Introduction

Through the deposition of size-selected atomic clusters consisting of a few to thousands of atoms on well-defined substrates, nanostructured surfaces can be produced through bottom-up fabrication, which is a promising method for creating low-dimensional nanomaterials with atomic-scale structural precision1,2,3,4. The properties of functionalized nanostructured surfaces can be controlled by designing cluster–surface interactions, which facilitates a nanoscale approach to developing nanomaterial-based modified electrodes for application in electrochemistry5. The cluster–surface interaction is a fundamental characteristic of such nanostructured materials3,6, and has been a focus in the preparation of heterogeneous catalysts through control of the physical and chemical properties, size, and dimensionality7,8,9,10. For example, Haruta indicated the importance of choosing a substrate in enhancing the catalytic activity of gold (Au) nanoparticles for low-temperature CO oxidation7. In addition, the substrate acidity has been reported to control the catalytic activity of size-selective platinum (Pt) clusters10. In these studies, localized cluster–surface interactions are enhanced using metal oxide substrates8,9 to avoid the generation of weakly bound nanoclusters (NCs) on a clean surface, because these NCs generally behave as a two-dimensional gas, ultimately resulting in aggregation6.

Interactions that take place through charge transfer (CT), or more explicitly, electron transfer11, are important in chemical reactions between two reactant molecules since they lead to the formation of intermolecular CT complexes that exhibit a new electronic transition known as a CT band12. Their segregated stacking can lead to molecular electrical conductivity, including superconductivity13,14. Such CT processes play an important role in cluster–surface interactions. More specifically, due to the CT interactions with pre-deposited organic molecules on a substrate, the NCs can exist in a monodisperse state on the surface15,16.

Among various gas phase NCs and their characteristic functionalities explored during the past several decades, NCs formed with a highly symmetrical geometry and an electronically closed shell are known as “superatoms” (SAs), which mimic the chemical properties of atoms with clusters17,18,19,20,21,22,23,24,25. In particular, anionic aluminum (Al) NCs with 13 atoms, i.e., Al13−, are promising candidates for the fabrication of SA assembled nanomaterials26,27,28,29,30,31, because Al13− simultaneously satisfies both icosahedral packing and the electronic shell closing32,33 of 40 electrons as (1S)2(1P)6(1D)10(2S)2(1F)14(2P)6, thereby facilitating the bottom-up fabrication of nanostructures with desired functionalities, similar to the case of building nanoblocks34,35,36.

In this study, we show that the choice of organic substrate can allow molecular control of the CT interactions at the cluster–surface interface and stabilize SAs on the surface. Since the localized interactions between the pre-decorated organic molecules and the deposited NCs are enhanced compared to those of a clean bulk metal or semiconductor substrate, the organic substrate is key to immobilization of the deposited NCs, in which the NC aggregation caused by two-dimensional gas behaviors is suppressed16,25. Thus, we deposit Al13− and boron-doped B@Al12− SAs17,26,37,38 on organic substrates of n-type C60 and p-type hexa-tert-butyl-hexa-peri-hexabenzocoronene (HB-HBC, C66H66 (see Supplementary Fig. 1 and Supplementary Note 1)). Spectroscopic characterization by X-ray photoelectron spectroscopy (XPS) and oxidative reaction measurements of the Al13− and B@Al12− SAs on the organic substrates are then conducted to reveal that superatomic behavior can be observed on the p-type organic substrates through CT interactions.

Results

Charge state of the Aln NCs on n-type C60 and p-type HB-HBC substrates

Through magnetron sputtering (MSP) of the Al targets, the generated Aln− NCs possessed a mass-to-charge ratio (m/z) predominantly in the range of 200−800 (see Supplementary Fig. 2). With an ion current of 300 pA, samples containing 2.9 × 1013 mass-selected NCs (~0.6 monolayers (MLs)) could be prepared within 3 h (see “Methods” section and Supplementary Note 2). The morphology of the deposited NCs on the organic substrate was confirmed by scanning tunneling microscopy (STM) imaging16,25, wherein the SAs were found to be monodispersively immobilized without aggregation (see Supplementary Fig. 3).

Figure 1a, b show the XPS spectra around Al 2p core levels for (a) Al13 on C60 and (b) Al13 on HB-HBC before (lower) and after (upper) O2 exposure, respectively. The binding energies (BEs) of Al 2p3/2 for the bulk Al (Al0) and oxidized Al (Al3+) have been previously reported (marked by vertical bars in the figure)39. As can be seen, without O2 exposure, the Al atoms on the C60 substrate are completely oxidized, while the Al atoms on HB-HBC are not oxidized. Following O2 exposure, the Al atoms on C60 remain unchanged, while Al atoms on HB-HBC are oxidized to Al3+. As shown in Fig. 1c, d, the corresponding O 1 s component can be observed in the lower trace of Fig. 1c even without O2 exposure. These results show that the Al13 NCs present on the C60 substrate are so reactive that the nascent NCs are oxidized after deposition by some residual gas in the vacuum chamber (<10−5 Pa) during the deposition process.

Fig. 1: XPS spectra for Al13 on the C60 and HB-HBC substrates.
figure 1

a–d XPS spectra around the Al 2p core levels for a Al13 on C60 and b Al13 on HB-HBC before (lower) and after (upper) O2 exposure (at 5 × 1010 Langmuir (L = 1.33 × 10−4 Pa·s)), along with c, d the O 1s spectra for each state. e, f The XPS spectra around the C 1s core level for the underlying e C60 and f HB-HBC are also shown for the deposition of Al13 and Ta@Si16 or Lu@Si16. Reference binding energies (BEs) of Al 2p3/2 for the bulk Al (Al0 and Al3+) and O 1s (O2−) are marked by vertical bars. The BEs for Al 2p show zerovalent Al0 only for Al13 on the HB-HBC substrate before O2 exposure, while the other BEs are in the vicinity of Al3+. After the deposition of 0.6 ML Al13 (blue) or Ta@Si16 (black) on C60, the C 1s peak in (e) shifts by ~0.3 eV toward a lower BE from that before deposition (light blue), indicating the presence of an anionic C60− state. After the deposition of 0.6 ML Al13 (red) or Lu@Si16 (black) on HB-HBC, the C 1s peak in (f) shifts by ~0.25 eV toward a higher BE from that before deposition (pink), indicating the presence of a cationic HB-HBC+ state.

Full size image

The contrasting oxidation behavior of these two systems correlates well with the C 1s XPS peaks from the underlying C60 or HB-HBC on highly oriented pyrolytic graphite (HOPG), where the C 1s signals are mainly derived from the topmost molecular layer (Fig. 1e, f). As shown in Fig. 1e, after the deposition of Al13 on C60, the C 1s peak shifts toward a lower BE by ~0.30 eV. Although Al13 is nascently oxidized, the shift to a lower BE shows that an anionic C60− state is formed by Al13 oxides through a CT interaction16,40; the degree of shift corresponds well to the formation of C60− as reported in the literature41 (see Supplementary Note 3 and Supplementary Fig. 4). A similar C 1s shift has been reported when the alkali-like Ta@Si16 SA25 is deposited on C60, wherein a shift attributable to Ta@Si16+C60− is observed40,41, as denoted in Fig. 1e. More quantitatively, when the C60-derived C 1s peak is deconvoluted into two peak components corresponding to C60 alone (non-interacted) and bound with the Aln oxide (interacted), the BE of the interacted C60 peak is 0.33 eV lower than that of non-interacted C60 (Supplementary Fig. 4). In addition to Al13, the Al7 NCs deposited on C60 is nascently oxidized, as can be observed from the Al 2p XPS spectrum, although Al7+ is regarded to complete the 2S shell (i.e., 20 e−)21. In contrast, after the deposition of 0.6 MLs of Al13 on the HB-HBC substrate, the C 1s peak shown in Fig. 1f shifts toward a higher BE by ~0.25 eV. Since a similar behavior can be observed for the deposition of the halogen-like Lu@Si16 SA25 onto HB-HBC, this shift suggests the formation of a cationic HB-HBC+ state, and in turn, an Al13−/HB-HBC+ CT complex.

In addition to Al13−, all Aln− NCs (n = 7–24) can be size-selectively deposited onto C60 and HB-HBC substrates. More specifically, the Al 2p XPS spectra show that these Aln NCs were successfully deposited onto HB-HBC without undergoing any oxidation reactions (see Supplementary Fig. 5). However, complete oxidation was observed for the Aln NCs deposited on C60. It should be emphasized that this contrast in the reactivity of Aln results from the different types of organic substrate molecules, i.e., n-type and p-type for C60 and HB-HBC, respectively. In the Al 2p XPS spectra for the Aln NCs on HB-HBC, peaks were observed in the range of 73.0–73.2 eV, which is close to the peak position for bulk Al (i.e., 73.0 eV)39 (see Supplementary Fig. 6). In addition, the small size dependence is consistent with that in the Al 2p core-level BEs for Aln+ (n = 12–15) obtained from the soft X-ray photoionization efficiency curves42. More precisely, the charge states of the Al atoms for the deposited Aln NCs can be discussed in terms of their Al 2p peak positions; the BEs of Al 2p for all Aln NCs are slightly higher than that of the bulk Al (zerovalent Al0), suggesting that the Aln NCs on HB-HBC are anionic rather than neutral. Recently, Kambe et al. have reported the Al 2p XPS spectra for several Aln species (n = 4, 12, 13, 28, and 60) synthesized with dendrimers43, and they revealed a size-dependent behavior in the Al 2p XPS spectra from 71.2 (n = 4) to 72.3 eV (n = 13) along with a particular shift of more than 0.6 eV between n = 12 and 13. However, our Al 2p spectra exhibit a cluster-size dependence within only 0.3 eV for n = 7–24, and no particular peak shift can be observed around n = 13. It should be noted here that the peaks in the C 1s XPS spectra for the Aln NCs on the C60 and HB-HBC/HOPG substrates exhibit a similar contrast shift; namely a decrease in the BE for the Aln NCs on C60 (−0.30 eV) and an increase in the BE for the Aln NCs on HB-HBC (+0.25 eV), with a small size-dependent shift being observed (see Supplementary Fig. 7).

Oxidative reactivity of Aln on the HB-HBC substrate

As shown in Fig. 1b, the Aln NCs deposited on HB-HBC are oxidized upon O2 exposure, and the oxidative rates are dependent on the NC size. Figure 2 shows the Al 2p XPS spectra for the Al13 on HB-HBC at several different O2 exposure amounts (i.e., 0–5 × 1010 L), where the O2 exposure amounts (in Langmuir units, L = 1.33 × 10−4 Pa·s)) are noted on the right-hand side in of the figure. With increasing the amount of O2 exposure, the intensity of the peak corresponding to the zerovalent Al0 component decreases, while that of the oxidized component Al3+ increases along with that of the O 1s component. The oxidative reactivity can therefore be quantitatively evaluated on the basis of its dependence on the O2 exposure amount from 0 to 1 × 104 L. It should be noted here that at 1 × 104 L O2, the Al0 component survives only in the case of n = 13 (Supplementary Fig. 5), which is peculiarly unreactive compared to NCs of other sizes and with Al single crystal surfaces, which are completely oxidized when exposed to 400 L O2 at room temperature44. Furthermore, both the Al 2p and O 1s peaks shift to a lower BE when the O2 exposure amount is increased from 1 × 104 to 5 × 1010 L, thereby implying that a structural change relevant to a phase transition from amorphous to crystalline Al2O3 takes place, such as the formation of α- or γ-Al2O345.

Fig. 2: Oxidative behaviors in the Al 2p and O 1s XPS spectra for Al13 on the HB-HBC substrate.
figure 2

a, b XPS spectra around a Al 2p and (b) O 1s. With increasing O2 exposure (from top red to bottom blue), the intensity of the zerovalent component (Al0) decreases, while those of the oxidized component (Al3+) and the O 1s component increase accordingly. c The oxidative reactivity is evaluated by the slope of the dependence against the logarithmic O2 exposure amount from 0 L to 1 × 104 L. At the highest exposure of 5 × 1010 L, the Al 2p peak shifts to a lower BE, likely due to a structural change relevant to the phase transition of aluminum oxide (see the main text for further details).

Full size image

The chemical reactivity of the Aln NCs toward O2 gas was then evaluated based on the oxidation rate, OAln, which is a simple index for investigating the size-dependent behavior of the oxidation reaction. More specifically, the peak area ratio, RAln, of the non-oxidized component (\({{{{{{\rm{S}}}}}}}_{A{l}^{0}}\)) to the oxidized component (\({{{{{{\rm{S}}}}}}}_{A{l}^{3+}}\)) for the Al 2p spectra is plotted against the logarithm of the O2 exposure amount in L (log10 O2), and the linear slope is evaluated as OAln, where RAln is expressed as follows:

$${R}_{A{l}_{n}}=\,\frac{{S}_{A{l}^{0}}}{{S}_{A{l}^{0}}+\,{S}_{A{l}^{3+}}}$$
(1)

In this analysis, the oxidation of Al atoms by O2 is modeled in terms of the dissociative adsorption of O2 on a single crystal Al surface, whose XPS peak appears at a BE close to that of the Al3+ component (i.e., 75−76 eV)44:

$${{{{{{\rm{A}}}}}}{{{{{\rm{l}}}}}}}_{n}\cdot {x{{{{{\rm{O}}}}}}}_{2}+{{{{{{\rm{O}}}}}}}_{2}\to {{{{{{\rm{Al}}}}}}}_{n}\cdot ({x+1}){{{{{\rm{O}}}}}}_{2},\,\left(x=0,\,1,\,2,\cdots \right)$$
(2)

The oxidation rates, OAln, are evaluated by considering the conversion of 1 L → 1s because the exposure amount can be converted to the corresponding reaction time for elementary reactions. The intersection of the linear line with the x-axis in Fig. 2c gives the O2 exposure amount of VAl13(O2) that is required to completely oxidize the Aln NCs; for Al13, 3.45 × 104 L O2 is obtained as the value of VAl13(O2). The higher the reactivity, the smaller the quantity of oxygen required to completely oxidize the Aln NCs; for example, VAln(O2) at n = 12 is 2.36 × 103 L O2, thereby showing that Al12 is 14.6 times more reactive than Al13 (further details regarding the OAln and VAln(O2) values can be found in Supplementary Note 4 and Supplementary Table 1).

Figure 3 shows the size dependence of the oxidative rates on the Aln NCs (n = 7–24) deposited on the HB-HBC substrate, where the relative reactivity is evaluated by dividing the VAln(O2) value at n = 13 by each individual VAln(O2) value. A local minimum is clearly found at n = 13, and a small local minimum is also found at n = 19, while an even–odd alternation relevant to spin conservation46 observed in the gas phase reaction17,30,47,48 is not obvious. According to previous experimental and theoretical works17,30,38,49,50,51, electronically stabilized Aln anions should appear at n = 19 and 23 as well as at n = 13. Since its interactions with HB-HBC induces an anionic character in the deposited Aln NCs, Al13 and Al19 complete their 2P (40 e−) and 1G (58 e−) shells, respectively. However, such stabilization was not observed for Al23 despite this species completing its 3S (70 e−) shell (see Fig. 3), and this was attributed to the fact that Al23 is geometrically deformed on the substrate owing to its relatively low rigidity having structural Cs symmetry50,51,52.

Fig. 3: Size dependent relative reactivities of Aln (n = 7–24) and B@Al12 on the HB-HBC substrate against the O2 exposure.
figure 3

The oxidative reactivity rates for Aln on the HB-HBC substrate are plotted (red open circles), and a clear local minimum is observed at n = 13 along with a small minimum at n = 19, while there is no apparent local minimum at n = 23. B@Al12 shows a low oxidative reaction rate (red solid square), similar to that of Al13.

Full size image

Oxidative reactivity of B@Al12 on the HB-HBC substrate

The structural rigidity of icosahedral Al13− is demonstrated by the boron (B) doped B@Al12− SAs. Boron belongs to the same group as Al in the periodic table, and it has been reported B@Al12− can be preferentially formed as an SA both experimentally and theoretically because the isoelectronic and geometrically small B atom facilitates relaxation of the icosahedral geometric strain when used as a central atom37,38,53,54. Thus, using a B-mixed Al target, B@Al12− was formed by MSP and was deposited onto the C60 and HB-HBC substrates (Supplementary Fig. 8).

Figure 4a, b show the XPS spectra around the Al 2p core levels for the B@Al12 deposited on C60 and the B@Al12 deposited on HB-HBC, respectively, before (lower) and after (upper) O2 exposures, similar to the spectra in Fig. 1. These XPS spectra show that Al atoms on C60 are substantially oxidized without O2 exposure, but the tailing peak in the Al0 region implies that some Al atoms survive without oxidation. In contrast, the Al atoms of B@Al12 deposited on HB-HBC are not oxidized in the same manner as those of Al13, as shown in Fig. 1b.

Fig. 4: XPS spectra for B@Al12 on the C60 and HB-HBC substrates.
figure 4

a–d XPS spectra around the core Al 2p levels for a B@Al12 on C60 and b B@Al12 on HB-HBC before (lower) and after (upper) O2 exposure (at 5 × 1010 L), along with c, d O 1s for each state. e, f XPS spectra around the B 1s core levels for e B@Al12 on C60 f B@Al12 on HB-HBC before (top) and after (lower) O2 exposure. g, h XPS spectra around the C 1s core levels of the underlying g C60 or h HB-HBC are also shown for the depositions of B@Al12 and Al13. The reference binding energies (BEs) of Al 2p and B 1s for the bulk Al and B (Al0/B0 and Al3+/B3+) and O 1s (O2−) are marked by vertical bars. The BEs for Al 2p show the presence of zerovalent Al0 only for Al13 on HB-HBC before O2 exposure, while the other BEs are in the vicinity of Al3+, indicating the presence of oxidized Al atoms. After the deposition of 0.6 ML B@Al12 (violet) and Al13 (blue) on C60, the C 1s peak in (g) shifts by ~0.3 eV toward a lower BE from that before deposition (light blue), showing an anionic C60− state. After the deposition of 0.6 ML B@Al12 (orange) and Al13 (red) on HB-HBC, the C 1s peak in (h) shifts by ~0.25 eV toward a higher BE than that before deposition (pink), showing a cationic HB-HBC+ state. Importantly, with B atom doping, B@Al12 is stabilized even in the cationic form, as shown by the tailing peak of Al 2p in (a) and the non-oxidized B0 component in (e).

Full size image

Upon O2 exposure, the Al 2p XPS peak for the Al atoms deposited on C60 becomes shaper upon oxidation, while the Al atoms on HB-HBC are sequentially oxidized to Al3+. As shown in Fig. 4c, d, the corresponding O 1s component can be observed in the lower trace of Fig. 4c even without O2 exposure, but the peak intensity is lower than that observed for the Al13 on C60 (see Fig. 1c). In fact, the intensity of the O 1s peak increases with O2 exposure, as shown in Fig. 4c. These results show that the B@Al12 NCs on C60 are reactive, but that the oxidation rate is surpressed because of the geometrical stabilization induced by B atom encapsulation.

As shown in Fig. 4e, f, the B 1s XPS spectra show the effect of such B atom encapsulation. More specifically, despite a similar oxidative reactivity between the Al and B atoms55,56, the B 1 s peak for the B@Al12 on C60 shows that a non-oxidized B0 component can be observed even for the nascent B@Al12 on C60, showing that the oxidation of B atoms to achieve the B3+ state is significantly slower than the corresponding oxidation of Al atoms under O2 exposure. More importantly, the B 1s peak for the B@Al12 on HB-HBC can be observed at an O2 exposure amount up to ~1 × 104 L, at which point the majority of Al atoms are oxidized. Furthermore, Fig. 4g, h show the C 1s XPS spectra of B@Al12 on the C60 and HB-HBC substrates, respectively, wherein a behavior similar to that of Al13 deposition can be observed. More specifically, for the B@Al12 on C60, the C 1s peak (Fig. 4g) shifts toward a lower BE by ~0.25 eV, while for the B@Al12 on HB-HBC, the C 1s peak (Fig. 4h) shifts toward a higher BE by ~0.25 eV, suggesting the formation of a B@Al12−/HB-HBC+ CT complex.

When the oxidation rate of B@Al12 is similarly evaluated based on the peak area ratio of the non-oxidized component (\({{{{{{\rm{S}}}}}}}_{A{l}^{0}}\)) to the oxidized component (\({{{{{{\rm{S}}}}}}}_{A{l}^{3+}}\)) (see Supplementary Fig. 9), the OB@Al12 value is the same with the OAl13 value within experimental uncertainties, resulting in similar VB@Al12(O2) and VAl13(O2) values, as plotted in Fig. 3. Upon B atom encapsulation, all Al atoms become surface Al atoms of the Al12 cage, while in contrast, Al13 consists of twelve surface Al atoms and one central Al atom. The same oxidative rates observed for B@Al12 and Al13 therefore indicate that B@Al12 is more robust because these equivalent rates were obtained despite the contribution of the central Al atom of Al13.

Theoretical calculations on the charge distributions for the 13-mer anions and cations

For Al13−, B@Al12−, Al13+, and B@Al12+, although theoretical calculations have been reported by several groups36,38,57,58, density functional theory (DFT) calculations are collectively performed to explain the different oxidation behaviors observed for Al13/B@Al12 on the C60 and HB-HBC substrates. The results are presented in Fig. 5, and the Cartesian coordinates are summarized in Supplementary Table 4. For the equilibrium structures, the averaged Al–Al bond lengths are 0.2794 nm for icosahedral Al13− and 0.2675 nm for icosahedral B@Al12−. The shortened Al–Al bond in B@Al12− is ascribed to relaxed geometric strains due to the presence of a small central B atom inside the Al12 cage. For both Al13+ and B@Al12+, the structural symmetry is lowered, giving C1 symmetry for Al13+ and Ci symmetry for B@Al12+, and this was attributed to the electron deficiency of 2P shell closure.

Fig. 5: Calculated NPA charge distributions for Al13 and B@Al12.
figure 5

a–d Natural population analysis (NPA) distributions for a Al13−, b Al13+, c B@Al12−, and d B@Al12+ for the optimized structures using PBE0 with 6-311+G(d) for Al13− and B@Al12− or with 6-311G(d) for Al13+ and B@Al12+. Along with the representative values, the charge amount is expressed by the color gradation: positive in red and negative in blue. The icosahedral Ih symmetries for Al13− and B@Al12− are lowered to C1 for Al13+ and Ci for B@Al12+ owing to electron deficiency in the 2P shell. In general, a central Al/B atom is negatively charged, while the surface Al atoms are positively charged, with the exception of a top Al atom in the distorted Al13+ in (b).

Full size image

In terms of the charge distributions of Al13−/B@Al12− and Al13+/B@Al12+, natural population analysis (NPA) shows that the central Al/B atom is negatively charged, while the surface Al atoms (ρ (Al)) have a positive charge of +0.06/+0.15 for Al13−/B@Al12−. Compared to the central Al atom (ρ (Al) = −1.67), the central B atom is more negatively charged (ρ (B) = −2.68), and the surrounding twelve Al atoms (ρ (Al) = +0.15) are more positively charged than those in Al13− (ρ (Al) = +0.06). For Al13+/B@Al12+, the positive charges are delocalized over all Al atoms in the cluster, with the exception of one negatively charged Al atom. Therefore, these theoretical calculations show common electronic features wherein the negatively charged central atoms are masked by the surrounding Al atoms.

Discussion

Charge state of the Al13/B@Al12 deposited on a p-type substrate and reaction mechanism

The charge states of the deposited Al13/B@Al12 SAs were found to be significantly influenced by the cluster–surface interactions, which in turn are affected by the molecular character of the organic substrate. More specifically, the p-type organic substrate of HB-HBC was found to electronically stabilize the halogen-like Al13/B@Al12 NCs on the surface by donating an electron, which led to electron shell closure.

The ultraviolet photoelectron spectroscopy (UPS) reveals the electronic states of the organic substrates (see the UPS spectrum for HB-HBC in Supplementary Fig. 10). More specifically, before the deposition of Aln, the HOMO energies of the C60 and HB-HBC are 2.3 eV59 and 1.6 eV (at the peak maximum) below the Fermi level (EF), respectively. In addition, the LUMO levels can be accessed by two-photon photoemission spectroscopy; LUMO energies of C60 and HB-HBC are 0.7 eV59 and 1.4 eV above the EF for C60 and HB-HBC, respectively (see Supplementary Fig. 11). These HOMO and LUMO energies indicate that C60 (HB-HBC) can be regarded as n-type (p-type) substrates, wherein C60 accepts an electron to its LUMO, while HB-HBC donates an electron from its HOMO. Indeed, the quantitative evaluations carried out for the energetics of CT complexation between Al13/B@Al12 and C60/HB-HBC reasonably explain the formation of Al13+C60−/B@Al12+C60− and Al13−HB-HBC+/B@Al12−HB-HBC+ (see Supplementary Note 5 and related contents, i.e., Supplementary Fig. 12, Supplementary Table 2, and Supplementary Table 3).

In the context of O2 chemisorption on the surfaces, the adsorbed O2 molecules with two unpaired electrons accept an electron from the surface, forming superoxide (O2−) or peroxide (O22−) ions60,61. Furthermore, the adsorption energy of two O atoms is larger than the dissociation energy of a single O2 molecule62, and therefore, O atoms are preferentially bound to the surface via a dissociative electron attachment process61. When O2 molecules react with CT complexes on a substrate, the O2 molecules preferentially attack the electron-rich sites of the anions. At the deposition of Al13−/B@Al12− SAs onto the C60 and HB-HBC substrates, as mentioned above, an electron transfer takes place to form Al13+C60−/B@Al12+C60− on C60 and Al13−HB-HBC+/B@Al12−HB-HBC+ on HB-HBC. Comparing the electron affinities (EAs) of the C60 (2.68 eV)63 and Al13/B@Al12 SAs (3.1–3.6 eV)33,64,65, it is easier to transfer an electron from C60− to O2 than from Al13−/B@Al12−. In other words, C60−, which is generated by the deposition of Al13/B@Al12, facilitates the dissociative electron attachment of O2, resulting in the immediate oxidation of Al13+/B@Al12+ cations with O2− or O− through a Coulombic attraction. Therefore, the SA nature of Al13−/B@Al12− is reinforced by p-type molecular decoration, which renders it possible to fabricate assembled surfaces of chemically robust Al-based SAs.

To conclude, we have successfully characterized the series of Aln NCs deposited on an n-type C60 and p–type HB-HBC substrates. The XPS results reveal that the n-type C60 substrate possessing a high EA withdraws an electron from the Aln NCs, resulting in a deviation from the electron shell closure. In contrast, the p-type HB-HBC substrate donates one electron to the Aln NCs, generating electronically stable Al13−/B@Al12− SAs (40 e−). The chemical stabilities of the deposited Aln examined by step-by-step O2 exposure are shown to be significantly influenced by their charge states on the surface, wherein the stability is enhanced in the 40 e− systems of Al13/HB-HBC and B@Al12/HB-HBC along with icosahedral rigidity.

Overall, we have demonstrated the importance of optimizing the cluster–surface interactions to achieve stable depositions of Al13−/B@Al12− SAs. It has also been demonstrated that the molecular decoration of a substrate aids in controlling the local electronic state through the generation of such cluster–surface interactions. We believe that this molecular strategy for the stable deposition of Al13/B@Al12 could facilitate the fabrication of SA assemblies for all functional SAs generated in the gas phase.

Methods

Sample preparation

The samples of Aln or AlnBm NCs deposited on organic C60 and HB-HBC substrates were prepared in an integrated vacuum chamber, including an MSP source, NC deposition, organic evaporation, and photoelectron spectroscopy systems16,40,41. The organic C60 and HB-HBC substrates were prepared on cleaned HOPG by thermal evaporation in ultrahigh vacuum (UHV) conditions (<3 × 10−8 Pa). The thicknesses were controlled at 2 and 5 MLs for C60 and HB-HBC, respectively, and were monitored using a quartz crystal microbalance. Commercially available C60 (Aldrich, sublimed, 99.9%) was used, while HB-HBC was synthesized (see Supplementary Note 1)66.

Anionic Al clusters (Aln−) were generated using an MSP system (Ayabo Corp. nanojima-NAP-01)25, in which the Al targets were sputtered with Ar+ ions in the MSP aggregation cell. After clustering atomic Al vapors into Aln− in a cooled (77 K) He gas flow, the Aln− NCs were introduced into a quadrupole (Q) mass filter (Extrel CMS; MAX-16000) through ion optics. The production conditions were optimized by monitoring the mass spectra of Aln− (see Supplementary Fig. 2) to maximize the ion intensities at the chosen m/z ratios. The mass-selected Aln− NCs were then deposited on the C60 and HB-HBC substrates with a mass resolution of m/Δm ~70, which was sufficient to exclude the co-deposition of minor products with neighboring m/z values (see Supplementary Fig. 2). The collision energy of the Aln− ions was controlled by applying a bias voltage to the substrates (typically +5 V), satisfying the soft-landing conditions (<10 eV/cluster). The number of deposited Aln− ions was counted as 2.9 × 1013 clusters, where the coverage of Aln on the substrates was estimated as 0.6 MLs, assuming a deposition area of 2.8 × 1013 nm2 (6 mm in diameter) and an Aln size estimated by a cubic-root interpolation between the sizes of the Al atom (n = 1) and the icosahedral Aln (n = 13 and 55) (i.e., 0.62 nm for n = 7 and 0.98 nm for n = 24 in diameter). The estimated coverage was verified by XPS and UPS measurements with the step-by-step deposition of NCs40,41. The deposited samples were transferred to the photoelectron spectroscopy system connected to the cluster deposition system while maintaining UHV conditions. More detailed procedures for sample preparation were described in Supplementary Note 2.

Photoelectron spectroscopy

XPS measurements were performed using an Mg Kα (hν = 1253.6 eV) X-ray source. Photoelectrons emitted from the sample surface were collected with a hemispherical electron energy analyzer (VG SCIENTA, R3000) at a detection angle of 45° from the surface normal. The BE was calibrated using the Au 4 f core level (84.0 eV). It was ensured that no charging effect was observed during any of the XPS measurements. In the XPS analyses, after subtracting the Shirley background, peak fitting was performed by instrumental broadening determined from the Au 4f peak profile (Voigt function with a full width at half maximum (FWHM) of 1.09 eV; the Gaussian and Lorentzian widths were 0.75 and 0.56 eV, respectively). A He–I discharge lamp (hν = 21.22 eV) was used for the UPS measurements.

To examine the oxidative reactivities of the deposited Aln NCs, the samples were exposed to O2. The amount of O2 exposure was defined as Langmuir units (L = 1.33 × 10−4 Pa·s). The O2 gas was introduced into the XPS/UPS system using a variable leak valve for low exposure levels (≤104 L). At higher exposure levels (>1010 L), the sample was exposed to O2 in a UHV chamber isolated from the XPS/UPS system. All XPS/UPS measurements and O2 exposures were performed at room temperature.

Density functional theory (DFT) calculations

Geometry optimizations for the Al13−, B@Al12−, Al13+, and B@Al12+ cluster ions with singlet spin states were performed by DFT implemented in the Gaussian 16 program67. All equilibrium geometries were optimized until no imaginary frequencies were found. The hybrid exchange-correlation function PBE068,69 was employed at 6-311+G(d) for Al13− and B@Al12− and at 6-311G(d) for Al13+ and B@Al12+. Population analyses were performed using NPA70 for the total electron density obtained at the same level of DFT calculations.