Magic Mathematical Relationships Governing Nanocluster Geometry and Properties

Introduction

The concept of magic numbers in nanoclusters traces back to van Hardeveld and Hartog’s 1969 study [1], predating the nanoscience era. Since then, magic sequences have appeared in two‑dimensional polygons, three‑dimensional polyhedra, fullerenes, and a broad spectrum of elemental clusters, with over a thousand Web of Science records citing the phenomenon [2‑6]. Understanding how the size and shape of a cluster dictate its physical and chemical properties—optical, catalytic, electronic, magnetic—is essential for modern nanotechnology [7].

Magic numbers arise when a cluster completes a full atomic shell; the resulting configuration exhibits a distinctive set of integers that govern nearest‑neighbor coordination, total atom count, and surface‑to‑bulk ratios. In graph‑theoretic terms, atoms are vertices and bonds are edges, revealing nested shells akin to onion layers. This study extends prior work on fcc, bcc, and hcp lattices [1, 2] by incorporating sc, diamond cubic, and Platonic solids, and by exploring topological indices and surface‑atom dispersion.

Surface atoms dominate the properties of small clusters. The fraction of surface atoms, FE, and its dependence on geometry influence catalytic activity, plasmonic response, and quantum‑dot photoluminescence [8‑9]. We quantify FE and compare cluster geometries by normalizing to a relative size parameter, d_rel, derived from unit‑cell dimensions.

Topological indices have long been employed to correlate molecular structure with physical properties. Starting with the Wiener index, researchers have developed a suite of graph‑based metrics—hyper‑Wiener, reverse Wiener, Szeged—that capture connectivity patterns [10‑12]. In this paper we derive exact magic formulas for these indices across all examined cluster types, revealing underlying polynomial relationships in n.

Methods

For each cluster geometry, we algorithmically generate atomic coordinates and then build the adjacency matrix A and distance matrix D. Two atoms i and j are considered bonded if their Euclidean distance r_ij satisfies r_ij < r_c, where r_c = 1.32·r_min and r_min is the shortest bond length. The adjacency matrix entries are defined as

$$ \mathbf{A}(i,j)=\begin{cases} 1 &\text{if } r_{ij}< r_c \text{ and } i\neq j\\ 0 &\text{otherwise} \end{cases} $$

The coordination number of vertex i equals the sum of the ith column of A. The distance matrix D records the length of the shortest path between all vertex pairs:

$$ \mathbf{D}(i,j)=\begin{cases} 0 & i=j\\ d_{ij} & i\neq j \end{cases} $$

Efficient algorithms exist for computing D from A [13]. With these matrices we evaluate the Wiener index W(G), hyper‑Wiener index WW(G), reverse Wiener index rW(G), and Szeged index Sz(G) following established protocols [14].

Cluster size is quantified by the number of shells n (with shells numbered 0 to n). The total number of atoms is

$$ N_T(n)=\sum_{c_n=1}^{c_{nM}} N_{c_n}(n) $$

where N_{c_n}(n) counts atoms with coordination c_n, and c_{nM} is the maximum coordination in the cluster. Surface atoms satisfy c_n < c_{ns}, yielding

$$ N_S(n)=\sum_{c_n=1}^{c_{ns}} N_{c_n}(n) $$

The total number of bonds follows from double counting:

$$ N_B(n)=\frac12\sum_{c_n=1}^{c_{nM}} c_n\,N_{c_n}(n) $$

These expressions, combined with the adjacency matrix data, provide a rigorous check against previously derived polynomial formulas for Platonic solids [2].

For each geometry we compute the first four integer values of n (or the appropriate parity subset for structures with alternating behavior) and fit the data to a cubic polynomial in n. Symbolic algebra ensures exact rational coefficients. The resulting magic formulas for N_T(n), N_B(n), and the coordination distribution are tabulated in the supporting tables.

The fraction of exposed surface atoms is defined as

$$ \text{FE}=\frac{N_S}{N_T}\times 100\% $$

Relative cluster size is expressed via

$$ d_{rel}=b\,N_T^{1/3},\qquad b=d_{at}^{-1}\left(\frac{6V_u}{\pi n_u}\right)^{1/3} $$

where d_at is the covalent atomic diameter, V_u the unit‑cell volume, and n_u the number of atoms per cell. The constant b equals 1.105 for fcc/hcp, 1.137 for bcc, 1.488 for sc, and 1.517 for diamond cubic lattices. This scaling allows cross‑comparison of clusters irrespective of crystal structure.

Results and Discussion

Table 20 (not reproduced here) summarizes key milestones in nanocluster research up to 2018, highlighting the rapid expansion of synthesis techniques for transition‑metal clusters.

FCC Clusters

Eight transition metals adopt the fcc lattice, including the noble metals and key catalytic species. Experimental synthesis of fcc clusters spans cubes, octahedra, cuboctahedra, and icosahedra, with the (111) facet favored for its lower surface energy [7]. Our magic formulas for fcc rhombic dodecahedra exhibit distinct even/odd sequences, matching the literature when shell numbering is aligned (see Table 1). The derived expressions agree with those reported in Refs. [2, 4] and extend them by providing explicit n‑dependent forms.

BCC Clusters

Seven transition metals crystallize in the bcc lattice; iron remains the only magnetic element with this structure. Nanocubes of iron have been synthesized, while fcc iron clusters are also reported [26]. We present magic formulas for five canonical bcc shapes (Table 2), noting that bcc structures exhibit more complex coordination patterns than fcc counterparts.

HCP Clusters

Twelve transition metals possess hcp structures, yet few have been successfully synthesized as discrete clusters due to oxidation and limited chemical interest. Gold clusters adopting hexagonal bipyramidal shapes have been observed [27], while the truncated hexagonal bipyramid appears only in α‑Fe₂O₃ [28]. Magic formulas for hcp geometries are provided (Table 3).

Platonic Clusters

The Platonic solids—cube, tetrahedron, octahedron, icosahedron, dodecahedron—serve as archetypal cluster geometries. We present magic formulas for icosahedral, dodecahedral, tetrahedral, and body‑centered tetrahedral clusters (Tables 4‑7). Notably, the dodecahedron exhibits a unique surface coordination of seven, leading to mixed surface/bulk coordination in successive shells. Gold nanoclusters have been synthesized in all Platonic shapes [29, 30], validating our theoretical predictions.

Diamond Cubic, Simple Cubic, and Decahedron Clusters

Silicon, germanium, and diamond‑carbon adopt the diamond cubic lattice, with hydrogen‑terminated silicon clusters achieving 8–15 nm cubic morphologies [31, 32]. We predict the composition of such Si–H clusters as Si_{8n³+6n²−9n+5}H_12n−8, where n is the shell count. For sc clusters, which are realized only in polonium, our analysis extends the previous one‑dimensional hypercube study [4] by detailing full magic formulas (Table 8).

Magic Topological Formulas

We compute the vertex‑degree entropy

$$ I_{vd}=\sum_{i=1}^{v} a_i \log_2 a_i $$

using ToposPro (v5.3.2.2) on cif files from the Crystallographic Open Database. Table 22 shows that the sc lattice has zero complexity, while fcc, hcp, bcc, and diamond cubic lattices display increasing values of I_vd. These metrics correlate with the difficulty of deriving stable topological indices; for instance, Szeged indices for bcc cubes could not be solved due to high complexity.

Magic formulas for the Wiener, hyper‑Wiener, reverse Wiener, and Szeged indices are listed in Tables 23‑25. The polynomial degrees reach 7–9, reflecting the rich connectivity of complex lattices. Notably, sc clusters yield the simplest forms, whereas fcc structures provide the most tractable expressions.

Dispersion

Figure 2 (see below) plots the fraction of exposed surface atoms, FE, against the relative size parameter d_rel for several cluster shapes. For platinum clusters of 2.2 nm (d_rel≈7.5), FE values are 47.9 % (icosahedral), 52.8 % (cuboctahedral), 57.5 % (decahedral), and 58.9 % (octahedral). The octahedral shape thus offers the highest surface exposure for a given size, correlating with its superior oxygen‑reduction reaction (ORR) activity [38]. Similar trends hold for larger 13 nm clusters (d_rel≈25–30), where FE differences narrow but strain effects in icosahedral clusters may still influence catalytic performance [40, 41].

Conclusions

We have catalogued 19 nanocluster geometries and derived exact magic formulas for atom counts, bond numbers, coordination distributions, and four key topological indices. These results, many of which are new to the literature, bridge the gap between cluster structure and property prediction. In particular, the quantified surface‑atom dispersion provides a useful metric for anticipating catalytic behavior. We anticipate that these comprehensive datasets will serve as a valuable resource for researchers designing nanoclusters with tailored properties.

Abbreviations

bcc:

body‑centered cubic

cif:

crystallographic information file

fcc:

face‑centered cubic

FE:

Fraction Exposed (surface‑atom dispersion)

hcp:

hexagonal close‑packed

ORR:

oxygen‑reduction reaction

rW(G):

reverse Wiener index

Sz(G):

Szeged index

W(G):

Wiener index

WW(G):

hyper‑Wiener index