and more rapid than the Soxhlet continuous-extraction procedure [13, 14]. Kr~itschmer et al. [3] have reported the isolation of fullerenes from arc-produced carbon dust by Soxhlet extraction with benzene. Our soot samples were held in the ultrasonicator for at least 1 h at 5 0 " - 6 0 ° C at a concentration of at least 5 mL of toluene per gram of soot. After sonication, the samples were filtered in a two-step procedure. The bulk of the solid soot was removed using a preextracted paper filter with a Biichner funnel. Filter paper was not sufficiently fine to remove all the soot, so a second filtration using preextracted chemically resistant 0.45-/~m
J . T . MCKINNON ET AL. cartridge filters (Gelman Acrodisc CR) was performed. The solids were reextracted with toluene in the ultrasonicator with the process being repeated until no visible fullerenes were present in the sample. The fullerene solutions have a dark reddish-brown color even at very low concentrations (less than 0.5 mg/mL), so visual detection of their presence in a solvent is relatively easy. We also tried extraction in dichloromethane, but it was not as effective as toluene. The IR analysis (described below) indicated the presence of a large amount of impurity in the soot extracts due to a carbonyl compound. Two methods are described below to purify the sample. Kr~itschmer et al. [3] reported using diethyl ether to remove a "ubiquitous hydrocarbon" that was present in their sample. We anticipated that the impurities in our sample would be some sort of oxygenated PAH (aldehyde or organic acid) so we expected that we would need a more powerful solvent than ether. We tried washing the dried fullerene sample with several solvents, such as methanol and cyclohexane, but found none to be effective at selectively removing the impurity. Upon reexamination of the IR spectra, we realized that the impurity was not aromatic in nature, as evidenced by the lack of aromatic C - H spectral features (discussed below). We then tried diethyl ether (DEE) and found it to work quite effectively at purifying the samples. In this method, the solvent from the soot extract was dried, leaving the fullerenes and the impurities on the bottom of a vial. A few milliliters of DEE were added and the vial was sonicated at room temperature for a few minutes. The ether was poured off and filtered. The ethersoluble material was discarded and the filtrate and the material still adhering to the vial bottom were redissolved in toluene. The ether was found to tenaciously cling to the fullerenes, as was indicated by IR analysis. To remove it, we went through a cycle of drying the solvent and redissolving the fullerenes in toluene until no traces of ether remained. We found that four of these dry-redissolve cycles were sufficient to eliminate traces of DEE. We also tried solid-phase-extraction (SPE), cyano (CN)-bonded phase columns for sample cleanup (Analytichem International Bond Elut CN column). The CN columns selectively remove polar compounds, so we reasoned that it would
COMBUSTION SYNTHESIS OF FULLERENES
105
be effective at removing the carbonyl compound more easily than the ether wash. We used cyclohexane as the mobile phase to carry the fullerenes through the column and methanol to elute the adsorbed polar compounds. This method was partially successful, but there was still an unacceptable level of impurity in the cyclohexane phase. Since the IR spectrum of C6o had been reported by several groups, IR became our primary analytical method for fullerenes. We used an IBM Instruments Model IR/30 FTIR spectrometer with a 2 c m - i resolution and averaged over 128 scans. The samples were prepared in one of two ways. For rapid checks of purity, we placed the fullerene sample in the toluene solvent on a KBr disk and allowed the solvent to air dry. This method gave us a good qualitative check, but the scattering due to crystallites on the salt disk caused a large baseline shift. For more accurate spectra, we dried about 1 mg of the fullerene sample onto about 75 mg of KBr powder and pressed a disk. Baseline shift was still a minor problem due to scattering from air bubbles in the disk (the spectra shown below have the baseline adjusted). We used mass spectrometry to confirm the IR results and as a quantitative measure of the relative amounts of different fullerenes. Following the methods described by Ajie et al. [15], we did both electron impact (El) and fast-atom bombardment (FAB) mass spectrometry. The instrument was a VG Instruments Model VG7070EQ-HF double-focusing magnetic-sector mass spectrometer. We also did a limited number of gas chromatography/mass spectrometry (GC/MS) runs. The EI spectra were done at 70 eV electron energy and a heated probe/source. Exact mass determinations were done for several isotopes at a resolution of 3500 using perfluorokerosene as a reference compound. For the FAB spectra, we used m-nitrobenzyl alcohol (NBA) as the matrix. The FAB atom was argon at 8 keV energy. Interestingly, the fullerenes did not directly dissolve in the NBA when the sample was prepared, but when the probe was removed from the instrument after several minutes of argon-atom bombardment, the fullerenes were dissolved, giving the characteristic red-brown color. The G C / M S work was done using a HewlettPackard Ultra # 2, cross-linked 5 % phenyl methylsilicone wide-bore (0.32-mm) thick-film (0.5/xm) column, 20 m long. The injector was held at
250"C and the column was started at 120"C, ramped at 10°C/min until 200"C, and then ramped at 6°C/min until 300"C where the ramp was stopped. A splitless injection was used. The column effluent went to the VG mass spectrometer and the data were stored in the computer system for analysis. Thin-layer chromatography (TLC) was used as a rapid method of checking our sample cleanup methods. We used both neutral alumina (Bakerflex A1203) and silica (Eastan 13181) TLC plates with a fluorescent indicator and a mobile phase of cyclohexane. The fullerenes had a retention factor ( R / ) of 0.71 with the impurities giving four distinct dots with Rf s of 0.0-0.61. Although we never identified the impurities, the fact that they were chromatographically separated into four peaks could be an indication that there was a small number of different materials present.
RESULTS
The extractable fractions of the soot samples are shown in Table 1 along with the fraction that was ether insoluble. The ether-insoluble portion was nearly 100% fullerenes. As can be seen, only the soot p r o d u c e d f r o m the low - p r e s s u r e benzene/oxygen/argon flame (HTLP soot) had any fullerenes present. As discussed above, this soot sample was produced from low-pressure flames that were very hot (2100 K), with about 1 ms residence time above 2000 K. The other soot samples had less well-characterized time-temperature histories, but they were all produced at lower temperatures and higher pressures than the HTLP soot. The other soots and carbon samples TABLE 1
Extractable Yields and Fullerene yields from Soot Samples Extrable
Sample
Fraction (%)
HTLP Soot Acetylene torch soot Caqrbon black (CabotCorp.) Activated charcoal Copy machinetoner Diesel bus soot Diesel forkliftsoot
1.9 2.5 ~0 ~0 0.01 0.3 0.2
Fullerene Fraction (%) 43
106
J.T. MCKINNON ET AL.
either had little or no extractable material or no indication whatsoever of fullerenes (as determined by IR). Our lower detection limit is estimated at 100 /~g of fullerenes per gram of soot. The oxidative isolation of fullerenes was unsuccessful. Although the soot slowly oxidized as was planned, leaving a grey-colored ashlike material, this material turned out to be inorganic contaminants in the soot sample rather than the fullerenes which we sought. There was no material carried over to the cold trap either. We have no explanation of why the fullerenes should oxidize so readily at 400°C. We still believe that the fullerenes should be resistant to oxidation due to the lack of edge carbons and C - H bonds. Possibly the inorganic contaminants in the soot acted as catalytic oxidation sites to enhance the oxidation rate of the carbon clusters.* In a similar form, we were unable to remove
any fullerenes from the soot sample by sublimation using an oxygen-free carrier stream. The solid soot sample from the sublimation experiment was then extracted in toluene. Interestingly, this sample that had been held at 400"C for about 200 h had no solvent-extractable fullerenes. Either our hypothesis about the thermal stability and oxidative resistance of fullerenes is incorrect, or there must be another process occurring such as catalytic degradation. Figure 2 shows an IR spectrum of the raw extract from the HTLP soot. The features at 1429, 1183, 577, and 528 cm - l due to C6o are clearly visible. The large peak at 1730 c m - t corresponds to a carbonyl compound. Although not shown in Fig. 2, there are C - H stretch absorption features in the vicinity of 3000 cm 1 These were quite surprising to us because there was no absorption in the aromatic C - H region above 3000 cm-1, only in the aliphatic region below 3000 c m - ]. The lack of aromatic material (with the exception of fullerenes, which are aromatic but contain no hydrogen) in the solvent extract is puzzling since most soot formation theories involve PAH compounds as key players
* It has subsequently been s h o w n that fullerenes are more, not less resistant to oxidation in air than c a r b o n black. [D. W. M c K e e , Carbon 2 9 : 1 0 5 7 - 1 0 5 8 (1991)].
8.1688
8.140
LU O Z < t:fl nO
0.100
flCl
6.669
9,955q
1
I
1500
I
I
I
I
I
1899
I
,
I
I
596
WAVENUMBER (CM "~) Fig. 2. Infrared s p e c t r u m o f the toluene extract o f H T L P soot. The sample is dried on a K B r disk. Resolution = 2 c m - i.
C O M B U S T I O N SYNTHESIS OF F U L L E R E N E S
107
in mass growth. Since this soot suffered a fairly severe time-temperature history ( = 1 ms above 2000 K), there may have been sufficient time for the PAH to totally decompose. We have no indication whether the carbonyl materials were formed in the flame, or were from low-temperature oxidation reactions over the several years that this sample was stored at ambient temperature in air. Figure 3 shows the IR spectrum of the fullerenes sample after the DEE wash. The four C6o peaks are clearly dominant. As shown in Fig. 3, several of the other peaks at 794, 671, 6 4 1 , 5 6 4 , and 534 c m - I can be attributed to C7o [16]. The large peaks at 458, 692, and 725 c m have not been identified and could possibly be due to C84 or C58. (Cox et al. have also reported an unassigned peak at 725 c m - l [16].) There is still some absorption due to C - H stretching, which indicates that our DEE-wash procedure was not 100% effective at removal of impurities.
Thus, the mass yield must be viewed as upper limits. (The G C / M S results put an upper limit on the impurities at about 6%, see below.) Figure 4 shows the IR spectrum of the DEE-soluble material. We see a large peak at 1730 c m - ' due to the carbonyl bond stretch, a peak at 1030 c m - ' due to the ether, and the aliphatic C - H stretch at 2930 c m - ]. The source of the peak at 1270 cm 1 was not identified. Figure 5 shows the FAB mass spectrum of the toluene extract/ether wash sample and Fig. 6 shows the EI mass spectrum of the same sample. The quantitative results are shown in Table 2. We were able to obtain EI spectra at temperatures as low as 40°C, as opposed to the temperatures of 2 5 0 " C - 4 0 0 ° C reported by Cox et al. [16]. We are not sure if this material represents solid sample that sloughed off the surface and was swept into the ionization section by the convective flow, or if this truly represents sublimed fullerenes at such a low temperature.
9.481 528
8.48
1429
8.38
577
w
o z < m 0 09 133
I].28
1183
O. 18
-8. el4
r
I
I
1588
/
i
1888
I 588
WAVENUMBER (CM "1) Fig. 3. Infrared spectrum of the DEE-washed toluene extract of HTLP soot. The sample is pressed into a KBr pellet. C6o and C7o peaks are identified. Resolution = 2 cm- ~.
108
J . T . MCKINNON ET AL.
8,1800
8. 168 -
8.14~I LI.I
o Z
8,128
rr
O cfl
8..'188
8.888
B,
8688
,
3588
,
3888
J
,
I
,
,
e
,
2588
I
I
~
,
288f1
1588
1888
SSB
WAVENUMBEA (CM") Fig. 4. Infrared spectrum of DEE soluble material from HTLP soot (i.e., the sample impurities). Resolution = 2 cm ~.
One of the most noteworthy features of these spectra is that the C7o/C6o ratio is much larger than that which has been reported for arc-produced carbon dust. FAB shows about 40% C70 and EI shows 36% C7o, as opposed to the 1%
111~1
C70 reported
by Kr~itschmer et al. [3] and 13% reported by Ajie et al. [15]. There are a number of uncertainties in these measurements, however. In the case of both MS methods, we do not know the relative ionization cross-sections of the two
190|
W.
71. $1.
70. 60.
W. 4@. 211. 111_
!
Fig. 5. Fast atom bombardment (FAB) mass spectra of fullerenes produced from HTLP soot sample. Left--C60; right--C70.
COMBUSTION SYNTHESIS OF FULLERENES
109
lllel se.
IMI Be. Be.
[L
40.
le- L .
.
7~. sip. 5lb.
41L
10_
. . . . . . . . . . .
.
,
7~
.
.
.
.
.
.
7~
.
A_
A ,
_~
. . . . . .
.
~
.
.
.
.
.
7~
83e
835
84e
845
Fig. 6. Electron impact (El) mass spectra of fullerenes produced from HTLP soot sample. Left--C6o; right--C7o.
fullerenes. In the case o f the EI method, we may be getting selective vaporization rates from the heated probe due to different vapor pressures. The C6o molecule also shows a strong tendency to form the doubly charged ion, M ++, which depletes the abundance o f the M + ion. The isotope ratios determined by F A B were different than that predicted from the abundance ratio o f 13C/12C, as can be seen from Table 2. W e always see more M + 1, M + 2, etc. ions in our signal. W e interpret this to mean that our F A B matrix is providing a proton source to form C6o H+ ions as well as more protonated species. The EI spectra are very close to the expected isotope ratios.
Table 3 shows the results of the exact-mass determination on the fullerene isotopes compared with the calculated values. This test determines the exact-mass difference between the analyzed peak and an added reference compound to within one part in 3500. The reference compound was perfluorokerosene, which had several peaks near the fullerene isotope masses. An exact mass determination resolves the different masses o f each nucleus (due to the mass defect) and thus determines the stoichiometry of the peak. Because carbon is the reference compound on which all other atomic weights are based (by definition, the atomic weight of 12C is 12.0000), we would expect ~2C6o to have a mass of 720.0000 amu. As
TABLE
2
FAB and E1 Mass Spectrometry Results
Isotope
FAB Relative Abundance
EI Relative Abundance
Expected Abundance
720 721 722 723
100 91 52 30
100 60 20 5
100 67.3 22.3 4.8
Total counts of base peak (FAB) Total counts of base peak (El) 840 841 842 843
91 100 57 32
Total counts of base peak (FAB) Total counts of base peak (EI)
3.606 x 107 1.009 × 107 100 78 29 11
100 78.6 30.4 7.7 2.761 5.719
× 10 7 × 10 6
110
J . T . MCKINNON ET AL. TABLE 3 Results of Exact Mass Determination of Fullerenes Measured Mass
Calculated Mass
12C 60
720.0003
720.0000
J2C 70
839.9954
840.0000
721.0028 722.0020 840.9911
721.0034 722.0068 841.0034
Isotope
12C 13tr~ 12~ 5913~--~ 58 ~2 12C6913C
can be seen from Table 3, this is the experimental result achieved to within three decimal places. The other isotopes of C60 and the C7o isotopes are not quite as close the calculated values because of lower signal-to-noise ratios. The G C / M S work did not isolate fullerenes because the column conditions were not optimized, but did identify many of the impurities in our sample. The results showed that we had about 30 to 50 types of PAH molecules in the sample. The largest PAHs with a nonnegligible peak area were several isomers with a mass of 252. This mass corresponds to benzopyrenes which have been previously reported to be formed in sooting flames. We did not do a PAH calibration procedure and thus do not have the response factors for the various PAHs in the sample. A very rough estimate of the upper limit on the total amount of PAH in the sample, however, can be made. We estimate that the largest mass 252 peak represents 10-100 ng and that the whole sample has about 10-20 times the amount of PAH as is in the largest 252 peak. The sample was at a concentration of about 5 m g / m L and 3 /~L were injected onto the column. The estimate of upper limit impurity is then 100 ng(in mass 252 peak)*10 / ~ mg 3 # L injected
/"
mL
=6%.
The IR spectra are consistent with this estimate in that they show a very small, but nonnegligible aromatic C - H absorption feature. DISCUSSION The "conventional" method of producing fuUerenes, that is, with an electric arc between carbon rods, is fairly well developed as of this
writing and there are reports of laboratories producing the material at rates of 10 g/day with a carbon yield of about 10% [15]. Our yield of fullerenes from fuel carbon is only of order 3 x 10 - 4 based on 3% soot yield from carbon, 2% extractable yield from soot, and approximately 50% of the extractables being fullerenes. Our flames, however, were in no way optimized for fullerene production. The equilibrium calculations give an estimate of the upper limit that may be obtained in synthesizing fullerenes in a hydrocarbon pyrolysis process. That limit is about 8% at C / H = 1.0 and a pressure of 1 atm. This limit increases as pressure decreases and the carbon elemental content increases and decreases as oxygen is added. If fullerenes become a commodity chemical, their production through combustion means rather than an arc process may be advantageous. First, with combustion, there is no need to supply expensive electricity as the heat source. Second, it is much easier to produce a pure liquid product (the fuel) than a pure solid product (the carbon rods used in the arc). Third, combustion processes should be inherently easier to scale-up than an electric-arc process. We can draw an analogy to carbon black furnaces that operate at rates of order 5000 kg/h [17]. Lastly, many of the suggested applications of fullerenes, such as for catalysts or batteries, suggest that the inclusion of a metal atom inside the cage would modify the properties in a favorable way. Synthesis of fullerenes in a combustion system may prove to be a controllable route to this structure modification by doping the burner inlet mixture. For example, the enclosure of iron in a cage could possibly be effected by doping the flame with ferrocene. The observation that fullerenes are present in some flame-produced soot samples, while totally absent in others, indicates that the spherical carbon clusters are not involved in the main channel of soot nucleation and growth. The only soot in which we found fullerenes has been produced in very hot flames with a peak temperature of 2100 K. This high temperature is in agreement with the equilibrium predictions discussed above; that is, at the experimental pressure, fullerenes are strongly favored at temperatures between 1900 and 2600 K. This is in the region of the adiabatic flame temperature for a q~ = 2.4 fuel-oxygen
COMBUSTION SYNTHESIS OF FULLERENES
111
flame (Tad for benzene/oxygen = 2711 K, for methane-oxygen T~d = 2177 K) but real flames have heat losses to burner surfaces and radiation. It may be advantageous to add energy through a microwave or radio-frequency discharge in order to attain temperatures higher than the adiabatic flame temperature. We have no indication if low pressure is required for fullerene production. In a burnerstabilized flame, the temperature and pressure are not independent quantities since the pressure of the flame determines distance between the flame front and the burner surface. This distance, in turn, has an enormous impact on the heat loss to the cooled burner surface, which affects the temperature. Varying the inlet velocity can affect this relationship, but within a fairly narrow limit of flame stability. The lack of fullerenes present in diesel soot also corresponds well with the pressure effect predicted by equilibrium. At pressures present in a diesel combustion chamber, the equilibrium yield of fullerenes from fuel pyrolysis is near zero. There are some interesting mechanistic speculations that can be made concerning soot formation and fullerene formation. Since the growth of PAH is so fast in a flame it seems unlikely that fullerenes are formed through a sequential addition of C2 units to a growing carbon network. That is, the probability of putting just the correct number of five-membered rings in just the correct locations of the network seems low. The results of Gerhardt et al. [4] indicate that the fullerenes appear to form from PAH of equivalent molecular weight. Exactly how the six-membered ring arrays (the structure of most PAH) convert to five- and six-membered ring arrays is unknown, but we can speculate that the networks can have a sort of "soap bubble" character. At low temperatures, this molecular rearrangement would never happen due to its high activation energy; too many covalent bonds must be broken and reformed. However, very large molecules at very high temperatures build up large amounts of vibration energy. It may be that the statistical likelihood of localizing enough energy in the correct bonds is high enough that rate of conversion from flat benzenoid arrays to curved fullerene arrays is high. As mentioned previously, a combustion system has an advantage in that it can be probed to elucidate these conversion mechanisms.
Arbogast et al. [18] have noted that C6o is a powerful photosensitizer for the production of singlet oxygen (~A 02) which has detrimental physiological properties. C6o was shown to be resistant to oxidation by singlet 02 (unlike large PAH) and is probably resistant to normal oxidative channels due to the entirely basal-plane nature of the carbon atoms. Thus the environmental lifetime of C6o may be very long, leading Arbogast et al. [18] to express concern about danger due to environmental C6o. Based on the results of this work, we feel that it is unlikely that fullerenes will form in natural processes such as forest fires due to the high temperatures required. However, many industrial processes, such as steel refining, use carbon in reducing atmospheres at temperatures in the C6o-optimized window, so there could be industrial fullerene emissions.
CONCLUSIONS We have demonstrated that fullerenes can be extracted from combustion soot. A critical parameter appears to be a very high combustion temperature. A low pressure may be important, but that is not clear at this time. The overall yields from carbon are very low (ca. 0.03%) but our soot samples were in no way optimized for fullerene production.
This project was partially supported by the Department o f Energy, under the Small Business Innovative Research Program, contract number DE-FGO3-90ER80999. We would like to thank Ms. A m y Schultz f or assisting in the soot extractions and Dr. Donald Cox f or providing us with a copy of his paper prior to publication.
REFERENCES 1. 2. 3. 4. 5.
Kroto, H. W., Heath, J. R., O'Brien, S. C., Curl, R. F., and Smalley, R. E., Nature 318:162-163 (1985). Kr~itschmer, W., Fostiropoulos, K., and Huffman, D. R., Chem. Phys. Lett. 170:167-170 (1990). Kr~itschmer, W., Lamb, L. D., Fostiropoulos, K., and Huffman, D. R., Nature 347:354-358 (1990). Stanton, R. E., and Newton, M. D., J. Phys. Chem. 92:2141-2145 (1988). Gerhardt, Ph., Loftier, S., and Homann, K. H., Chem. Phys. Lett 137:306-310 (1987).
112
6. 7. 8. 9. 10. ll. 12. 13.
14.
J.T. MCKINNON
McKinnon, J. T., J. Phys. Chem. (in press). Newton, M. D., and Stanton, R. E., J. Am. Chem. Soc. 108:2469-2470 (1986). Benson, S. W., Thermochemical Kinetics, Wiley, New York, 1976. Kee, R. J., Rupley, F. M., and Miller, J. A., Sandia National Laboratories Report SAND87-8215 (1987). Stein, S. E., and Fahr, A., J. Phys. Chem. 89:3714-25 (1985). McKinnon, J. T., Ph.D. thesis, Massachusetts Institute of Technology, Cambridge, MA, 1989. Walker, P. L., Taylor, R. L., and Ranish, J. M., Carbon 29:411-421 (1991). Greist, W. H., Canton, J. E., Guerin, M. R., Yeatts, L. B., and Higgins, C. E., in Polynuclear Aromatic Hydrocarbons: Chemistry and Biological Effects, Proceedings of 4th International Symposium on PAH (A. Bjorseth and A. J. Dennis, Eds.), Battelle Columbus Laboratories, 1979. Greist, W. H., Tompkins, B. A., Epler, J. L., and Rao,
15.
16.
17. 18.
ET AL.
T. K., in Polynuclear Aromatic Hydrocarbons (P. W. Jones and P. Leber, Eds.), Ann Arbor Sciences Publishers, Ann Arbor, MI, 1979. Ajie, H., Alvarez, M. M., Anz, S. J., Beck, R. D., Diederich, F., Fostiropoulos, K., Huffman, D. R., Kr~itschmer, W., Rubin, Y., Schriver, K., Sensharama, D., and Whetten, R. L., J. Phys. Chem. 94:8630-8633 (1990). Cox, D. M., Behal, S., Disko, M., Gorun, S., Greaney, M., Hsu, C. S., Kollin, E., Millar, J., Robbins, J., Robbins, W., Sherwood, R., and Tindall, P., J. Am. Chem. Soc. 113:2940-2944 (1991). Shreve, R. N., and Brink, J. A., Chemical Process Industries, McGraw-Hill, New York, 1977. Arbogast, J. W., Darmanyank, A. P., Foote, C. S., Rubin, Y., Diederich, F. N., Alvarez, M. M., Anz, S. J., and Whetten, R. L., J. Phys. Chem. 95:11-12 (1991).
Received 23 May 1991; revised 5 September 1991
C O M B U S T I O N A N D F L A M E 88: 102-112 (1992)
Combustion Synthesis of Fullerenes J. T H O M A S M C K I N N O N * and W I L L I A M L. BELL TDA Research, Inc., Wheat Ridge, CO 80033
and R O B E R T M. B A R K L E Y Cooperative Institute f o r Research on Environmental Sciences, Boulder, CO 80309 We report the isolation of C6o and C7o from combustion soot that is produced in high-temperature, low-pressure premixed flat flames. A critical parameter for high fullerene yields in combustion appears to be a very high flame temperature. Equilibrium calculations indicate that low pressures are important, but the experimental evidence is not clear at this time. Combustion synthesis yields fullerenes with a C 70/C 60 ratio of about 40 %, as compared with the 12% reported for electric-arc-generated fullerenes. The overall yields from carbon are very low (ca. 0.03%) but the soot studied had been produced in flames that were in no way optimized for fullerene production.
INTRODUCTION In 1985, Harold Kroto and co-workers, while conducting experiments by vaporizing a carbon substrate with a pulsed Nd:YAG laser, discovered a new class of all-carbon molecules that have now come to be known as fullerenes, with C6o and C7o being the most abundant [1]. The first macroscopic synthesis method for fullerenes was reported by Kr~itschmer et al. using an apparatus in which carbon was vaporized in an electric arc in a helium environment below atmospheric pressure [2, 3]. The IR absorption of the carbon dust produced was fairly featureless with the exception of four sharp absorption bands at 1429, 1183, 577, and 528 cm - l . Theoretical chemists had predicted that C6o would have only four IR absorption bands due to its I n point group and had made calculations of their location (to within about 100 c m - I ) [4]. Kr~itschmer et al. recognized that their IR spectrum fit well the description of C6o. In 1987, Gerhardt et al. observed C6o ions in sooting flames of benzene and acetylene using a molecular beam/mass spectrometer technique [5].
* Present address: Department of Chemical Engineering and Petroleum Refining, Colorado School of Mines, Golden, CO 80401. To whom correspondence should be sent. 0010-2180/92/$5.00
Only the concentration of the flame ions was measured, not the total carbon cluster concentration. The flames were at 20 torr and equivalence ratios (normalized fuel-air ratio--~b) from 1.9 to 2.63. The authors concluded that carbon clusters were produced from the PAH of similar size. The C6o ions, and to a lesser extent C7o ions, were much more resistant to oxidation in the postflame zone, compared with the other high-mass species present. We have calculated the equilibrium yield of C6o from the pyrolysis and oxidation of a hydrocarbon. This has been presented elsewhere [6], but will be summarized here. The thermodynamic properties of C6o were estimated from the MNDO calculations of Newton and Stanton [7] and derived from the vibrational frequencies of Stanton and Newton [4] and statistical mechanical formulas [8]. The thermodynamic properties of the other species in the mix (PAH, olefins, acetylenes, oxygenates, etc.) were taken from standard sources [9, 10]. Figure 1 shows the mass fraction of C6o in equilibrium with hydrocarbon gases when the global C:H ratio is unity as a function of pressure and temperature; no oxygen is present in this calculation. Figure 1 demonstrates the window of thermodynamic stability of C6o predicted by the MNDO calculation between 2200 to 2600 K at a pressure of 1 arm. At lower pressures, the optimum temperature range is larger. Copyright © 1992 by The Combustion Institute Published by Elsevier Science Publishing Co., Inc.
COMBUSTION SYNTHESIS OF FULLERENES
103
0.70
0.01 atm 0
0.60
0.10 atm D
cO 0.50
0.50A atm
. m
1.00 atm
2.(X) atm
i~" 0.40 or) ¢/)
0.30
0 0
o.20 0.10
0.00
1600
1800
2000
220O
2400
Temperature (K) Below the optimum temperature, PAHs are the dominant product and above about 2600 K acetylene and polyacetylenes dominate. The presence of oxygen in the calculation suppresses the equilibrium C6o yields drastically. It should be noted that solid carbon (graphite) has been excluded from the product mix; its inclusion would totally eliminate PAH and C6o from the picture. The exclusion of graphite was based on kinetic grounds since PAH are produced in rich flames under conditions in which equilibrium would predict only solid carbon formation. EXPERIMENTAL
We have tested a number of different soot samples for the presence of fullerenes. The only sample found to contain fullerenes came from a set of premixed, low-pressure benzene/oxygen/ argon flames studied at MIT in a Ph.D. dissertation [11]. These flames varied in equivalence ratio from 4 ) = 2.2 to 0 = 2.5 and pressures from 20 to 40 torr. The argon mole fraction varied from zero to 45 %. This high-temperature, low-pressure soot sample (called HTLP soot) was collected over a number of years in the process of cleaning the combustion chamber used in the research on soot formation kinetics, so soot from
2600
Fig. 1. Computed global equilibrium yield of C6o from a 1:1 C/H element ratio mixture as a function of temperature and pressure.
all the different flames was mixed together. The dominant portion of sample, however, was from a flame at
104 sumption that the fullerenes should be more resistant to oxidation than the other components of soot since the fullerenes have no edge carbons or C - H bonds to attack. In the oxidation of graphite, it is the edge carbons that are attacked first by oxygen; basal-plane carbons are much more resistant [12]. Thus, by submitting a sample of soot to mild oxidizing conditions, we thought that the fullerenes could be separated from the background soot matrix. The oxidative isolation was attempted by holding the soot sample in a tube furnace at 400"C to allow the soot matrix to burn away, leaving the fullerenes behind. Our first oxidation experiments were done with soot samples in beakers in a muffle furnace. We learned later, in the report of Kr~itschmer et al. [3], that C6o will sublime at this temperature, so we changed the experimental configuration to a flow tube in a furnace at 400"C over which a low flow of air (--- 50 cm 3/min) was passed. The exhaust of the reactor was passed through a trap immersed in a dry ice/isopropyl alcohol bath to collect any fullerenes that may have been sublimed and convected with the air flow. The oxidation experiments typically took several days to consume a few grams of soot. A variant on this experiment was tried in which we attempted to isolate the fullerenes by sublimation alone. We used the same apparatus and procedures described for oxidation, but used a flowing stream of argon that had been passed over an oxygen trap (Oxyclear, Model DGP-250R1). This experiment was carried out for 1 week. In the solvent extraction method, soot was treated with toluene in an ultrasonic bath to remove the soluble fraction. Greist et al. have reported that solvent extraction in an ultrasonic bath is more reproducible and more rapid than the Soxhlet continuous-extraction procedure [13, 14]. Kr~itschmer et al. [3] have reported the isolation of fullerenes from arc-produced carbon dust by Soxhlet extraction with benzene. Our soot samples were held in the ultrasonicator for at least 1 h at 5 0 " - 6 0 ° C at a concentration of at least 5 mL of toluene per gram of soot. After sonication, the samples were filtered in a two-step procedure. The bulk of the solid soot was removed using a preextracted paper filter with a Biichner funnel. Filter paper was not sufficiently fine to remove all the soot, so a second filtration using preextracted chemically resistant 0.45-/~m
J . T . MCKINNON ET AL. cartridge filters (Gelman Acrodisc CR) was performed. The solids were reextracted with toluene in the ultrasonicator with the process being repeated until no visible fullerenes were present in the sample. The fullerene solutions have a dark reddish-brown color even at very low concentrations (less than 0.5 mg/mL), so visual detection of their presence in a solvent is relatively easy. We also tried extraction in dichloromethane, but it was not as effective as toluene. The IR analysis (described below) indicated the presence of a large amount of impurity in the soot extracts due to a carbonyl compound. Two methods are described below to purify the sample. Kr~itschmer et al. [3] reported using diethyl ether to remove a "ubiquitous hydrocarbon" that was present in their sample. We anticipated that the impurities in our sample would be some sort of oxygenated PAH (aldehyde or organic acid) so we expected that we would need a more powerful solvent than ether. We tried washing the dried fullerene sample with several solvents, such as methanol and cyclohexane, but found none to be effective at selectively removing the impurity. Upon reexamination of the IR spectra, we realized that the impurity was not aromatic in nature, as evidenced by the lack of aromatic C - H spectral features (discussed below). We then tried diethyl ether (DEE) and found it to work quite effectively at purifying the samples. In this method, the solvent from the soot extract was dried, leaving the fullerenes and the impurities on the bottom of a vial. A few milliliters of DEE were added and the vial was sonicated at room temperature for a few minutes. The ether was poured off and filtered. The ethersoluble material was discarded and the filtrate and the material still adhering to the vial bottom were redissolved in toluene. The ether was found to tenaciously cling to the fullerenes, as was indicated by IR analysis. To remove it, we went through a cycle of drying the solvent and redissolving the fullerenes in toluene until no traces of ether remained. We found that four of these dry-redissolve cycles were sufficient to eliminate traces of DEE. We also tried solid-phase-extraction (SPE), cyano (CN)-bonded phase columns for sample cleanup (Analytichem International Bond Elut CN column). The CN columns selectively remove polar compounds, so we reasoned that it would
COMBUSTION SYNTHESIS OF FULLERENES
105
be effective at removing the carbonyl compound more easily than the ether wash. We used cyclohexane as the mobile phase to carry the fullerenes through the column and methanol to elute the adsorbed polar compounds. This method was partially successful, but there was still an unacceptable level of impurity in the cyclohexane phase. Since the IR spectrum of C6o had been reported by several groups, IR became our primary analytical method for fullerenes. We used an IBM Instruments Model IR/30 FTIR spectrometer with a 2 c m - i resolution and averaged over 128 scans. The samples were prepared in one of two ways. For rapid checks of purity, we placed the fullerene sample in the toluene solvent on a KBr disk and allowed the solvent to air dry. This method gave us a good qualitative check, but the scattering due to crystallites on the salt disk caused a large baseline shift. For more accurate spectra, we dried about 1 mg of the fullerene sample onto about 75 mg of KBr powder and pressed a disk. Baseline shift was still a minor problem due to scattering from air bubbles in the disk (the spectra shown below have the baseline adjusted). We used mass spectrometry to confirm the IR results and as a quantitative measure of the relative amounts of different fullerenes. Following the methods described by Ajie et al. [15], we did both electron impact (El) and fast-atom bombardment (FAB) mass spectrometry. The instrument was a VG Instruments Model VG7070EQ-HF double-focusing magnetic-sector mass spectrometer. We also did a limited number of gas chromatography/mass spectrometry (GC/MS) runs. The EI spectra were done at 70 eV electron energy and a heated probe/source. Exact mass determinations were done for several isotopes at a resolution of 3500 using perfluorokerosene as a reference compound. For the FAB spectra, we used m-nitrobenzyl alcohol (NBA) as the matrix. The FAB atom was argon at 8 keV energy. Interestingly, the fullerenes did not directly dissolve in the NBA when the sample was prepared, but when the probe was removed from the instrument after several minutes of argon-atom bombardment, the fullerenes were dissolved, giving the characteristic red-brown color. The G C / M S work was done using a HewlettPackard Ultra # 2, cross-linked 5 % phenyl methylsilicone wide-bore (0.32-mm) thick-film (0.5/xm) column, 20 m long. The injector was held at
250"C and the column was started at 120"C, ramped at 10°C/min until 200"C, and then ramped at 6°C/min until 300"C where the ramp was stopped. A splitless injection was used. The column effluent went to the VG mass spectrometer and the data were stored in the computer system for analysis. Thin-layer chromatography (TLC) was used as a rapid method of checking our sample cleanup methods. We used both neutral alumina (Bakerflex A1203) and silica (Eastan 13181) TLC plates with a fluorescent indicator and a mobile phase of cyclohexane. The fullerenes had a retention factor ( R / ) of 0.71 with the impurities giving four distinct dots with Rf s of 0.0-0.61. Although we never identified the impurities, the fact that they were chromatographically separated into four peaks could be an indication that there was a small number of different materials present.
RESULTS
The extractable fractions of the soot samples are shown in Table 1 along with the fraction that was ether insoluble. The ether-insoluble portion was nearly 100% fullerenes. As can be seen, only the soot p r o d u c e d f r o m the low - p r e s s u r e benzene/oxygen/argon flame (HTLP soot) had any fullerenes present. As discussed above, this soot sample was produced from low-pressure flames that were very hot (2100 K), with about 1 ms residence time above 2000 K. The other soot samples had less well-characterized time-temperature histories, but they were all produced at lower temperatures and higher pressures than the HTLP soot. The other soots and carbon samples TABLE 1
Extractable Yields and Fullerene yields from Soot Samples Extrable
Sample
Fraction (%)
HTLP Soot Acetylene torch soot Caqrbon black (CabotCorp.) Activated charcoal Copy machinetoner Diesel bus soot Diesel forkliftsoot
1.9 2.5 ~0 ~0 0.01 0.3 0.2
Fullerene Fraction (%) 43
106
J.T. MCKINNON ET AL.
either had little or no extractable material or no indication whatsoever of fullerenes (as determined by IR). Our lower detection limit is estimated at 100 /~g of fullerenes per gram of soot. The oxidative isolation of fullerenes was unsuccessful. Although the soot slowly oxidized as was planned, leaving a grey-colored ashlike material, this material turned out to be inorganic contaminants in the soot sample rather than the fullerenes which we sought. There was no material carried over to the cold trap either. We have no explanation of why the fullerenes should oxidize so readily at 400°C. We still believe that the fullerenes should be resistant to oxidation due to the lack of edge carbons and C - H bonds. Possibly the inorganic contaminants in the soot acted as catalytic oxidation sites to enhance the oxidation rate of the carbon clusters.* In a similar form, we were unable to remove
any fullerenes from the soot sample by sublimation using an oxygen-free carrier stream. The solid soot sample from the sublimation experiment was then extracted in toluene. Interestingly, this sample that had been held at 400"C for about 200 h had no solvent-extractable fullerenes. Either our hypothesis about the thermal stability and oxidative resistance of fullerenes is incorrect, or there must be another process occurring such as catalytic degradation. Figure 2 shows an IR spectrum of the raw extract from the HTLP soot. The features at 1429, 1183, 577, and 528 cm - l due to C6o are clearly visible. The large peak at 1730 c m - t corresponds to a carbonyl compound. Although not shown in Fig. 2, there are C - H stretch absorption features in the vicinity of 3000 cm 1 These were quite surprising to us because there was no absorption in the aromatic C - H region above 3000 cm-1, only in the aliphatic region below 3000 c m - ]. The lack of aromatic material (with the exception of fullerenes, which are aromatic but contain no hydrogen) in the solvent extract is puzzling since most soot formation theories involve PAH compounds as key players
* It has subsequently been s h o w n that fullerenes are more, not less resistant to oxidation in air than c a r b o n black. [D. W. M c K e e , Carbon 2 9 : 1 0 5 7 - 1 0 5 8 (1991)].
8.1688
8.140
LU O Z < t:fl nO
0.100
flCl
6.669
9,955q
1
I
1500
I
I
I
I
I
1899
I
,
I
I
596
WAVENUMBER (CM "~) Fig. 2. Infrared s p e c t r u m o f the toluene extract o f H T L P soot. The sample is dried on a K B r disk. Resolution = 2 c m - i.
C O M B U S T I O N SYNTHESIS OF F U L L E R E N E S
107
in mass growth. Since this soot suffered a fairly severe time-temperature history ( = 1 ms above 2000 K), there may have been sufficient time for the PAH to totally decompose. We have no indication whether the carbonyl materials were formed in the flame, or were from low-temperature oxidation reactions over the several years that this sample was stored at ambient temperature in air. Figure 3 shows the IR spectrum of the fullerenes sample after the DEE wash. The four C6o peaks are clearly dominant. As shown in Fig. 3, several of the other peaks at 794, 671, 6 4 1 , 5 6 4 , and 534 c m - I can be attributed to C7o [16]. The large peaks at 458, 692, and 725 c m have not been identified and could possibly be due to C84 or C58. (Cox et al. have also reported an unassigned peak at 725 c m - l [16].) There is still some absorption due to C - H stretching, which indicates that our DEE-wash procedure was not 100% effective at removal of impurities.
Thus, the mass yield must be viewed as upper limits. (The G C / M S results put an upper limit on the impurities at about 6%, see below.) Figure 4 shows the IR spectrum of the DEE-soluble material. We see a large peak at 1730 c m - ' due to the carbonyl bond stretch, a peak at 1030 c m - ' due to the ether, and the aliphatic C - H stretch at 2930 c m - ]. The source of the peak at 1270 cm 1 was not identified. Figure 5 shows the FAB mass spectrum of the toluene extract/ether wash sample and Fig. 6 shows the EI mass spectrum of the same sample. The quantitative results are shown in Table 2. We were able to obtain EI spectra at temperatures as low as 40°C, as opposed to the temperatures of 2 5 0 " C - 4 0 0 ° C reported by Cox et al. [16]. We are not sure if this material represents solid sample that sloughed off the surface and was swept into the ionization section by the convective flow, or if this truly represents sublimed fullerenes at such a low temperature.
9.481 528
8.48
1429
8.38
577
w
o z < m 0 09 133
I].28
1183
O. 18
-8. el4
r
I
I
1588
/
i
1888
I 588
WAVENUMBER (CM "1) Fig. 3. Infrared spectrum of the DEE-washed toluene extract of HTLP soot. The sample is pressed into a KBr pellet. C6o and C7o peaks are identified. Resolution = 2 cm- ~.
108
J . T . MCKINNON ET AL.
8,1800
8. 168 -
8.14~I LI.I
o Z
8,128
rr
O cfl
8..'188
8.888
B,
8688
,
3588
,
3888
J
,
I
,
,
e
,
2588
I
I
~
,
288f1
1588
1888
SSB
WAVENUMBEA (CM") Fig. 4. Infrared spectrum of DEE soluble material from HTLP soot (i.e., the sample impurities). Resolution = 2 cm ~.
One of the most noteworthy features of these spectra is that the C7o/C6o ratio is much larger than that which has been reported for arc-produced carbon dust. FAB shows about 40% C70 and EI shows 36% C7o, as opposed to the 1%
111~1
C70 reported
by Kr~itschmer et al. [3] and 13% reported by Ajie et al. [15]. There are a number of uncertainties in these measurements, however. In the case of both MS methods, we do not know the relative ionization cross-sections of the two
190|
W.
71. $1.
70. 60.
W. 4@. 211. 111_
!
Fig. 5. Fast atom bombardment (FAB) mass spectra of fullerenes produced from HTLP soot sample. Left--C60; right--C70.
COMBUSTION SYNTHESIS OF FULLERENES
109
lllel se.
IMI Be. Be.
[L
40.
le- L .
.
7~. sip. 5lb.
41L
10_
. . . . . . . . . . .
.
,
7~
.
.
.
.
.
.
7~
.
A_
A ,
_~
. . . . . .
.
~
.
.
.
.
.
7~
83e
835
84e
845
Fig. 6. Electron impact (El) mass spectra of fullerenes produced from HTLP soot sample. Left--C6o; right--C7o.
fullerenes. In the case o f the EI method, we may be getting selective vaporization rates from the heated probe due to different vapor pressures. The C6o molecule also shows a strong tendency to form the doubly charged ion, M ++, which depletes the abundance o f the M + ion. The isotope ratios determined by F A B were different than that predicted from the abundance ratio o f 13C/12C, as can be seen from Table 2. W e always see more M + 1, M + 2, etc. ions in our signal. W e interpret this to mean that our F A B matrix is providing a proton source to form C6o H+ ions as well as more protonated species. The EI spectra are very close to the expected isotope ratios.
Table 3 shows the results of the exact-mass determination on the fullerene isotopes compared with the calculated values. This test determines the exact-mass difference between the analyzed peak and an added reference compound to within one part in 3500. The reference compound was perfluorokerosene, which had several peaks near the fullerene isotope masses. An exact mass determination resolves the different masses o f each nucleus (due to the mass defect) and thus determines the stoichiometry of the peak. Because carbon is the reference compound on which all other atomic weights are based (by definition, the atomic weight of 12C is 12.0000), we would expect ~2C6o to have a mass of 720.0000 amu. As
TABLE
2
FAB and E1 Mass Spectrometry Results
Isotope
FAB Relative Abundance
EI Relative Abundance
Expected Abundance
720 721 722 723
100 91 52 30
100 60 20 5
100 67.3 22.3 4.8
Total counts of base peak (FAB) Total counts of base peak (El) 840 841 842 843
91 100 57 32
Total counts of base peak (FAB) Total counts of base peak (EI)
3.606 x 107 1.009 × 107 100 78 29 11
100 78.6 30.4 7.7 2.761 5.719
× 10 7 × 10 6
110
J . T . MCKINNON ET AL. TABLE 3 Results of Exact Mass Determination of Fullerenes Measured Mass
Calculated Mass
12C 60
720.0003
720.0000
J2C 70
839.9954
840.0000
721.0028 722.0020 840.9911
721.0034 722.0068 841.0034
Isotope
12C 13tr~ 12~ 5913~--~ 58 ~2 12C6913C
can be seen from Table 3, this is the experimental result achieved to within three decimal places. The other isotopes of C60 and the C7o isotopes are not quite as close the calculated values because of lower signal-to-noise ratios. The G C / M S work did not isolate fullerenes because the column conditions were not optimized, but did identify many of the impurities in our sample. The results showed that we had about 30 to 50 types of PAH molecules in the sample. The largest PAHs with a nonnegligible peak area were several isomers with a mass of 252. This mass corresponds to benzopyrenes which have been previously reported to be formed in sooting flames. We did not do a PAH calibration procedure and thus do not have the response factors for the various PAHs in the sample. A very rough estimate of the upper limit on the total amount of PAH in the sample, however, can be made. We estimate that the largest mass 252 peak represents 10-100 ng and that the whole sample has about 10-20 times the amount of PAH as is in the largest 252 peak. The sample was at a concentration of about 5 m g / m L and 3 /~L were injected onto the column. The estimate of upper limit impurity is then 100 ng(in mass 252 peak)*10 / ~ mg 3 # L injected
/"
mL
=6%.
The IR spectra are consistent with this estimate in that they show a very small, but nonnegligible aromatic C - H absorption feature. DISCUSSION The "conventional" method of producing fuUerenes, that is, with an electric arc between carbon rods, is fairly well developed as of this
writing and there are reports of laboratories producing the material at rates of 10 g/day with a carbon yield of about 10% [15]. Our yield of fullerenes from fuel carbon is only of order 3 x 10 - 4 based on 3% soot yield from carbon, 2% extractable yield from soot, and approximately 50% of the extractables being fullerenes. Our flames, however, were in no way optimized for fullerene production. The equilibrium calculations give an estimate of the upper limit that may be obtained in synthesizing fullerenes in a hydrocarbon pyrolysis process. That limit is about 8% at C / H = 1.0 and a pressure of 1 atm. This limit increases as pressure decreases and the carbon elemental content increases and decreases as oxygen is added. If fullerenes become a commodity chemical, their production through combustion means rather than an arc process may be advantageous. First, with combustion, there is no need to supply expensive electricity as the heat source. Second, it is much easier to produce a pure liquid product (the fuel) than a pure solid product (the carbon rods used in the arc). Third, combustion processes should be inherently easier to scale-up than an electric-arc process. We can draw an analogy to carbon black furnaces that operate at rates of order 5000 kg/h [17]. Lastly, many of the suggested applications of fullerenes, such as for catalysts or batteries, suggest that the inclusion of a metal atom inside the cage would modify the properties in a favorable way. Synthesis of fullerenes in a combustion system may prove to be a controllable route to this structure modification by doping the burner inlet mixture. For example, the enclosure of iron in a cage could possibly be effected by doping the flame with ferrocene. The observation that fullerenes are present in some flame-produced soot samples, while totally absent in others, indicates that the spherical carbon clusters are not involved in the main channel of soot nucleation and growth. The only soot in which we found fullerenes has been produced in very hot flames with a peak temperature of 2100 K. This high temperature is in agreement with the equilibrium predictions discussed above; that is, at the experimental pressure, fullerenes are strongly favored at temperatures between 1900 and 2600 K. This is in the region of the adiabatic flame temperature for a q~ = 2.4 fuel-oxygen
COMBUSTION SYNTHESIS OF FULLERENES
111
flame (Tad for benzene/oxygen = 2711 K, for methane-oxygen T~d = 2177 K) but real flames have heat losses to burner surfaces and radiation. It may be advantageous to add energy through a microwave or radio-frequency discharge in order to attain temperatures higher than the adiabatic flame temperature. We have no indication if low pressure is required for fullerene production. In a burnerstabilized flame, the temperature and pressure are not independent quantities since the pressure of the flame determines distance between the flame front and the burner surface. This distance, in turn, has an enormous impact on the heat loss to the cooled burner surface, which affects the temperature. Varying the inlet velocity can affect this relationship, but within a fairly narrow limit of flame stability. The lack of fullerenes present in diesel soot also corresponds well with the pressure effect predicted by equilibrium. At pressures present in a diesel combustion chamber, the equilibrium yield of fullerenes from fuel pyrolysis is near zero. There are some interesting mechanistic speculations that can be made concerning soot formation and fullerene formation. Since the growth of PAH is so fast in a flame it seems unlikely that fullerenes are formed through a sequential addition of C2 units to a growing carbon network. That is, the probability of putting just the correct number of five-membered rings in just the correct locations of the network seems low. The results of Gerhardt et al. [4] indicate that the fullerenes appear to form from PAH of equivalent molecular weight. Exactly how the six-membered ring arrays (the structure of most PAH) convert to five- and six-membered ring arrays is unknown, but we can speculate that the networks can have a sort of "soap bubble" character. At low temperatures, this molecular rearrangement would never happen due to its high activation energy; too many covalent bonds must be broken and reformed. However, very large molecules at very high temperatures build up large amounts of vibration energy. It may be that the statistical likelihood of localizing enough energy in the correct bonds is high enough that rate of conversion from flat benzenoid arrays to curved fullerene arrays is high. As mentioned previously, a combustion system has an advantage in that it can be probed to elucidate these conversion mechanisms.
Arbogast et al. [18] have noted that C6o is a powerful photosensitizer for the production of singlet oxygen (~A 02) which has detrimental physiological properties. C6o was shown to be resistant to oxidation by singlet 02 (unlike large PAH) and is probably resistant to normal oxidative channels due to the entirely basal-plane nature of the carbon atoms. Thus the environmental lifetime of C6o may be very long, leading Arbogast et al. [18] to express concern about danger due to environmental C6o. Based on the results of this work, we feel that it is unlikely that fullerenes will form in natural processes such as forest fires due to the high temperatures required. However, many industrial processes, such as steel refining, use carbon in reducing atmospheres at temperatures in the C6o-optimized window, so there could be industrial fullerene emissions.
CONCLUSIONS We have demonstrated that fullerenes can be extracted from combustion soot. A critical parameter appears to be a very high combustion temperature. A low pressure may be important, but that is not clear at this time. The overall yields from carbon are very low (ca. 0.03%) but our soot samples were in no way optimized for fullerene production.
This project was partially supported by the Department o f Energy, under the Small Business Innovative Research Program, contract number DE-FGO3-90ER80999. We would like to thank Ms. A m y Schultz f or assisting in the soot extractions and Dr. Donald Cox f or providing us with a copy of his paper prior to publication.
REFERENCES 1. 2. 3. 4. 5.
Kroto, H. W., Heath, J. R., O'Brien, S. C., Curl, R. F., and Smalley, R. E., Nature 318:162-163 (1985). Kr~itschmer, W., Fostiropoulos, K., and Huffman, D. R., Chem. Phys. Lett. 170:167-170 (1990). Kr~itschmer, W., Lamb, L. D., Fostiropoulos, K., and Huffman, D. R., Nature 347:354-358 (1990). Stanton, R. E., and Newton, M. D., J. Phys. Chem. 92:2141-2145 (1988). Gerhardt, Ph., Loftier, S., and Homann, K. H., Chem. Phys. Lett 137:306-310 (1987).
112
6. 7. 8. 9. 10. ll. 12. 13.
14.
J.T. MCKINNON
McKinnon, J. T., J. Phys. Chem. (in press). Newton, M. D., and Stanton, R. E., J. Am. Chem. Soc. 108:2469-2470 (1986). Benson, S. W., Thermochemical Kinetics, Wiley, New York, 1976. Kee, R. J., Rupley, F. M., and Miller, J. A., Sandia National Laboratories Report SAND87-8215 (1987). Stein, S. E., and Fahr, A., J. Phys. Chem. 89:3714-25 (1985). McKinnon, J. T., Ph.D. thesis, Massachusetts Institute of Technology, Cambridge, MA, 1989. Walker, P. L., Taylor, R. L., and Ranish, J. M., Carbon 29:411-421 (1991). Greist, W. H., Canton, J. E., Guerin, M. R., Yeatts, L. B., and Higgins, C. E., in Polynuclear Aromatic Hydrocarbons: Chemistry and Biological Effects, Proceedings of 4th International Symposium on PAH (A. Bjorseth and A. J. Dennis, Eds.), Battelle Columbus Laboratories, 1979. Greist, W. H., Tompkins, B. A., Epler, J. L., and Rao,
15.
16.
17. 18.
ET AL.
T. K., in Polynuclear Aromatic Hydrocarbons (P. W. Jones and P. Leber, Eds.), Ann Arbor Sciences Publishers, Ann Arbor, MI, 1979. Ajie, H., Alvarez, M. M., Anz, S. J., Beck, R. D., Diederich, F., Fostiropoulos, K., Huffman, D. R., Kr~itschmer, W., Rubin, Y., Schriver, K., Sensharama, D., and Whetten, R. L., J. Phys. Chem. 94:8630-8633 (1990). Cox, D. M., Behal, S., Disko, M., Gorun, S., Greaney, M., Hsu, C. S., Kollin, E., Millar, J., Robbins, J., Robbins, W., Sherwood, R., and Tindall, P., J. Am. Chem. Soc. 113:2940-2944 (1991). Shreve, R. N., and Brink, J. A., Chemical Process Industries, McGraw-Hill, New York, 1977. Arbogast, J. W., Darmanyank, A. P., Foote, C. S., Rubin, Y., Diederich, F. N., Alvarez, M. M., Anz, S. J., and Whetten, R. L., J. Phys. Chem. 95:11-12 (1991).
Received 23 May 1991; revised 5 September 1991