As discussed earlier, Smalley and Kroto initially believed that fullerenes were formed from graphite fragments that were ripped from the graphite target during laser irradiation (Kroto, 1985). However, this mechanism of fullerene formation has subsequently been disproven by a variety of experiments. The most compelling have involved studies of isotopic distribution of C produced from 1:1 mixtures of pure C and pure C graphite powders. The C molecules formed under these conditions contain both C and C atoms, strongly suggesting that they are produced by aggregation of carbon atoms or small carbon chains, not by curling of preformed graphite sheets (Meijer, 1990).

In the Smalley/Kroto laser vaporization experiment, a plasma of carbon vapor is generated over the irradiated spot, where temperatures over 10,000°C are attained. In order to cool this superhot plasma and generate carbon clusters, a burst of helium gas is introduced from a pulsed gas nozzle. As the carbon atoms cool, clustering reactions occur. By adjusting the relative timing between the vaporization laser and the helium gas pulses, the residence time in the source can be extended to allow these growing clusters to aggregate in the soup of carbon atoms.

The growth process that generates the fullerenes (Smalley, 1992) probably begins with small linear chains that add other linear chains and carbon atoms until they become large enough to take the form of monocyclic rings. These rings, in turn, could grow through further additions of atoms or small chains until they were in the 25-35 atom size range. Then polycyclic networks like the open graphite sheets discussed earlier would begin to be favored.

In order to explain the formation of fullerenes, the open graphitic sheets must rearrange to incorporate pentagons as well as hexagons in the bonding pattern. The pentagons would cause the sheet to curl and enable some of the edge carbon atoms with unsatisfied valences to bond together. The loss of p-p (-bond) overlap resulting from curling would be more than offset by the formation of good -bonds from coupling edge carbons.

Key to the success of this process is adequate "annealing time" so that the growing cluster can incorporate enough pentagons (12) to close. If the rate of cooling the carbon plasma is too rapid, most of the clusters will grow out well beyond C, becoming giant fullerenes or soot particles.

Although C is a remarkably stable carbon cluster, it will photodissociate when pulsed with laser light. Interestingly, the fragmentation proceeds cleanly via successive C losses all the way down to C. Apparently, under these conditions, C shrinks to ever smaller fullerenes until it reaches C, where strain energy becomes too great and it explodes into open fragments (O'Brien, 1988).

One possible mechanism for C extrusion is shown in Figure VI.B. A C spheroidal shell with a 5-5 ring junction could lose C and rearrange into a C spheroidal shell. The resulting structure has one less hexagon but the same number of pentagons, as required for closure. Critical to this mechanism is the existence of a process for rapid surface geometry reorganization, because clusters like C do not have pentagons sharing an edge.

Figure VI.B: Proposed mechanism for the extrusion of a C unit from a fullerene.

Early on, it was discovered that metal atoms could be placed inside C by simply carrying out the fullerene synthesis in the presence of metal atoms. For example, in the original Smalley/Kroto apparatus, if the graphite disk is impregnated with lanthanum by exposure to a boiling saturated solution of LaCl in water, carbon clusters of the form CLa (where n is an even number ranging from 44 to more than 76) are observed (see Figure VI.C) (Heath, 1985).

Figure VI.C: Distribution of CLa clusters (black bars) and C clusters (gray bars) produced by laser vaporization of a lanthanum-impregnated graphite disk in an apparatus similar to that shown in Figure II.B.

Bare C and C are also observed but, significantly, there is no evidence of any cluster picking up more than one La atom. This suggests that there is only one highly stable binding site - presumably inside the cage.

Further evidence for the metal on the inside comes from the photofragmentation behavior of CK and CCs, produced from K and Cs-impregnated graphite and then ionized in the mass spectrometer. When these clusters are laser-irradiated with sufficiently high-energy pulses, they photodissociate by successive C losses to produce smaller and smaller even carbon-number clusters (Weiss, 1988). For C K, the smallest stable member in the product family is CK, while for C Cs, the smallest fragment cluster is CCs.

These results are exactly what would be predicted based on calculated fullerene cavity sizes and known ionic radii, i.e., C and C are the smallest clusters that can accommodate K and Cs, respectively, while still maintaining a van der Waals radius of 1.65 Å for carbon. K and Cs have been essentially "shrink-wrapped" by the photofragmentation process.