1: \begin{abstract}

2: Multi-shell fullerenes "buckyonions" were simulated, starting from initially random configurations, using a density-functional-theory (DFT)-trained machine-learning carbon potential within the Gaussian Approximation Potential (GAP) Framework [Volker L. Deringer and Gábor Csányi, Phys. Rev. B 95, 094203 (2017)]. Fullerenes formed from seven different system sizes, ranging from 60 $\sim$ 3774 atoms, were considered. The buckyonions are formed by clustering and layering starting from the outermost shell and proceeding inward. Inter-shell cohesion is partly due to interaction between delocalized \text{$\pi$} electrons protruding into the gallery. The energies of the models were validated \textit{ex post facto} using density functional codes, \texttt{VASP} and \texttt{SIESTA}, revealing an energy difference within the range of 0.02 - 0.08 eV/atom after conjugate gradient energy convergence of the models was achieved with both methods.

3: 

4: \end{abstract}

5: