Coupled composition–geometry structure prediction and synthesis-accessible near-UV gap tuning in mixed MgO–ZnO nanoclusters

Chen, Shen, Saranya and Dixon (J. Phys. Chem. C 2019) predicted, via a two-stage search — a genetic algorithm for pure-cluster morphologies followed by combinatorial-ion-exchange (CIE) basin hopping over homotops at fixed morphology, with the aggregation simulations performed in conjunction with density functional theory — that mixed Zn$_x$Mg$_{60-x}$O$_{60}$ nanoclusters switch their most stable morphology from MgO-like cuboid to ZnO-like double-layer cage near $x = 42$ and show essentially monotonic HOMO–LUMO gaps spanning the near ultraviolet. That prediction rests on decoupling geometry from cation distribution, on low-cost homotop energetics whose ranking reliability at $n = 60$ is not independently established, and on short, high-temperature Born–Oppenheimer molecular dynamics as a kinetic proxy, leaving open whether the structure–gap picture and its synthesis accessibility hold where geometry and composition are strongly coupled. The open problem is to establish, computationally and/or experimentally, whether mixed MgO–ZnO nanoclusters can serve as composition-tunable near-UV gap materials: jointly sampling morphology and homotop space at larger sizes and fluxional regimes with quantified energetic reliability, and mapping the achievable gap range including metastable isomers and other sizes, with realistic synthesis routes (low-temperature aggregation kinetics, ligand/solvent/support effects, kinetic control of specific homotops) as the motivating practical context for any materials claim.

Background

ZnO and MgO are wide-band-gap oxides (bulk wurtzite ZnO about 3.4 eV, rocksalt MgO about 7.8 eV), and their alloy Mg$_x$Zn$_{1-x}$O has long been studied as a route to band gaps in the near ultraviolet: epitaxial wurtzite films reach roughly 4–4.5 eV (Makino et al. 2001) and metastable cubic-phase films exceed 5.0 eV (Choopun et al. 2002), but only within restricted solid solubilities, with phase separation at intermediate compositions, so the 4–5 eV window is hard to access reproducibly in equilibrium bulk or film material. At the nanoscale, gas-phase and computational studies of stoichiometric (ZnO)$_n$ and (MgO)$_n$ clusters — including the systematic survey by Chen, Straatsma, Fang and Dixon (2016) up to $n \le 168$ — find 'magic-number' clusters (sizes of exceptional stability and abundance relative to neighbors), with (MgO)$_{60}$ predicted as a rocksalt-like cuboid and (ZnO)$_{60}$ as a double-layer cage. This suggests that discrete mixed-cation clusters Zn$_x$Mg$_{60-x}$O$_{60}$ might interpolate between the two morphologies and their electronic structures, providing composition-tunable HOMO–LUMO gaps (HOMO/LUMO = highest occupied / lowest unoccupied molecular orbital, the molecular analogue of valence- and conduction-band edges) free of bulk phase constraints.

For a mixed cluster, a 'homotop' is a distinct way of placing the two cation species on the sites of one fixed geometry; the count grows combinatorially as $\binom{n}{x}$, so brute-force enumeration over all geometries times all species distributions is impossible at relevant sizes. Chen, Shen, Saranya and Dixon (2019) therefore introduced a two-stage partition: first find low-energy pure-cluster morphologies with a genetic algorithm (a population-based global optimizer evolving candidate structures through crossover and mutation), then sample homotops of each fixed morphology with combinatorial-ion-exchange (CIE) basin hopping — basin hopping being Monte Carlo global optimization on a locally minimized potential-energy surface (PES) — in which Zn–Mg neighbor exchanges are proposed with Boltzmann-like fitness weights and assembled into multi-exchange moves. Because every trial move needs an energy, the CIE stage relied on low-cost electronic-structure energetics, while the aggregation simulations were performed in conjunction with density functional theory; the same group had earlier shown that machine-learned structure–energy relationships can accelerate such searches for ZnO nanoparticles (Chen and Dixon 2018), and the methodology is reviewed in Chen's 2019 Annual Reports in Computational Chemistry chapter. The 2019 study found a most-stable-morphology crossover at $x = 42$ (Zn/Mg $\approx$ 2:1; MgO-like cuboid stable at low $x$, ZnO-like double-layer cage at high $x$), Zn preference for the lowest Zn–O coordination sites and Mg preference for the highest Mg–O coordination sites, thermodynamically favorable aggregation of two small nanoclusters to form the large magic-number composite, Born–Oppenheimer molecular dynamics (BOMD: classical nuclear dynamics on the instantaneous electronic ground state) evidence that aggregation depends on the structural motifs of the reactants, and HOMO–LUMO gaps essentially monotonic in $x$ and highly tunable — 3.1–4.8 eV over all sampled Zn$_x$Mg$_{60-x}$O$_{60}$ isomers and 3.2–4.2 eV over the global minima.

The strategy deliberately exploits a weak coupling between geometry and species distribution, which is what makes the two-stage decoupling valid; the ranking reliability of its low-cost homotop energetics is not established against higher-level theory at $n = 60$; it samples a limited set of structural motifs at the single size $n = 60$; and it uses short, high-temperature BOMD as a kinetic proxy far above realistic growth temperatures. Each of these choices defines a boundary of what the study established, and each boundary marks an open issue beyond that verified scope: generalization to fluxional morphologies and larger sizes where geometry and composition are strongly coupled, the reliability of surrogate-guided homotop sampling, gap tunability across other sizes, unsampled morphologies and metastable isomers, and — crucially for any materials claim — whether specific homotops or morphologies can actually be synthesized and stabilized under experimental conditions, where supports, solvents and ligands can perturb both structure and gap.

Problem Statement

Establish whether mixed MgO–ZnO nanoclusters can serve as composition-tunable near-ultraviolet gap materials, by extending the 2019 Zn$_x$Mg$_{60-x}$O$_{60}$ benchmark beyond the regime where its two-stage, weakly-coupled search strategy is valid. One overarching objective ties the components together: a predictive, synthesis-aware structure–property map for these clusters. The linked open questions, all of which a resolution of the overarching objective must address, are: (1) structure sampling that treats geometry and cation distribution jointly — fluxional morphologies, larger sizes, and regimes where species placement and morphology strongly influence each other — instead of decoupling morphology discovery from homotop search; (2) quantified energetic reliability, i.e. benchmarked or actively corrected (surrogate, density-functional, or better) energetics such that homotop rankings, morphology-stability crossovers, and site-distribution conclusions do not depend on uncontrolled error of the low-cost methods driving homotop sampling; and (3) the composition-, size-, morphology-, and homotop-resolved HOMO–LUMO/optical gap map, including whether metastable isomers beyond the original study's sampled morphologies and sizes widen the achievable near-UV gap range beyond its reported ground-state values, and how gaps respond there. Because the source study found aggregation to be thermodynamically favorable and dependent on the reactants' structural motifs, synthesis accessibility — aggregation thermodynamics and kinetics under realistic low-temperature conditions including ligand, solvent, or support effects, and the kinetic control required to obtain particular homotops or morphologies — remains essential motivating context for the materials claim. Answers may be computational, experimental, or combined.

References

1. Mingyang Chen, Kangqi Shen, Govindarajan Saranya, David A. Dixon, 'ZnxMg60–xO60 Nanoclusters with Tunable Near-Ultraviolet Energy Gaps', The Journal of Physical Chemistry C 123, 13083–13093 (2019). DOI: 10.1021/acs.jpcc.9b00314, https://doi.org/10.1021/acs.jpcc.9b00314 (source paper)
2. T. Makino, Y. Segawa, M. Kawasaki, A. Ohtomo, R. Shiroki, K. Tamura, T. Yasuda, H. Koinuma, 'Band gap engineering based on MgxZn1−xO and CdyZn1−yO ternary alloy films', Applied Physics Letters 78, 1237–1239 (2001). DOI: 10.1063/1.1350632, https://doi.org/10.1063/1.1350632
3. S. Choopun, R. D. Vispute, W. Yang, R. P. Sharma, T. Venkatesan, H. Shen, 'Realization of band gap above 5.0 eV in metastable cubic-phase MgxZn1−xO alloy films', Applied Physics Letters 80, 1529–1531 (2002). DOI: 10.1063/1.1456266, https://doi.org/10.1063/1.1456266
4. Mingyang Chen, T. P. Straatsma, Zongtang Fang, David A. Dixon, 'Structural and Electronic Property Study of (ZnO)n, n ≤ 168: Transition from Zinc Oxide Molecular Clusters to Ultrasmall Nanoparticles', The Journal of Physical Chemistry C 120, 20400–20418 (2016). DOI: 10.1021/acs.jpcc.6b06730, https://doi.org/10.1021/acs.jpcc.6b06730
5. Mingyang Chen, David A. Dixon, 'Machine-Learning Approach for the Development of Structure–Energy Relationships of ZnO Nanoparticles', The Journal of Physical Chemistry C 122, 18621–18639 (2018). DOI: 10.1021/acs.jpcc.8b01667, https://doi.org/10.1021/acs.jpcc.8b01667
6. Mingyang Chen, 'Structures and evolution of metal oxide nanoclusters: Bottom-up genetic algorithm and fragment-based energy decomposition model', Annual Reports in Computational Chemistry 15, 105–169 (2019). DOI: 10.1016/bs.arcc.2019.08.001, https://doi.org/10.1016/bs.arcc.2019.08.001
7. Giovanni Barcaro, Luca Sementa, Alessandro Fortunelli, 'A grouping approach to homotop global optimization in alloy nanoparticles', Physical Chemistry Chemical Physics 16, 24256–24265 (2014). DOI: 10.1039/c4cp03745f, https://doi.org/10.1039/c4cp03745f
8. Sooraj Kunnikuruvan, Arup Chakraborty, Dan Thomas Major, 'Monte Carlo- and Simulated-Annealing-Based Funneled Approach for the Prediction of Cation Ordering in Mixed Transition-Metal Oxide Materials', The Journal of Physical Chemistry C 124, 27366–27377 (2020). DOI: 10.1021/acs.jpcc.0c08579, https://doi.org/10.1021/acs.jpcc.0c08579
9. Rajesh K. Raju, 'Exploring Nanocluster Potential Energy Surfaces via Deep Reinforcement Learning: Strategies for Global Minimum Search', The Journal of Physical Chemistry A 128, 9122–9134 (2024). DOI: 10.1021/acs.jpca.4c04416, https://doi.org/10.1021/acs.jpca.4c04416
10. Bart Klumpers, Emiel J. M. Hensen, Ivo A. W. Filot, 'Transferable, Living Data Sets for Predicting Global Minimum Energy Nanocluster Geometries', Journal of Chemical Theory and Computation 20, 6801–6812 (2024). DOI: 10.1021/acs.jctc.4c00572, https://doi.org/10.1021/acs.jctc.4c00572
11. Jinyuan Shi, Qinghua Ren, Yi Gao, 'Accelerating Global Optimization of Cerium Oxide Nanocluster Structures with High-Dimensional Neural Network Potential', The Journal of Physical Chemistry A 129, 2190–2199 (2025). DOI: 10.1021/acs.jpca.4c07840, https://doi.org/10.1021/acs.jpca.4c07840
12. Eddy M. Lontchi, Monica Vasiliu, Ryan Backman-Flamerich, Virgil E. Jackson, David A. Dixon, 'Hydrolysis Reaction Pathways of Thorium Oxide Nanoclusters', The Journal of Physical Chemistry A 129, 4051–4067 (2025). DOI: 10.1021/acs.jpca.5c00981, https://doi.org/10.1021/acs.jpca.5c00981
13. Y. S. Wang, 'ZnO and MgxZn1−xO nanocrystals grown by non-hydrolytic route', Journal of Crystal Growth 304, 393–398 (2007). DOI: 10.1016/j.jcrysgro.2007.03.012, https://doi.org/10.1016/j.jcrysgro.2007.03.012