Introduction

Superatoms, clusters that exhibit properties similar to those of elements different to those of which they are composed, have the potential to serve as building blocks for unprecedented materials with tunable properties. Studies on superatoms have been mainly theoretical and have been supported by mass detection in the gas phase under high-vacuum conditions1,2. With regard to further progress in this research field, achieving a solution-phase synthetic method for superatoms is important. The approach using noble metals Au, Ag, and Cu produces ligand-protected clusters, and although they have produced cluster crystals3,4, these are thermodynamically stable magic clusters. Therefore, a more advanced synthetic strategy is required.

Experimental studies on superatoms started with the synthesis of vaporized sodium and aluminum clusters5,6, and have attracted wide attention since the discovery of the Al13 halogen-like superatom by Khanna et al7. This was followed by the synthesis of Al7, As7, and KO3 8,9,10. Superatom properties can be estimated using a simple theory referred to as the Jellium model11 that defines superatomic orbitals (e.g., 1S, 1P, 1D, 2S, 1F, 2P) in a highly symmetric cluster. For example, the Al13 cluster has 39 valence electrons and exhibits a halogenic nature based on its superatomic orbitals (1S2, 1P6, 1D10, 2S2, 1F14, 2P5), resulting in a reactivity with I− and the high stability of the mono anionic species Al13 − 7,12.

The aluminum clusters with anionic or cationic states have been extensively studied and confirmed by gas-phase synthesis and quantum chemical calculations13,14,15. Such a study also revealed photoelectron spectroscopy about the energy levels of the valence electrons using laser light, and the superatom character of Al13 − containing an extraordinary stability against oxidation16,17,18. On the other hand, the solution-phase synthesis has advantages because the synthesis is scalable without any special apparatus. Although poly(vinylpyrrolidone) or tetrahydrofuran (THF) units have been reported to theoretically stabilize the Al13 cluster19,20, the solution-phase synthesis of Al13 − has not yet been achieved, but is expected21.

Specially designed dendrimers are one of the effective tools for the synthesis of superatoms because metal units can be set inside the structure and are protected by the shell-effect of the dendrons22,23,24,25,26. Dendritic poly-phenylazomethines25,27,28,29,30,31,32 (DPAs) have been developed that can provide various size-controlled metal clusters24. From these features, the DPAs have a potential to be a template for superatom synthesis though the capsuling dendrimer has a risk to prevent some features from the superatoms.

Here, we use this method to synthesize the aluminum clusters including Al13 −, which is the best-known superatom. This approach enables core-level X-ray photoelectron spectroscopy (XPS) and electrochemical measurements, which are not compatible with the gas-phase synthetic approaches. By investigating a series of aluminum clusters, the superatomic nature of Al13 − in solution is confirmed.

Results

Fabrication of aluminum clusters using DPAG4

The preparation method is illustrated in Fig. 1a. The number of Al atoms in the clusters was controlled using the stepwise assembly feature of the DPAs. The fourth-generation DPA dendrimer (DPAG4) and that with a pyridine core (pyDPAG4) were used (Supplementary Fig. 1). The assembly of AlCl3 was confirmed by UV–Vis titration; the absorption change in a THF solution of the DPAG4 or pyDPAG4 during the addition of AlCl3 indicated complexation behavior between the imine site and AlCl3 in a radial fashion, and a multi-step shift of the isosbestic point provided evidence that the targeted amounts of the AlCl3 were coordinated in the DPAs24. The DPAG4 enabled the controlled assembly of 4, 12, 28, and 60 units of AlCl3 (Supplementary Fig. 2). In contrast, the pyDPAG4, which has a pyridine core that can coordinate one additional metal salt, enabled the assembly of 13 units of AlCl3 in the dendritic capsule. Supplementary Figure 3 shows assembly of AlCl3 from 0 to 61 equivalents. The obvious change in the isosbestic point at 13 equivalents demonstrates preparation of 13AlCl3–pyDPAG4 in solution. The AlCl3 units positioned in the DPAs were reduced into clusters by a benzophenone ketyl radical prepared using sodium metal in THF. The fabricated aluminum clusters were then subjected to various experiments. The aluminum clusters with 14 atoms were prepared using the DPAG4. Even though the branched structure of the DPAG4 was not the best for 14 atom clusters, the enough ability to fabricate the cluster as the major product with small deviation was already reported in our previous work29.

Fig. 1
figure 1

Fabrication of Al13 − using the dendrimer template. a Illustration of dendritic poly-phenylazomethines including an aluminum cluster in the solution state. b Negative-mode matrix-assisted laser desorption/ionization mass spectra of Al13 − and blank. 1,8,9-Trihydroxyanthracene was used as the matrix. c A high-angle annular dark-field scanning transmission electron microscopy image of Al13 − on a TEM grid covered with a thin carbon film (scale bar: 5 nm). d Histogram of the observed particle sizes in the STEM sample

Full size image

The fabricated cluster was protected by the dendrimer shell; however, the superatomic Al13 − could be detected using negative-mode matrix-assisted laser desorption/ionization (MALDI) mass spectrometry at m/z = 350.9 via dissociation from the DPA template during the laser desorption process. This peak was overlapped by a fragment of the pyDPAG4, but the formation of Al13 − in solution was still confirmed (Fig. 1b). Al13 − was specifically observed when the 13AlCl3-pyDPAG4 complex was reduced, whereas it was not detected for 12AlCl3-pyDPAG4 and 14AlCl3-pyDPAG4 (Supplementary Fig. 4). In addition, the Al13 − cluster was detected using high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM). The observed cluster sizes were measured to be 1.0 ± 0.1 nm (Fig. 1c, d). The cluster sizes were reasonable considering the calculated diameter of Al13 − (0.95 nm, see Methods). These results confirm that the Al13 − superatom exists in solution.

Electronic state

The oxidation states of the clusters were estimated from XPS Al 2p spectra. The XPS samples were prepared on glassy carbon substrates by repetitive drop-casting. The spectra and binding energies (BEs) of the aluminum clusters which were synthesized using 4, 12, 13, 28, and 60 AlCl3 at the DPAs are summarized in Fig. 2a, b, respectively. All the aluminum clusters possessed BEs equal or lower than that of bulk aluminum, thereby indicating the presence of reduced clusters. In addition, the BE for n = 13 cluster seems to be different from the BEs for n = 4, 12, and 28 atomic clusters (Fig. 2b). Although the XPS spectra generally reflect the surface state of materials, it is not applied to these aluminum clusters because they have lower than or equal to 60 atoms with the diameters of about 1 nm. Considering that the photoelectrons can escape from about a 1–4 nm thick material33, almost all atoms in the clusters were analyzed.

Fig. 2
figure 2

X-ray photoelectron spectroscopy results of the aluminum clusters. a XPS spectra and b binding energies of the Al n cluster samples, aluminum metal and Al2O3. n ( = 4, 12, 13, 28, and 60) is the number of assembled AlCl3 units in the DPAs. Error bars in b stand for the standard deviation of the mean of three independent experiments

Full size image

To interpret these XPS results in a theoretical manner, population analyses for aluminum clusters including Al13 – were performed (Supplementary Fig. 5). The employed structures were based on the literatures34,35. To estimate the charge distribution effect on the 2p binding energies, we have calculated the 2p orbital levels. Although core-hole relaxation in the final state has not been taken into account, simple correlation between core levels and binding energies has been reported36. In the case of Al13 –, the 2p levels of the inner atom are apart from those of surface atoms (Supplementary Fig. 6). This difference offers possibility of two peaks in the XPS spectrum. However, only one broad peak was detected in our experiments. The observed peak could be ascribed to 12 surface atoms or the mixture of 12 surface atoms and an inner atom. Also, in the cases of anionic clusters in other sizes, the non-uniform charge distributions yield different 2p levels, as shown in Supplementary Fig. 7a. However, all the 2p levels are close and overlap to be one broad peak. The size dependence of the 2p levels of anionic clusters is consistent with our experimental results. This is in contrast to the neutral clusters (Supplementary Fig. 7b).

These shift of binding energy in the XPS measurements and calculated energy levels could be considered by the influence of the negative charges. The charges per atom were simply calculated based on the assumption that the anionic charges are fully delocalized in the aluminum clusters (Supplementary Fig. 8). The anionic charges per atom increased as the number of atoms decreased. This tendency was in agreement with the observed BEs. Therefore, these BEs are assignable to monovalent or more than divalent anionic states.

Stability

The reactivity of the aluminum clusters with oxygen was evaluated by monitoring the change in the XPS Al 2p peaks. The samples deposited on the substrates with the DPAs were exposed to air and subjected to Ar sputtering (2 kV, 2 min) to remove any surface impurities before the measurement. The results are shown in Fig. 3. The peaks corresponding to the oxidized species from Al12 and Al14 clusters increased within several tens of hours, whereas the peak for Al13 − maintained its position. This result demonstrates the specific stability of the Al13 − clusters toward oxygen and is consistent with the superatomic nature observed in the gas phase18. The clusters observed within the DPAs were deposited with ketyl radical salts; however, they did not affect the stability because the radical anions were quickly oxidized in air within several minutes (Supplementary Fig. 9). The extraordinary stability of Al13 − was well theoretically investigated in previous reports37,38,39. In addition, inclusive studies of aluminum clusters were also reported with their local and global minima34,40,41. They fully supported our result.

Fig. 3
figure 3

Change in the XPS Al 2p peaks upon air exposure. The graphs show the spectral change for the Al n clusters. a n = 13, b n = 12, and c n = 14. The exposure times were 1–2 min, 5 h, 24–25 h, and 46–48 h. The vertical axes are normalized intensity. The spectra for Al12 and Al14 samples could be fitted by two peaks derived from the initial reduction peak (blue dashed line) and oxidation peak (red dashed line). Ratio of the oxidized (red squares) to the as-prepared (blue diamonds) peaks for d Al12 and e Al14 samples

Full size image

Electrochemical oxidation

The oxidation processes of the prepared aluminum clusters in the DPAs were investigated by potentiodynamic electrochemical measurements in THF solutions containing the sodium benzophenone ketyl radical (Fig. 4). To unify the materials in the cell, DPAG4 was used for all the aluminum clusters. All steps of the measurements were conducted under an Ar atmosphere. The open circuit potentials of the samples were approximately −1.7 V vs. Ag/Ag+. The potential was then swept to higher values. The Al13 − cluster exhibited a shoulder-like peak (i) at −0.97 V and a second oxidation peak (ii) at −0.73 V (Fig. 4b, black arrows). The cyclic voltammogram of AlCl3 measured in THF showed a reduction wave of AlCl3 below −1.0 V and an increase of the oxidation current from −0.6 V (Supplementary Fig. 10). Considering these results, the first oxidation step observed in the Al13 − sample could be assigned to oxidation from Al13 − to Al13 0. The oxidation current of Al13 − was maintained after the process, which suggests that the oxidized clusters were then re-reduced by the benzophenone ketyl radicals in the cell. This one electron oxidation of Al13 − suggested tuning of the electronic state by the applied electronic potential in solution. In contrast, both Al12 and Al14 clusters exhibited different phenomena, and only an intense oxidation peak that corresponded to the second oxidation process was observed for Al13 − (Fig. 4a, c). As their oxidation potentials were nearly the same, these intense peaks are considered to be derived from the oxidation of the aggregated aluminum clusters on the surface of the working electrode and organic materials arising from the DPAG4.

Fig. 4
figure 4

Oxidation waves of the fabricated anionic aluminum cluster samples. They correspond to Al n clusters. a n = 12 (6 μM), b n = 13 (4 μM), and c n = 14 (4 μM) in THF. The potential sweep was started from −1.7 or −1.8 V (open circuit potentials) at the scan speed of 0.025 V/s. Black bars indicate 10 μA in a and c, and 5 μA in b. NaPF6 was used as an electrolyte (0.05 M)

Full size image

Discussion

The fabricated clusters were protected by the DPA template. It is useful for synthesis of clusters in solution27,28,29, however, the features have possibility to be different from the inherent character. Considering the isolation of clusters from the DPAs, stability of the clusters is a crucial factor.

In conclusion, aluminum clusters, including the superatomic Al13 − species, were fabricated in solution by utilizing the controlled metal assembly of DPA templates. To the best of our knowledge, this is the first example of the synthesis of the benchmark Al13 − superatom in solution. The Al13 − cluster was identified by mass spectrometry and STEM measurements. In addition, specific features of the Al13 − in the XPS, oxidation reaction and electrochemical measurement demonstrated its successful synthesis in solution. The oxidation states of the aluminum clusters were determined from BEs measured in XPS Al 2p spectra. The stability and reactivity of Al13 − were compared with those of other aluminum clusters. The particularly high stability of Al13 − was demonstrated by air exposure and through theoretical calculations. In addition, the aluminum clusters in solution exhibited redox behavior derived from Al13 −/Al13 in DPAG4. The solution-phase synthesis of Al13 − superatoms has a significant role for the experimental development of cluster science.

Methods

Materials

pyDPAG4 and DPAG4 were synthesized by a convergent method. The fourth-generation phenylazomethine dendrons were connected to the core molecules. The detailed procedures are described in previous reports42,43. AlCl3 (ultra dry) was purchased from Alfa Aesar, a Johnson Matthey Company. Dehydrated THF (stabilizer-free) was obtained from Kanto Chemicals. Sodium metal, benzophenone and tetrabutylammonium hexafluorophosphate were obtained from Wako Pure Chemical Industries, Ltd. and TCI Co., Ltd.

Assembly of AlCl3 in pyDPAG4

The experiment was conducted in an N2-filled glove box. A THF solution of AlCl3 (2.60 mM) and pyDPAG4 (4.01 μM) was prepared. An appropriate equivalent of AlCl3 solution was then added to the DPAG4 solution. Complexation between AlCl3 and DPAG4 was monitored using UV–Vis spectroscopy.

Assembly of AlCl3 in DPAG4

The experimental procedure was conducted in the same way as that with pyDPAG4. The concentrations of AlCl3/THF and DPAG4/THF were 5.41 mM and 5.06 μM, respectively.

Fabrication of aluminum clusters

pyDPAG4 was used for the Al13 − cluster, and DPAG4 was used for the Al4, Al12, Al14, Al28, and Al60 clusters. The inferior number (n) of Al n reflects the added equivalent of the AlCl3 in the DPAs. Dendrimer solutions in THF (3–8 μM) were prepared in an Ar-filled glove box. The required equivalents of AlCl3 in THF were added to each dendrimer solution to prepare the AlCl3-complex dendrimer. The benzophenone ketyl radical was prepared by the addition of an excess amount of sodium metal into the solution of benzophenone in THF (0.3 M). The THF solution of benzophenone ketyl radicals (1 mL) was added to the prepared AlCl3-complex solution (1 mL) to fabricate aluminum clusters at the DPAs. Samples for STEM measurements were prepared by the use of 60 equivalents of benzophenone to AlCl3 for reduction.

Characterization

UV–Vis spectra were recorded at 20 °C using Shimadzu UV-3600 and UV-3100PC spectrometers with a quartz cell having an optical length of 1 cm. XPS spectra were measured with a PHI 5000 VersaProbe (Ulvac-Phi, Inc.). Al Kα (15 kV, 25 W) radiation was used as the X-ray source. The beam was focused on a 100 μm2 area. Samples were sputtered with an Ar ion gun to remove the oxidized surface prior to the measurements. The spectra were analyzed with the MultiPak software (Physical Electronics), and were standardized according to the Au 4f 7/2 peak at 84.0 eV. Background subtract, peak smoothing and fitting were used to estimate peak areas. STEM images were obtained using a transmission electron microscope (JEOL, ARM-200F) and the HAADF method. STEM samples were deposited on a super high-resolution carbon film with a Cu mesh (Okenshoji Co.). Cyclic voltammetry was performed using a BAS ALS750B analyzer. A glassy carbon disc electrode and platinum wire were used as the working and counter electrodes, respectively. An Ag+/Ag (0.01 M AgNO3 in 0.1 M Bu4NClO4/acetonitrile) electrode was used as the reference electrode. NaPF6 (0.05 M) was used as the electrolyte.

Calculations

Geometry optimizations for Al4, Al12, Al13, and Al28 in anionic and neutral states were carried out with the B3LYP/6–31 G(d, p) level of theory. The employed structures were based on the past literatures34,35. Vibrational analyses were also conducted for all the optimized structures and no imaginary frequency appeared. For the theoretical interpretation of the observed XPS data, Mulliken and natural population analyses were performed. In addition, the 2p orbital levels were calculated by single-point energy calculations with the GB3LYP/DZP-DKH level of theory within fourth-order Douglas–Kroll–Hess relativistic approximation, which is aimed at the incorporation of spin-orbit coupling. The diameter of Al13 – was estimated using the Monte-Carlo method, in which the molecular volume was defined as the volume inside a contour with a density of 10–3 electrons/Bohr3. All the calculations were conducted with Gaussian 1644.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.