DOI: 10.1002/chem.201000327 Gold Sulfide Nanoclusters: A Unique Core-In-Cage Structure De-en Jiang,* [a] Michael Walter, [b, c] and Sheng Dai [a] Nanoparticles of chalcogenides of group 12 elements, such as CdS, ignited a revolution in nanoscience. [1] These quan- tum dots have the bulk bonding structure and a larger tun- ACHTUNGTRENNUNGable optical gap than the bulk, and find wide applications in biomedical imaging, [2] electronic devices, [3] and solar cells. [4] However, few studies have been directed toward the nano- particles of Au chalcogenides, such as Au 2 S. Although nano- crystals of Cu 2 S and Ag 2 S have recently been synthesized in a controllable way, [5] wet-chemistry preparation of Au 2 S nanoparticles by reduction of Au III often led to a mixture that contained Au nanoparticles. [6] Typical sizes of prepared Au 2 S nanoparticles range from 4 to 7 nm. Little is known of Au 2 S nanoparticles below 4 nm. In contrast, monodisperse thiolated gold nanoparticles can be made in a variety of sizes, from sub-nanometer to several nanometers; [7] in fact, Au 102 (SR) 44 and Au 25 (SR) 18  have been crystallized (SR = thiolate group). [8] The Au–S framework of Au 25 (SR) 18  is about 1 nm in size. Jin and co- workers [9] found that when the Au 25 (SR) 18  clusters are sub- jected to in-source fragmentation in matrix-assisted laser de- sorption/ionization (MALDI) mass spectrometry, a series of Au x S y  clusters are recorded in the mass spectrum due to se- lective breaking of the S  C bonds [10] and loss of all R groups and some S atoms. The most abundant species is Au 25 S 12  , followed by Au 23 S 11  and Au 27 S 13  . Au 25 S 12  can also lose a single Au atom to give Au 24 S 12  . [9] The abundance of Au 25 S 12  is independent of the RS group, [9] which indi- cates that this cluster has some magic nature. Because these Au x S y  clusters have an Au/S ratio of close to 2, they pro- vide a means to fill our knowledge gap regarding Au 2 S nanoclusters in this size range. To understand these Au x S y  clusters, one needs to discov- er their structures. However, the only available experimental data are the mass spectra. [9] Given the absence of knowl- edge, a global minimum search is needed to elucidate these Au x S y  clusters. To that end, we have employed the basin- hopping technique [11] for a global minimum search with ge- ometry optimization by density functional theory (see the Computational Methods section and the Supporting Infor- mation for details). This DFT-based basin-hopping is a pow- erful tool for finding global minima for clusters in which classical potentials are either unavailable or not accurate enough to be predictive; it has been successfully employed for many systems, such as gold, silicon, and carbon–boron clusters. [12] Herein, we used this technique and found global minima for the three most abundant Au x S y  clusters; a struc- tural pattern emerges that indicates these clusters are sym- metric with a unique core-in-cage construction, which can explain the observed mass peaks. [9] Our approach to finding the global minima of Au x S y  fol- lowed the history of the sample in the MALDI experiment. We started with the structure of the Au 25 (SR) 18  cluster with 18 R groups removed (Figure 1a). We then assumed that the middle S atoms of the six S-Au-S-Au-S motifs are lost (Fig- ure 1b); [9b] this is a convenient assumption in that these six S atoms belong to the same group (roughly equivalent in sym- metry) and are 1.1  farther from the cluster center than the other 12 S atoms. We then performed a geometry opti- mization of the Au 25 S 12  structure in Figure 1b and obtained a local minimum (Figure 1c). The structural change from Figure 1b to c is rather dramatic and is accompanied by a huge lowering in energy (1785.0 kJ mol 1 ); the local mini- mum is a distorted structure with two (AuS) 3 triangular rings and a longer Au–S oligomer on the cluster surface. Starting with Figure 1c as our initial guess, we performed a DFT-based basin-hopping search for the global minimum. After about 600 Monte-Carlo steps, we found the highly [a] Dr. D.-e. Jiang, Dr. S. Dai Chemical Sciences Division, Oak Ridge National Laboratory Oak Ridge, TN, 37831 (USA) Fax: (+ 1) 865-576-5235 E-mail : jiangd@ornl.gov [b] Dr. M. Walter Physics Department, University of Freiburg Hermann-Herder-Strasse 3 79106 Freiburg (Germany) [c] Dr. M. Walter Freiburg Materials Research Center, University of Freiburg Stefan-Meier-Strasse 21 79104 Freiburg (Germany) Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/chem.201000327. Chem. Eur. J. 2010, 16, 4999 – 5003  2010 Wiley-VCH Verlag GmbH&Co. KGaA, Weinheim 4999 COMMUNICATION