The Journal of
Physical Chemistry
0 Copyright 1994 by the American Chemical Society
VOLUME 98, NUMBER 39, SEPTEMBER 29, 1994
LETTERS Laser Desorption Mass Spectrometry of Fullerene Derivatives: Laser-Induced Fragmentation and Coalescence Reactions Rainer D. Beck* and Patrick Weis Institut fur Physikalische Chemie, Universitat Karlsruhe, 76128 Karlsruhe, Germany
Andreas Hirsch and Iris Lamparth Institut fiir Organische Chemie, Universitat Tiibingen, 72076 Tubingen, Germany Received: June 15, 1994; In Final Form: August 4, 1994@
Laser desorption mass spectrometry was carried out for several fullerene derivatives (C600 and a series of 6-6 ring-bridge closed adducts of Cm with bis(ethoxycarbony1)methylenes C61(COOEt)2,&(COOEt)4, and C63(COOEt)6) to investigate laser-induced growth reactions between the desorbed species. A clear correspondence between the size distribution of fragments of the parent species and the desorption-induced reaction products indicates an efficient mechanism by which reactive fragments, produced by the desorption laser, aggregate or coalesce. The reaction mechanism and nature of the products were probed through fluence dependence studies and the detection of delayed electron emission.
Introduction Since the first preparation of c 6 0 and C ~ in O macroscopic quantities,' many fullerenes* and fullerene derivatives3 have been synthesized, isolated, and characterized with the list growing rapidly. Laser desorption mass spectrometry (LDMS) has played an important role as an analytical technique for fullerene materials, both for detecting new species and for the determination of sample composition in the development of separation schemes. Reasons for the wide use of LDMS with fullerene materials include the simple sample preparation, relatively soft desorptiodionization conditions, and the highly efficient data collection when applied in a time-of-flight (TOF) mass spectrometer with its large mass range and the ability to record complete mass spectra on each laser shot. LDMS has also revealed some unusual and interesting properties of fullerenes, such as the ability to undergo delayed i o n i ~ a t i o n ~ , ~ after laser excitation or the coalescence reactions recently @Abstractpublished in Advance ACS Absrrucfs, September 15, 1994.
reported by Whetten et in which giant, highly stable, carbon clusters are formed during laser desorption of Cm films.' Since such reactions produce new species during the laser desorption which are not present in the original sample, LDMS as an analytical technique should be used in combination with other methods such as high-performance liquid chromatography (HPLC) or nuclear magnetic resonance (NMR). In an effort to extend the application of LDMS from the relatively stable fullerenes to more labile fullerene derivatives, we applied LDMS to several samples of pure, well-characterized fullerene derivatives and studied laser-induced fragmentation and aggregation processes.
Experimental Section The TOF mass spectrometer has been described in detail previ~usly.~Briefly, the source consists of a rotating and translating stainless steel target disk which is suspended about 10 mm below the orifice of a pulsed molecular beam valve (Lasertechnics) through which Ar or He gas is expanded to cool
0022-3654/94/2098-9683$04.50/0 0 1994 American Chemical Society
Letters
9684 J. Phys. Chem., Vol. 98, No. 39, 1994 and transport the laser-desorbed species. The sample is prepared by solvent casting or by pressing pellets from powder in a simple pressing device. The desorption laser is the fourth harmonic of a Nd:YAG laser (266 nm), focused by a movable lens with 50 cm focal length to a circular spot of 200 pm diameter on the target disk. A portion of the beam is split off before it enters the vacuum chamber and directed into a joulemeter for pulse energy measurements. The ions produced by the desorption laser travel 150 mm with the carrier gas into the acceleration region of the mass spectrometer where either cations or anions are accelerated with a high-voltage pulse of suitable polarity. After 1 m of free flight distance the ions are detected by a matched pair of microchannel plates which are connected through a preamplifier-discriminator to a PC-based multichannel data collection system (EG&G). To distinguish cations formed promptly in the laser desorption (LD) step from those which are formed after a time delay At of several microseconds, we used a pulsed electric field between the desorption source and the acceleration region of the mass spectrometer. In the delayed ion mode this field was turned on before the desorption laser was fired and kept on a variable time At after the desorption laser pulse. The field deflected any ions created during At and thereby prevented them from reaching the acceleration region and being detected. With a special trigger circuit it was possible to switch between the normal TOF mode (no deflecting field) and the delayed ion mode on alternating laser shots and accumulate the respective mass spectra separately. The C600 sample was synthesized by ozonolysis of c60, separated by HPLC, and characterized by NMR, IR, and UVvis s p e c t r o ~ c o p y . A ~ ~detailed ~ description of the synthesis, separation, and structurd characterization of the C6o-ethOXycarbonylmethylene adducts has been published elsewhere.1°
Results and Discussion LD Mass Spectra of Fullerene Derivatives. Samples of several pure fullerene derivatives were analyzed by LDMS to test the suitability of LDMS for analytical work with fullerene derivatives. The samples were separated by high-performance liquid chromatography (HPLC), and purity was checked by 'H (where applicable) and 13Cnuclear magnetic resonance (NMR) which also allowed unambiguous structural determination of the molecules used in this study. Figure 1 shows low fluence (1 mJ/cm2) laser desorption (LD) cation mass spectra of the mono-, bis-, and tris-adduct species of c 6 0 with bis(ethoxycarbony1)methylene. In each case the >C(COOEt)2 groups are added across the c 6 0 double bond (6-6 bridge). The structure of the parent species is shown to the right of the respective mass spectrum. With the desorption laser fluence close to the threshold for ion detection, we find a clear peak at the mass of each of the parent species as well as several fragment signals corresponding to loss of the (exohedral) ethoxycarbonyl (COOEt) moieties to various degrees. The COOEt groups appear to be lost somewhat more easily than the methylene carbon atoms bridging a 6-6 bond in the fullerene cage, which leads to abundant fragment signals at the mass of C60+, with n = 1 for the mono-adduct, n = 1, 2 for the bis-adduct, and n = 1,2, 3 for the tris-adduct.ll Note that the fullerene cage remains essentially intact toward fragmentation to C,, n 60, at this low fluence, reflecting the much higher stability of the fullerene cage compared to the external "ligand" bonds. Aggregation Reaction in LD. With increasing fluence the degree of fragmentation of the parent species increases, and at the same time mass peaks of species heavier than the parent molecules appear. Figure 2 shows cation mass spectra at several
EicOC-CCCEI
E1OCCVCCCEl
CCCEl
800
600
1000 mass [amu]
1200
Figure 1. LD cation mass spectra (266 nm, 1 ml/cm2) of the fullerene derivatives (top to bottom) Gl(COOEt)z,G2(COOEt)d,and c63(cOOEt)6. The mass peak of the parent molecule is marked by an arrow for each spectrum. The structure of the sample molecule is shown to the right of the respective mass trace.
fluences for the mono-adduct of Cm with bis(ethoxycarbony1)methylene, C61(COOEt)2. At the lowest fluence the parent peak at 878 amu is clearly detected but already dominated by the fragments c 6 0 and c 6 1 , indicating facile loss of the COOEt groups leaving an essentially intact Cm and a c60 with an additional carbon atom either dangling from the cage or incorporated perhaps in the form of a heptagon. Increasing the laser fluence causes more fragmentation (c58, c 5 6 , ...), and a series of peaks at higher mass than the parent molecule appears, similar to the fullerene coalescence products observed for LD of c 6 0 samples6,' but with a maximum at C122 instead of c118. Because we used samples of known purity and composition, we conclude that these signals are due to species created in the laser desorption process. We observed quite generally (see below) that the size distributions of the aggregate species (heavier than the parent mass) and of the fragments (below the parent mass) are related. The aggregate masses at low fluence can be obtained from combinations of the fragments (i.e., C61 f c61 C122). Among the fragments, c 6 0 appears to be the least reactive as judged from the low relative abundance of the C121 signal. c61 is the second most abundant fragment of C61(COOEt)2 (besides Cm), and correspondingly we find a maximum at C122 in the narrow size distribution of the aggregates at low fluence. This correlation between fragment and aggregate sizes is also found for the other derivatives of Cm studied (see Figure 3) and suggests that the aggregates are formed by addition of the reactive fragments of the fullerene derivatives. Enhanced reactivity of C5sn+and C56n+relative to C60n+ has been observed previously12 and was attributed to adjacent pentagons. In contrast to Cm, no structures with isolated pentagons (IPR structures) exist for c58, c56, and certain other C, clusters with n < 70. An increased reactivity of the non-IPR structures has also been invoked to explain their absence in the fullerene soot generated by the KratschmerHuffman method. Furthermore, closed-shell C, species with n = odd should be even more reactive as they must contain rings other than hexagons or pentagons. Analogous mass spectra were recorded for several other c 6 0 derivatives, and in each case we detected adduct species with characteristic size distributions. Figure 3 shows a comparison +
Letters
J. Phys. Chem., Vol. 98,No. 39, 1994 9685
I
I
500
700
I
I
900
I
1100
I
1300
1500
m a s [amu]
Figure 2. LD cation mass spectra at different fluences (top to bottom: 1,2,5, 18, and 43 d/cm2)for Csl(COOEt)2in the mass range of the parent molecule (left) and the dimeric aggregates (right). The intensity scale is scaled to the strongest peak in each spectrum.
L A C,(COOEt),
mass in carbon units
11' '20124
Figure 3. Comparison of LD mass spectra of different fullerene derivatives and pure C~O.Sample molecules are (top to bottom) Cm0, Cm, C6i(COOEt)zrC6z(COOEt)4, and C6s(COOEt)6. Left: parent and fragment mass distribution scaled to the strongest fragment peak, which is C ~ for S the Cm sample (the parent peak is truncated in this trace). Right: aggregates and coalescence products, scaled up to the strongest peak. The spectra were recorded at a fluence were the peak aggregate signal is approximately 1/10 of the strongest fragment signal (approximately 10, 100, 5 , 2, and 6 mJ/cm2,from top to bottom). of fragment sizes and aggregate size distributions for LD of CmO, Cm, and three adducts of Cm with bis(ethoxycarbony1)methylene (C61(COOEt)2, C62(COOEt)4, and C63(COOEt)6). Note that only for the latter three species which show abundant odd-numbered fragments ((261, c63) do we detect odd-numbered aggregates (Clu, C125) with significant intensity. The correlation between the distribution of fragments and the size distribution of the coalescence products is striking and strongly suggests that the reactive fragments produced by the desorption are the source of the aggregate species. The reverse interpretation, that
the fragments are in fact due to the aggregate species present in the sample, is inconsistent with the observed fluence dependence in which the fragments appear well before the aggregates are detected (see Figure 2 ) . The aggregation reactions could occur in the samples either by thermal decomposition or phototran~formation~~ of the solid or by collisions between the reactive fragments during desorption. We favor the latter mechanism based on the following observations. (I) We find no effect of extended exposure of a sample surface to the desorption laser which argues against phototransformation or thermal decomposition of the solid sample. Even after irradiation with the highest fluence conditions in the fluence dependences shown, we recover the original mass distribution when the laser intensity is reduced to the starting value. (II) The fluence dependence in which the intensity of the adduct signal increases with increasing desorption fluence is inconsistent with a phototransformation of the film into a fullerene polymer and subsequent desorption of these species. In our experience intact desorption of weakly bonded species (see Figure 1) is only possible, if at all, at the lowest fluence. A polymer signal in the mass spectra should quickly decrease in intensity with increasing fluence due to increasing fragmentation. Furthermore, steric hindrance by the bulky COOEt moieties makes it difficult to envisage a potential polymerization of the sample. (111) Finally, at a fluence close to the threshold for cation detection, abundant desorption of neutral molecules takes place14 leading to gas phase collision between the desorbing species.15 It therefore seems likely to us that collisions between reactive fragments produced in the desorption step are the source of the aggregates detected in our experiments. The fact that no aggregates are detected which incorporate any of the COOEt moieties is also consistent with this mechanism and the idea that the species must be activated by fragmentation in order to react. With increasing fluence the size distribution of the aggregates broadens to smaller sizes but without significant intensity for odd-numbered clusters below C l ~ o(see Figure 2 and 4). Simultaneously, even-numbered fragments below Cm grow in intensity, signaling extensive fragmentation of the Cm cage. If
Letters
9686 J. Phys. Chem., Vol. 98, No. 39, 1994 '122
C60
c124
DI
DI
,n,L 900
760
11'00
mass [amu]
A
DI
D 1100
I 1300
. A. 1500
, 1700
mass [amu]
Figure 4. Comparison of total ion yield and delayed ion yield for LD of C62(COOEt)4 at different fluences (top to bottom: 2,4,6, and 20 mJ/cm2). For each fluence the delayed ion yield (DI) is shown below the total yield on the same intensity scale, indicating the fraction of cations formed by
delayed ionization. this broadening of the aggregate distribution were due mainly to reactions between the c 6 0 fragments (c5&c56, ...), we would expect to also see odd C, for n < 120 because the signals at C122, C123, and C124 show that the fragments c61 and c62 still play a significant role at higher fluence. Therefore, we conclude the C, species with IZ < 120 are mostly due to fragmentation of the larger aggregates. The C, fragments, n < 60, do not participate in the growth reactions presumably because of the different fragmentation rates for fullerenes and fullerene derivatives. The more weakly bonded fullerene derivatives are expected to fragment quickly16 at times where the density of the desorption plume is still high enough for collisions to occur. Fullerenes, on the other hand, are known to fragment exceptionally slowly even at high internal energy,17 which leads to fragmentation at times when the desorption cloud has expanded far enough into the vacuum so that no more collisions occur. Since the mass analysis occurs only after 140 ps, we detect both the fast and slow fragmentation processes. The changes in the mass spectra with increasing laser fluence from narrow distributions with even- and odd-numbered C, of nearly equal intensity to broad distributions with the same size maximum, but much fewer odd-numbered species and fragmentation only into even-numbered C,, suggest that, depending on the laser fluence, both weakly bound dimer structures at the lowest fluence (dumbbells18)and strongly bound "coalesced" fullerenes7at high fluence can be formed during laser desorption. To test this hypothesis concerning the structure of the aggregates and coalescence products, we used the delayed electron emission process to distinguish between species of different stability as described in the next paragraph. Detailed studies of the weakly bound species by surface-induced dissociation are underway and will be reported in a forthcoming publication. Delayed Electron Emission. For strongly bound molecules or clusters in which the lowest fragmentation energy is higher than or at least comparable to the ionization potential, delayed
emission of an electron due to statistical redistribution of internal vibrational energy is a possible relaxation channel. This process is thought to be the microscopic analog of thermionic electron emission from a hot metal surface.19 Delayed electron emission in microscopic systems was first studied for transition metal clusters20.21(W and Nb) and has also been observed for C 6 0 ~ 3 ~ * ~ ~ and higher f ~ l l e r e n e confirming s ~ ~ ~ ~ their exceptional stability against dissociation. We have used the delayed ionization process in an attempt to distinguish strongly bound fullerene-like species from species which have low-energy fragmentation channels that outcompete delayed ionization. Figure 4 compares the total cation signal with the signal obtained in the delayed ion mode for LDMS of C62(COOEt)4 for four different laser fluences. In the delayed ion mode a deflection field is used to block all cations formed in the first 10 ps from reaching the acceleration region. Only those ions which form between 10 and 140 ps after the desorption laser are detected. At low fluence we find C, species with odd and even n with almost equal intensity without any contribution of delayed electron emission. The absence of delayed electron emission at this fluence indicates either that the products are too cold to emit an electron on the time scale of the experiment or that fragmentation dominates delayed electron emission because of low-energy fragmentation pathways. With increasing fluence the internal energy of the desorbed species should increase to the point where delayed electron emission becomes detectable if it competes with fragmentation. Figure 4 shows that delayed electron emission is detected only for the even-numbered C, but never for any of the odd-numbered Cnclusters, the parent molecule, or its non-fullerene fragments which all quickly diminish in intensity at higher laser fluence. We conclude that all of the odd-numbered and the majority of the even-numbered C, species which are detected at low fluence are not closed single shell fullerenes but are more weakly bound with
Letters
J. Phys. Chem., Vol. 98, No. 39, 1994 9687
fragmentation pathways that outcompete delayed electron emission. The observed onset of delayed electron emission for the even-numbered C, species including (2122 and CIX suggests that the weakly bound adducts, formed for example from c61 and C,Q, can anneal to highly stable fullerene structures, capable of undergoing delayed electron emission. Our earlier findings that the coalescence products formed by high fluence desorption of Cm efficiently undergo delayed electron emission and fragment by CZ loss subsequent to heating by surface impact7 are consistent with these ideas.
Acknowledgment. The authors gratefully acknowledge valuable discussions with Manfred M. Kappes and the support through his institute as well as the contributions of Rudi Michel, Carsten Stoermer, and Gotz Brauchle in earlier experiments leading to this research.
Summary and Conclusions
References and Notes
Efficient growth reactions have been detected in the laser desorption of several fullerene derivatives. The distribution of reaction products varies for different fullerene derivatives. A correlation between the fragmentation pattern of the parent species and the distribution of coalescence products indicates that the coalescence products are formed from reactive fragments produced in the desorption step and that the fragmentation rates of the parent species influence the efficiency of the growth reactions. From the observation of delayed electron emission at higher desorption fluence only for the even-numbered C, species and from the observed size distribution, we conclude that, depending on desorption conditions, both weakly bound adducts and strongly bound fullerenes can be formed in LD. Using our conclusion that aggregation and coalescence require reactive fragments, it is tempting to speculate on the difference in coalescence efficiency between samples of pure fullerenes and of fullerene derivatives.8 For a given fluence the amount of aggregation or coalescence will largely be a matter of the time scale for the reactive fragments to be produced since the collisions quickly cease in the expanding desorption plume. At very high fluence even Cm fragments rapidly by losing CZand possibly other small even-numbered C, fragments which can be incorporated in the hot coalescence products resulting in broad, almost symmetric coalescence distributions. The true onset of coalescence for pure Cm samples may be hard to detect due to the (unknown) presence of oxide impurities in the fullerene samples which lead to an onset at lower fluence and to less symmetric distribution of products because loss of species other than C, dominates the fragmentation in this case.z4 The specificity and efficiency of the growth reactions for fullerene derivatives might be useful to produce specific giant fullerenes by generating suitable precursors through LD. The observations also indicate that LDMS, as an analytical method for unknown fullerene materials, needs to be applied with caution and together with other methods such as HPLC or NMR, since it is clearly possible to generate new species during LDMS which are not present in the original sample. However, the
(1) Kratschmer, W.; Lamb, L. D.; Fostiropoulos, K.; Huffman, D. R. Nature 1990, 347, 162. (2) Thilgen, C.; Diederich, F.; Whetten, R. L. In Buchinsterfullerenes; Billups, W. E., Ciufolini, M. A., Eds.; VCH: New York, 1993; pp 59-82. (3) Wudl, F. In Buckminsterfullerenes; Billups, W. E., Ciufolini, M. A., Eds.; VCH. New York, 1993; pp 317-334. (4) Campbell, E. E. B.; Ulmer, G.; Hertel, I. V. Phys. Rev. Lett. 1991, 67, 1986. (5) Lykke, K. R.; Wurz, P. J. Chem. Phys. 1991, 95, 7008; J. Phys. Chem. 1992, 96, 3191. (6) Yeretzian, C.; Hansen, K.; Diederich, F.; Whetten, R. L. Nature 1992, 359, 44. (7) Beck, R. D.; Weis, P.; Brauchle, G.; Kappes, M. M. J. Chem. Phys. 1994, 100, 5684. (8) Beck, R. D.; Stoermer, C.; Schulz, C.; Michel, R.; Weis, P.; Gotz, B.; Kappes, M. M. J. Chem. Phys., in press. (9) Stoermer, C.; et al. To be published. (10) Hirsch, A.; Lamparth, I.; Karfunkel, H. R. Angew. Chem. Int. Ed. Engl. 1994, 33, 431. (1 1) Thermal decomposition by partial or complete loss of the COOEt moieties would by expected for heating under equilibrium conditions. (12) Petrie, S.; Bohme, D. K. Nature 1993, 365, 426. (13) Rao,A.M.;Zhou,P.;Wang,K.;Hager,G.T.;HoIden,J.M.;Wang, Y.; Lee, W.-T.; Bi, X.-X.; Eklund, P. C.; Comett, D. S.; Duncan, M. A.; Amster, I. J. Science 1993, 259, 955. (14) Wurz, P.; Lykke, K. R.; Pellin, M. J.; Gruen, D. M.; Parker, D. H. Vacuum 1992, 43, 381. (15) Kelly, R. J. Chem. Phys. 1990, 92, 5047. (16) Yeretzian, C.; Alvarez, M. M.; DiCamillo, B.; Whetten, R. L. In Laser and Optics for Sutjace Analysis; SPIE-Procedings Vol. 1857; De Vries, M. S., Ed.; SPIE: Bellingham, WA, 1993; pp 60-70. (17) Fragmentation times of several microseconds have been observed in surface collision experiments with Cw containing about 40 eV of intemal energy. At similar impact energy adduct species of Cw are found to fragment in less than 100 ns. Beck, R. D.; St. John, P.; Alvarez, M.; Diederich, F.; Whetten, R.L. J. Phys. Chem. 1991,95,8402. Beck, R. D. To be published. (18) Matsuzawa, N.; Ata, M.; Dixon, D.; Fitzgerald, G. J. Phys. Chem. 1992, 98, 2555. (19) Klots, C. E. Chem. Phys. Lett. 1991, 186, 73. (20) Amrein, A,; Simpson, R.; Hacket, P. J. Chem. Phys. 1991,95,4663. (21) Leisner. T.: Athanassenas. K.: Kreisle. D.: Recknaeel. E.: Echt. 0. J . Chem. Phys. 1993, 90, 9670. (22) Mmvama. S.: Lee, M. Y.; Haufler, R. E.; Chai, Y.;Smalley, R. E. 2.Phys. D-1991, 19, 409. (23) Echt, 0.;Kennedy, E. W. J . Phys. Chem. 1993, 97, 7088. (24) Deng, J. P.; Ju, D. D.; Her, G. R.; Mou, C. Y.; Chen, C. J.; Lin, Y. Y.; Han, C. C. J. Phys. Chem. 1993, 97, 11575.
ability to induce growth reactions and study the pathways and kinetics makes LDMS a potentially valuable tool in studying the formation and growth of fullerenes and other carbon cluster species.
I
