Deposition of Nanosized Amino Acid Functionalized Bismuth Oxido Clusters on Gold Surfaces

Bismuth compounds are of growing interest with regard to potential applications in catalysis, medicine, and electronics, for which their environmentally benign nature is one of the key factors. One thing that currently hampers the further development of bismuth oxido-based materials, however, is the often low solubility of the precursors, which makes targeted immobilisation on substrates challenging. We present an approach towards the solubilisation of bismuth oxido clusters by introducing an amino carboxylate as a functional group. For this purpose, the bismuth oxido cluster [Bi38O45(NO3)20(dmso)28](NO3)4·4dmso (dmso = dimethyl sulfoxide) was reacted with the sodium salt of tert-butyloxycabonyl (Boc)-protected phenylalanine (L-Phe) to obtain the soluble and chiral nanocluster [Bi38O45(Boc–Phe–O)24(dmso)9]. The exchange of the nitrates by the amino carboxylates was proven by nuclear magnetic resonance, Fourier-transform infrared spectroscopy, as well as elemental analysis and X-ray photoemission spectroscopy. The solubility of the bismuth oxido cluster in a protic as well as an aprotic polar organic solvent and the growth mode of the clusters upon spin, dip, and drop coating on gold surfaces were studied by a variety of microscopy, as well as spectroscopic techniques. In all cases, the bismuth oxido clusters form crystalline agglomerations with size, height, and distribution on the substrate that can be controlled by the choice of the solvent and of the deposition method.


Film Preparation
The bismuth oxido cluster 2 was dissolved in ethanol as a protic, polar solvent and for comparison in acetonitrile as an aprotic, polar solvent. A concentration of 20 g·L −1 was applied to all samples. As substrates, 10 mm × 10 mm pieces of silicon wafers with a 150 nm silicon oxide layer, a 20 nm sputtered titanium layer as an adhesion layer, and 100 nm gold (Au) were used.
Drop coating was performed by covering the substrate with 20 µL of solution and the evaporation of the solvent under ambient conditions. Spin coating was performed using a Laurell WS-650MZ-23NPPB spin coater (North Wales, USA) with a 4.5 mm vacuum chuck and a pressure of 1 mbar. The coater was operated with a rotation time of 30 s at a rotation speed of 2000 rpm. Moreover, 20 µL of the solution was dropped onto the substrate before the rotation was started.
The dip coating experiments were performed using a SOLGELWAY ACEdip 2.0 dip coater (Sceaux, France). The samples were dipped into 3 mL of the solution at a constant speed of 5 mm·s −1 for 100 s. Afterwards, the solution was removed with a constant speed of 2 mm·s −1 and dried under ambient conditions. The parameters used for the ethanol and acetonitrile samples are shown in Table 1, where the sample name is composed of compound 2, the deposition technique and the solvent, respectively.

Characterization Methods
Optical microscopy imaging was performed with a Nikon IC inspection Microscope ECLIPSE L200 from Nikon Metrology GmbH (Düsseldorf, Germany) with 1× as well as 50× objectives. Atomic force microscopy (AFM) measurements were performed with an atomic force microscope (NanoWorld Arrow TM NCPt AFM, supplied by NanoAndMore GmbH, 35578 Wetzlar, Germany) operating in intermediate topography mode using a PtIr 5 cantilever, with resonant frequencies in the range of (240-380) kHz. The analysis of the surface coverage was performed either using Image J (version 2.5.1) [43] (for the optical microscopy images) or Gwyddion software (version 1.52) [44] (for the AFM images).
FTIR spectroscopy was performed in the wavenumber range of (450-5000) cm −1 using a VERTEX 80v FTIR spectrometer with an attenuated total reflectance (ATR) unit from Bruker, Germany. The XPS analysis was performed with an ESCALAB 250Xi photoelectron spectrometer from Thermo Fisher Scientific TM (Waltham, MA, USA)in an ultra-high vacuum (UHV) chamber using a monochromatic Al-Kα (1486.68 eV) X-ray source and a beam diameter of 300 µm. The binding energies of all spectra were referenced to the binding energy of C1s (248.8 eV). UV-vis spectra were measured in the diffuse reflection geometry at ambient temperature with an Agilent Technologies Cary 60 UV-vis spectrometer in a range from 200 nm to 800 nm using an optical probe. The data were collected with the Cary WinUV Software (version 5.0.0.1005, Santa Clara, CA, USA). Powder X-ray diffractograms were measured at ambient temperature with an STOE Stadi P diffractometer (Darmstadt, Germany) using Ge(111)-monochromatized Cu-K α radiation (1.54056 nm, 40 kV, 40 mA). The full width at half maximum (FWHM) is corrected for instrumental broadening using a LaB 6 standard (SRM 660, Gaithersburg, MD, USA) purchased from the National Institute of Standards and Technology (NIST). The value β was corrected using β 2 = β measured 2 -β instrument 2 , where β measured and β instrument are the FWHMs of measured and standard profiles, respectively. Particle size distribution (PSD) based on dynamic light scattering (DLS) was determined using a Zetasizer Nano ZS (Malvern Instruments, Worcestershire, UK, WR14 1XZ), allowing the characterization of suspension in a size range of 0.6 nm to 6 µm. A red laser (633 nm, 4 mW) was used as a light source, and the analyses were performed at an angle of 173 • (backscatter NIBS default). The compounds were dissolved in appropriate solvents (20 g·L −1 ), filtered, filled into glass cuvettes (DTS0012), and measured at 20 • C. Calculation of the PSD was carried out according to the "Mie theory", assuming the presence of spherical particles. 1 H and 13 C{ 1 H} NMR spectra were recorded at room temperature in dmso-d 6 (dried over 4 Å molecular sieve) with an Avance III 500 spectrometer (BRUKER, Billerica, MA, USA) at 500.30 MHz and 125.80 MHz, respectively, and were referenced internally to the deuterated solvent relative to Si(CH 3 ) 4 (δ = 0.00 ppm). Scanning electron microscopy was performed using a NanoNovaSEM200 (Thermo Fisher Scientific, Waltham, MA, USA) device with an electron beam energy of 5 keV and different magnifications.  24 (dmso) 9 ] (2) was obtained after a few days of crystallisation under ambient conditions as a colourless solid with yields > 90%. The elemental analysis confirms the stoichiometry of compound 2; complete ligand substitution was proven using NMR in combination with IR spectroscopy. The 1 H-NMR spectrum reveals all expected signals for the Boc-Phe-O − ligand and is indicative of dmso coordination to the bismuth oxido cluster in solution (dmso-d 6 ), showing a characteristic signal at δ = 2.54 ppm. The IR spectrum shows all expected vibrations of the Boc-protected amino acid-based ligands (Figure S3a). Compared to the sodium salt 1, the bismuth oxido cluster 2 shows a characteristic broad Bi-O vibration at 580 cm −1 [45]. The absence of vibrations at approximately 1740 cm −1 (ν(N=O) free ) for free nitrates, at 1430 cm −1 (ν as. (NO 2 )), 995 cm −1 (ν sym. (NO 2 )) for monodentate, and at 1265 cm −1 (ν sym. (NO 2 )), 1385 cm −1 (ν as. (NO 2 )) for bidentate nitrates, demonstrates the complete exchange of the nitrate ligands ( Figure S3b, Table S1) [13,46]. The symmetric valence S=O vibration at 948 cm −1 is indicative of dmso in the resulting compound, which is in line with the NMR results [47]. The presence of Boc-Phe-OH is confirmed by the δ C-H vibration of the tert-butyl group at 1386 cm −1 and 1363 cm −1 , the δ C=C-H vibration for a monosubstituted aromatic ring at 699 cm −1 , 753 cm −1 , and 1018 cm −1 and the ν C-N , as well as the ν C-O at 1162 cm −1 and 1047 cm −1 , respectively [45,48,49]. All these vibrations are only slightly shifted compared to Boc-Phe-OH and the sodium salt 1, however, the symmetric and asymmetric ν C=O at 1465 cm −1 and 1554 cm −1 show significant shifts. In particular, the asymmetric vibration in 2 is shifted to lower wavenumbers by approximately 30 cm −1 compared to 1 and by 90 cm −1 compared to Boc-Phe-OH, which indicates different coordination to the cluster core.
The preservation of the [Bi 38 O 45 ] cluster core upon the ligand exchange reaction is confirmed using PXRD ( Figure S4), which shows a typical diffraction pattern for bismuth oxido nanoclusters, showing one intensive diffraction peak and a much weaker one for 2θ < 10 • [18,27]. Compared to the reflections of [Bi 38 O 45 (NO 3 ) 20 (dmso) 28 ](NO 3 ) 4 ·4dmso (2θ = 5.23 • and 2θ = 6.24 • ), the main diffraction peak of 2 is shifted to a lower angle at 2θ = 4.47 • , indicating an increase in the cluster size of 2 due to the larger ligand shell, compared with the nitrate cluster. The previously reported [Bi 38 O 45 (Boc-Phe-OH) 22 (OH) 2 ] shows similar diffraction peaks, with 2θ = 4.51 • and 2θ = 5.19 • [26] The main reflection corresponds to the interlayer distance, assuming a closed, packed nanocluster arrangement [20,50]. Thus, an interlayer distance for 2 of 1.975 nm (2θ = 4.47 • ) was calculated using Bragg's equation. Compared to the precursor [Bi 38 O 45 (NO 3 ) 20 (dmso) 28 ](NO 3 ) 4 ·4dmso (A) with its main reflection at 2θ = 5.23 • (d = 1.688 nm), the resulting interlayer distance increases according to the molecule size. In dmso solution, the mean hydrodynamic diameter was observed in the range of d h = (2.2-2.9) nm and of d h = (1.2-1.5) nm for compounds 2 and A, respectively. The hydrodynamic diameter was determined from the fitted particle size distribution (PSD) curve based on dynamic light scattering (DLS); the exemplarily results for each solvent are shown in Figure S5. The DLS analysis for the nitrate-substituted bismuth oxido cluster A is restricted to the usage of dmso as a solvent, as a result of its low solubility, whereas cluster 2 is soluble in various organic solvents. Using ethanol as a solvent, the resulting hydrodynamic diameter of 2 is in the range of d h = (2.2-2.9) nm and of d h = (1.4-1.6) nm, when acetonitrile was used. In the former case, we assume that the bismuth oxido cluster is solvated by the ethanol molecules, providing an increased periphery. In acetonitrile, the partial dissociation of coordinating ligands and solvent, as well as the partial hydrolysis and partial deprotection of the Boc-group, are assumed to give smaller cluster molecules and an asymmetrical molecule shape, which results in a reduced hydrodynamic diameter. Notably, the solubility of 2 is twice as high in ethanol (β = 65 g·L −1 ) as in acetonitrile (β = 30 g·L −1 ). However, the cluster core of 2 is stable in both solvents, which makes it suitable for the investigation of solution-based deposition Nanomaterials 2022, 12, 1815 6 of 17 methods. The influence of solubility as well as of the stabilization and partial dissociation of cluster 2 in solution onto the growth mode of the films on Au is discussed in the following.
In order to test the chemical and structural integrity of compound 2 upon dissolving in various solvents followed by deposition (dip coating, spin coating and drop coating) onto the Au-coated substrates, XRD, ATR-FTIR, and XPS measurements were performed. In the following, the observations made upon dip coating are exemplarily discussed.
The grazing incidence XRD results of the samples deposited by dip coating are shown in Figure 1, and compared to the reference PXRD pattern for 2 measured in the powder diffraction mode. A small shift in the reflections might be expected, considering a slightly different crystal packing in the powder and in the films. The XRD patterns over the full range from 2θ = 3 • -85 • are dominated by Au(111), Au(222), Au(311), and Si(400) reflections. For 2-DiE and 2-DiA, only one strong reflection at 2θ = 4.38 • and 2θ = 4.35 • , respectively, is observed, which corresponds to the main reflection peak of 2 at 2θ = 4.47 • . Thus, we conclude that the deposition of compound 2 was successful.
the former case, we assume that the bismuth oxido cluster is solvated by the ethanol molecules, providing an increased periphery. In acetonitrile, the partial dissociation of coordinating ligands and solvent, as well as the partial hydrolysis and partial deprotection of the Boc-group, are assumed to give smaller cluster molecules and an asymmetrical molecule shape, which results in a reduced hydrodynamic diameter. Notably, the solubility of 2 is twice as high in ethanol (β = 65 g·L −1 ) as in acetonitrile (β = 30 g·L −1 ). However, the cluster core of 2 is stable in both solvents, which makes it suitable for the investigation of solution-based deposition methods. The influence of solubility as well as of the stabilization and partial dissociation of cluster 2 in solution onto the growth mode of the films on Au is discussed in the following.
In order to test the chemical and structural integrity of compound 2 upon dissolving in various solvents followed by deposition (dip coating, spin coating and drop coating) onto the Au-coated substrates, XRD, ATR-FTIR, and XPS measurements were performed. In the following, the observations made upon dip coating are exemplarily discussed.
The grazing incidence XRD results of the samples deposited by dip coating are shown in Figure 1, and compared to the reference PXRD pattern for 2 measured in the powder diffraction mode. A small shift in the reflections might be expected, considering a slightly different crystal packing in the powder and in the films. The XRD patterns over the full range from 2θ = 3°-85° are dominated by Au(111), Au(222), Au(311), and Si (400) reflections. For 2-DiE and 2-DiA, only one strong reflection at 2θ = 4.38° and 2θ = 4.35°, respectively, is observed, which corresponds to the main reflection peak of 2 at 2θ = 4.47°. Thus, we conclude that the deposition of compound 2 was successful. XPS measurements were carried out to analyse the chemical composition of the films. The survey spectrum (cf. Figure 2a) of a representative sample deposited by dip coating XPS measurements were carried out to analyse the chemical composition of the films. The survey spectrum (cf. Figure 2a) of a representative sample deposited by dip coating from acetonitrile solution (2-DiA) reveals the typical signals for bismuth (Bi 5d, Bi 5p 3/2 , Bi 5p 1/2 , Bi 4f, Bi 4d 5/2 , Bi 4d 3/2 , Bi 4p 3/2 , Bi 4p 1/2 ), oxygen (O 1s) as well as carbon (C 1s) [47,51,52]. A peak at 399.7 eV arising from nitrogen N 1s in the amide groups [53] was observed in the high resolution spectrum (Figure 2b). By contrast, the N 1s peak for the starting material [Bi 38 O 45 (NO 3 ) 20 (dmso) 28 ](NO 3 ) 4 ·4dmso (A) is observed at 405.8 eV (cf. Figure 2b), which indicates the complete ligand exchange from nitrate to the Boc-protected amino acid carboxylate [46]. In the high resolution spectrum for O 1s, the main peak at 531.4 eV is associated to the C−O bond; the satellite peak at a higher binding energy of 532.9 eV corresponds to the C=O bond (cf. Figure 2c) [54]. The peak at the lower binding energy of 530.1 eV stems from Bi-O bonds in the cluster core [55] (Figure 2c), which proves the existence of the bismuth oxido core after the deposition on the Au-coated substrate. In Figure 2d, the high-resolution spectrum for C 1s is shown. The main peak corresponds to the sp 3 hybridized carbon of the Boc-protecting group, as well as the CH 3 group in Phe and dmso. ure 2b), which indicates the complete ligand exchange from nitrate to the Boc-protected amino acid carboxylate [46]. In the high resolution spectrum for O 1s, the main peak at 531.4 eV is associated to the C−O bond; the satellite peak at a higher binding energy of 532.9 eV corresponds to the C=O bond (cf. Figure 2c) [54]. The peak at the lower binding energy of 530.1 eV stems from Bi-O bonds in the cluster core [55] (Figure 2c), which proves the existence of the bismuth oxido core after the deposition on the Au-coated substrate. In Figure 2d, the high-resolution spectrum for C 1s is shown. The main peak corresponds to the sp 3 hybridized carbon of the Boc-protecting group, as well as the CH3 group in Phe and dmso.  Table next to it. FTIR data for spin and drop-coated samples are shown in Figure S6 and the assigned vibrations are listed in Table S2. The FTIR spectra of 2-DiA and 2-DiE along with that of 2 are shown exemplarily in Figure 3 and the assigned vibrations are listed in the Table next to it. FTIR data for spin and drop-coated samples are shown in Figure S6 and the assigned vibrations are listed in Table S2.
For all samples, intensive vibrational bands were observed in the range from 570 cm −1 -1700 cm −1 . The broad band at approximately 582 cm −1 is assigned to the stretching vibration of Bi-O [45], which is slightly broadened and shifted to higher wavenumbers in the films compared to the powder of 2. In contrast, the bands assigned to the Bocprotected amino acid do not show significant changes. The bands observed at 700 cm −1 , 753 cm −1 and 1020 cm −1 correspond to deformation vibrations of the mono-substituted aromatic system from the Phe unit [45,48]. The intense vibrational modes at 1363 cm −1 and 1386 cm −1 belong to the symmetric bending mode of C−H (methyl) groups [50]. The bands appearing at 1495 cm −1 and 1554 cm −1 correspond to the symmetric C=O stretching of carboxylate and at 1688 cm −1 to the C=O stretching of amide. The aromatic C−H stretching vibrations above 3000 cm −1 and the aliphatic C−H stretching vibration below 3000 cm −1 can be assigned to the Boc-Phe-O − ligands [50]. The characteristic symmetric S=O vibration band of compound 2 located at approximately 948 cm −1 is not present in the films [47]. We conclude that dmso as solvate of compound 2 is removed upon dissolving in ethanol or acetonitrile, respectively, due to the very low concentration of dmso in the resulting solutions. This behaviour was recently reported for [Bi 38 O 45 (OMc) 24 (dmso) 9 ]·2dmso·7H 2 O, for which the prior solvate dmso was displaced by appropriate alcohol solvate molecules according to the solvent used for recrystallization (EtOH or i PrOH) [56].
In conclusion, 2 was solubilized in the considered solvents and deposited from solution while retaining the main composition and cluster structure, except for the substitution of coordinating solvent dmso molecules, whose vibrational signature is not observed in the deposited films. In the following, we discuss the macroscopic growth mode of the clusters on gold surfaces using optical and scanning electron microscopy. Nanomaterials 2022, 12, x FOR PEER REVIEW 8 of 18 For all samples, intensive vibrational bands were observed in the range from 570 cm −1 -1700 cm −1 . The broad band at approximately 582 cm −1 is assigned to the stretching vibration of Bi-O [45], which is slightly broadened and shifted to higher wavenumbers in the films compared to the powder of 2. In contrast, the bands assigned to the Boc-protected amino acid do not show significant changes. The bands observed at 700 cm −1 , 753 cm −1 and 1020 cm −1 correspond to deformation vibrations of the mono-substituted aromatic system from the Phe unit [45,48]. The intense vibrational modes at 1363 cm −1 and 1386 cm −1 belong to the symmetric bending mode of C−H (methyl) groups [50]. The bands appearing at 1495 cm −1 and 1554 cm −1 correspond to the symmetric C=O stretching of carboxylate and at 1688 cm −1 to the C=O stretching of amide. The aromatic C−H stretching vibrations above 3000

Drop-coated Films
The contact angle of the droplets of the investigated solutions on the Au surface varied from 2 • to 8 • , which indicates wetting between the Au substrate and all solvents used. According to ref. [57] three main competing flow patterns can occur in droplets of solutions containing colloidal particles when deposited onto a substrate. On one hand, a fast evaporation at the wetting border of the droplet to the unwetted substrate induces a radial flow from the middle of the droplet towards the border, leading to a ring-like agglomeration of the nanoparticles. On the other hand, the presence of a dominant Marangoni recirculating flow would lead to a strong agglomeration of nanoparticles in the middle of the droplet. Third, the transport of particles towards the substrate driven by Derjaguin-Landau-Verwey-Overbeek (DLVO) interactions may lead either to ring-like patterns or to more homogeneous layers, depending on the pH of the solution.
Depending on the thermophysical properties of the substrate, on the concentration, and on the viscosity as well as on the pH-value of the solution, various structural properties of the films formed by the nanoclusters remaining on the substrate after the solvent evaporation can be achieved. Despite the fact that the phase diagram in ref. [57] was proposed for colloidal nanoparticles with dimensions which are two orders of magnitude larger than those of compound 2, some similarities between the agglomeration patterns observed in this study and those reported in [57] can be observed.
For a better understanding of the difference in growth behaviour on the edges and the middle of the droplet, SEM images are shown in Figure 4 Figure S7. The maximum height of agglomerations of 2 is observed at the borders of the dried droplet in the case of acetonitrile, while compound 2 is more homogeneously distributed in the case of the deposition from the ethanol solution. The tendency for agglomeration can be understood as a consequence of the stronger intermolecular interaction between the bismuth oxido clusters in solution compared to rather weak interactions between 2 and the gold surface [59].
On the edges of the initial droplets (cf. Figures 4b and 5b) rough agglomerations can be observed. Judging from the form and size of the bismuth oxido cluster agglomerations it can be inferred that in the middle of the droplets (cf. Figures 4c and 5c) the dominant growth mode is the coffee stain effect [57] for both solvents discussed here. Due to the low evaporation rate, agglomerations in the middle of the droplet are observed as well (cf. Figures 4c and 5c).     Furthermore, experiments with higher amount of solution (50 µL), covering the whole substrates were made to determine the difference in growth mode shown in Figures  6 and 7 for ethanol (2-DrE-2) and acetonitrile (2-DrA-2), respectively. As can be seen from the microscopy image in Figure 6a block-like structures separated by cracks are observed on the edge of the substrate which is attributed to the removal of solvent residuals during the drying process, and which may lead to hierarchical crack networks [60]. This phenomenon was also observed for annealed films of bismuth oxide [38]. The volume reduction led to surface tensions and microcracking of the films. Additionally, in Figure  6b-e island growth can be observed. In particular, at the edges of the substrate (Figure  6b), the islands occur as white spots in the SEM images, which is an indicator for charged structures and related to the height. The islands occur as a result of the interaction between several nanoclusters of compound 2. Furthermore, experiments with higher amount of solution (50 µL), covering the whole substrates were made to determine the difference in growth mode shown in Figures 6 and 7 for ethanol (2-DrE-2) and acetonitrile (2-DrA-2), respectively. As can be seen from the microscopy image in Figure 6a block-like structures separated by cracks are observed on the edge of the substrate which is attributed to the removal of solvent residuals during the drying process, and which may lead to hierarchical crack networks [60]. This phenomenon was also observed for annealed films of bismuth oxide [38]. The volume reduction led to surface tensions and microcracking of the films. Additionally, in Figure 6b-e island growth can be observed. In particular, at the edges of the substrate (Figure 6b), the islands occur as white spots in the SEM images, which is an indicator for charged structures and related to the height. The islands occur as a result of the interaction between several nanoclusters of compound 2.  In contrast to the growth behaviour observed for 2-DrE-2 from ethanol solution, the one obtained from acetonitrile is indicating diffusion-limited aggregation in the central droplet region, as presented in Figure 7. Fewer cracks occur on the borders as compared to 2-DrE-2, most likely due to the slightly different evaporation rates of the two solvents. Nevertheless, the dominant growth mode in this sample is diffusion-limited aggregation, since this mechanism is known to lead to the formation of characteristic dendritic, floral structures [61]. Such structures can be reproduced by Monte-Carlo simulations of rate since this mechanism is known to lead to the formation of characteristic dendritic, floral structures [61]. Such structures can be reproduced by Monte-Carlo simulations of rate equations for aggregates of atoms, molecules, or ions, and even for defect centres via correlation functions [62], as well as for the wetting of porous media. This occurs when diffusion on the surface is the primary mechanism of mass transport in the system and desorption/adsorption play a minor role [61]. The diffusion leads to the sticking of particles to the nucleation sites at the rim of the aggregates, where the floral-like structure starts growing. The floral-like agglomerations of 2 only occur during the drop and dip coating procedure.  In contrast to the growth behaviour observed for 2-DrE-2 from ethanol solution, the one obtained from acetonitrile is indicating diffusion-limited aggregation in the central droplet region, as presented in Figure 7. Fewer cracks occur on the borders as compared to 2-DrE-2, most likely due to the slightly different evaporation rates of the two solvents. Nevertheless, the dominant growth mode in this sample is diffusion-limited aggregation, since this mechanism is known to lead to the formation of characteristic dendritic, floral structures [61]. Such structures can be reproduced by Monte-Carlo simulations of rate equations for aggregates of atoms, molecules, or ions, and even for defect centres via correlation functions [62], as well as for the wetting of porous media. This occurs when diffusion on the surface is the primary mechanism of mass transport in the system and desorption/adsorption play a minor role [61]. The diffusion leads to the sticking of particles to the nucleation sites at the rim of the aggregates, where the floral-like structure starts growing. The floral-like agglomerations of 2 only occur during the drop and dip coating procedure.

Spin-coated and Dip-coated Films
In the following, a comparison between the films of the bismuth oxido cluster 2 deposited by spin and dip coating techniques is presented.
Optical microscopy images of spin-coated and dip-coated films from ethanol and acetonitrile solutions are shown in Figure 8 with a magnification of x1 as well as SEM images for a better resolution with a magnification of ×2000. While the mechanisms discussed in ref. [57] for the formation of nanoparticle agglomerations can be well applied to describe the morphology of the films grown by drop coating (just for 2-DrE and 2-DrA), the images in Figure 8 indicate that films deposited by spin and dip coating are not subjected to severe concentration gradients during growth.
In the case of spin-and dip-coated samples the coffee stain effect was not observed. Compared to the case of drop coating, a tendency of agglomeration of compound 2 is observable, with the main difference that in the case of spin and dip-coated samples homogeneously distributed agglomerations over the whole substrate are observed (for dip coating-over the whole dipped area). The homogeneous distribution of the agglomerations can be related to the faster spreading of the solution; hence the droplet size is reduced in the case of spin and dip coating. This, in turn, reduces the contribution of the radial flow towards the droplet edge, and hence the radial concentration gradients.
Compared to the case of drop coating, a tendency of agglomeration of compound 2 is observable, with the main difference that in the case of spin and dip-coated samples homogeneously distributed agglomerations over the whole substrate are observed (for dip coating-over the whole dipped area). The homogeneous distribution of the agglomerations can be related to the faster spreading of the solution; hence the droplet size is reduced in the case of spin and dip coating. This, in turn, reduces the contribution of the radial flow towards the droplet edge, and hence the radial concentration gradients. The area covered by the single aggregates is larger in the case of acetonitrile than for ethanol solution (see Figure 8b,d, respectively). The larger size of agglomerations grown from acetonitrile solution is consistent with the fact that the solubility in ethanol is more than twice as high as for acetonitrile. As the two solutions are equimolar, precipitation of 2 will start earlier from acetonitrile and the aggregates have more time to grow and coagulate. This effect is enhanced by the lower evaporation rate of acetonitrile compared to ethanol. Notably, the solubility experiments proved that the resulting agglomerates of 2 dissolved in ethanol are smaller due to the better solubility. Next to the island formation, Ostwald ripening [61] was observed, especially for 2-SpA (cf. Figure 8e). This phenomenon occurs in a liquid environment and leads to bigger agglomerates separated by an empty zone from the smaller ones, which get dissolved during the growth process of the larger ones. Ostwald ripening is a thermodynamically driven process, which minimizes the island edge line, and the high surface to volume ratio is decreased by the formation of big agglomerates. By comparing the particle area of the two samples obtained from ethanol and acetonitrile, the difference between the two solvents used is quite obvious, and the calculated particle density for acetonitrile samples is much lower than that of ethanol ones (cf. Table 2). The highest possible coverage of (59.95 ± 3.00)% was achieved for the 2-SpE sample. Overall, full coverage of the substrate by 2 cannot be achieved with the chosen deposition parameters. The coverage extracted from all microscopy images shown in Figure 8 are presented in Table 2.
AFM measurements were performed to determine the height and the shape of the agglomerations. Micrometer-sized 3D aggregate shapes are obtained in all cases. The results are shown in Figure 9 and the associated maximum heights are summarized in Table 2. The floral structure grown during dip coating from ethanol solution (Figure 9c) shows flatter maximum height than the smaller islands formed (cf. Figure 9a,b,d). The origin of the island height difference observed for the two deposition methods is not fully understood. The fractal boundary shape again suggests further growth on the gold surface by diffusion-limited aggregation. In both cases indications for Ostwald ripening of larger particles at the expense of smaller ones are present. Depending on the solvent evaporation rate one or the other mechanism dominates, as discussed in more detail below. larger ones. Ostwald ripening is a thermodynamically driven process, which minimizes the island edge line, and the high surface to volume ratio is decreased by the formation of big agglomerates. By comparing the particle area of the two samples obtained from ethanol and acetonitrile, the difference between the two solvents used is quite obvious, and the calculated particle density for acetonitrile samples is much lower than that of ethanol ones (cf. Table 2). The highest possible coverage of (59.95 ± 3.00) % was achieved for the 2-SpE sample. Overall, full coverage of the substrate by 2 cannot be achieved with the chosen deposition parameters. The coverage extracted from all microscopy images shown in Figure 8 are presented in Table 2.
AFM measurements were performed to determine the height and the shape of the agglomerations. Micrometer-sized 3D aggregate shapes are obtained in all cases. The results are shown in Figure 9 and the associated maximum heights are summarized in Table  2. The floral structure grown during dip coating from ethanol solution (Figure 9c) shows flatter maximum height than the smaller islands formed (cf. Figure 9a,b,d). The origin of the island height difference observed for the two deposition methods is not fully understood. The fractal boundary shape again suggests further growth on the gold surface by diffusion-limited aggregation. In both cases indications for Ostwald ripening of larger particles at the expense of smaller ones are present. Depending on the solvent evaporation rate one or the other mechanism dominates, as discussed in more detail below.  For the spin-coated samples shown in Figure 9a,b islands with well-defined terraces are formed. For the ethanol spin-coated sample shown in Figure 9 a combination of large and small islands can be observed. Interestingly, in the vicinity of large particles, a large area is depleted from islands; this dark red area of the image probably corresponds to pure Au. A "void" depletion region surrounding the cluster, which is typical for Ostwald ripening, can also be seen in the microscopy image of samples spin-coated from acetonitrile in Figure 9b as a light white shadow of the agglomerates. In some cases, those shadows, corresponding to the gold surface, are elliptically shaped with the long axis of the ellipse pointing in the same direction. It is well possible that this characteristic shape is a result of the outward flow of the solution, fast evaporation time during the spinning process and the lower solubility of compound 2 in acetonitrile.
The dominant growth mode in the sample shown in Figure 9c is the diffusion-limited aggregation [61]. The floral-like agglomerations of 2 only occurs during the drop (see above, different scales) and dip coating procedure. Therefore, it can be assumed that this For the spin-coated samples shown in Figure 9a,b islands with well-defined terraces are formed. For the ethanol spin-coated sample shown in Figure 9 a combination of large and small islands can be observed. Interestingly, in the vicinity of large particles, a large area is depleted from islands; this dark red area of the image probably corresponds to pure Au. A "void" depletion region surrounding the cluster, which is typical for Ostwald ripening, can also be seen in the microscopy image of samples spin-coated from acetonitrile in Figure 9b as a light white shadow of the agglomerates. In some cases, those shadows, corresponding to the gold surface, are elliptically shaped with the long axis of the ellipse pointing in the same direction. It is well possible that this characteristic shape is a result of the outward flow of the solution, fast evaporation time during the spinning process and the lower solubility of compound 2 in acetonitrile.
The dominant growth mode in the sample shown in Figure 9c is the diffusion-limited aggregation [61]. The floral-like agglomerations of 2 only occurs during the drop (see above, different scales) and dip coating procedure. Therefore, it can be assumed that this aggregates shape is related to the available time for the formation of this floral pattern. If the solvent is not fully evaporated during the retraction of the substrate, the nanoclusters of 2 have sufficient time to agglomerate. Table 2. Particle densities of the films extracted from the optical microscopy images shown in Figure 8 and maximum height for agglomerations extracted from the AFM images shown in Figure 9.

Sample
Particle Density/% (Microscopy Images) Summarizing, the deposition of bismuth oxido cluster 2 from different solutions and using three different deposition methods can lead to the formation of agglomerations or islands of different shape and size. Considering the XRD results, these structures are crystalline (see Figures S8 and S9).

Conclusions
In this work, the high-yield conversion of the bismuth subnitrate [Bi 38 24 (dmso) 9 ] (2) is demonstrated. The nanocluster 2 shows a hydrodynamic diameter of (1.4-1.6) nm (in CH 3 CN) and (2.2-2.9) nm (in Ethanol) according to DLS. PXRD results are in line with a diameter of approximately 2 nm. Cluster 2 is soluble in a variety of organic solvents, which makes deposition methods from solution such as spin, dip and drop coating attractive. Using XRD, XPS and FTIR spectroscopy, it is demonstrated that the cluster core stays intact upon deposition on gold surfaces. Macroscopically, the films obtained by drop coating are neither homogeneous nor do they spread over a large area by using the same amount of solution compared to spin and dip coating. Optical microscopy as well as SEM images show that the growth behaviour in drop coating can be explained by the coffee ring effect. By the usage of more solution, cracked, very rough structures with a tendency to diffusion-limited aggregation were observed. By using spin and dip coating methods, bismuth oxido cluster agglomerations with a lateral size on the micrometre scale were obtained. The lateral dimensions of the crystalline agglomerates are nearly monodisperse and the distance between the neighbouring micro-agglomerates is nearly constant. Since their size can be controlled by the choice of solvent and of the deposition method, compound 2 could be suitable for the growth of high-density bismuth oxido micro or nanostructures on prepatterned surfaces, where the prepatterning provides nucleation centres. Finally, this study provides a basis for further studies on wafer scale deposition of micro and nanostructures made of neutral metal oxido nanoclusters on metal surfaces, that are promising heterostructures for electronic or photonic applications.  Table S1. FTIR vibration modes