1 Introduction

Carbon monoxide (CO) is an odorless, colorless gas that is lethal in small quantities. It has a permissible exposure limit (PEL) of 50 ppm over an 8-h work period, referred to as the time-weighted average (TWA). The PEL is the maximum amount of chemical substance or physical agent a worker may be exposed to under OSHA regulations. One such solution to diminish CO concentrations is to oxidize it to carbon dioxide (CO2). The PEL for CO2 is 5000 ppm TWA, resulting in rare cases of poisoning [1]. This method has practical applications in the industrial sector, such as car emissions, CO2 lasers for welding, medical surgery and spectroscopy, fuel technology, gas sensors, and various chemical processes [2,3,4,5,6,7].

It was discovered by Haruta et al. [8] that gold nanoparticles are effective for low-temperature CO oxidation [8]. Since this discovery, gold nanoparticles have become a big topic of discussion for catalytic reactions. Valden et al. [9] used a sputtering technique to study gold particles of various sizes with CO oxidation. They found that gold nanoparticles with dimensions of 3.5 by 1.0 nm2 (approximately 300 atoms per particle) and smaller exhibited a metal-to-nonmetal transition. This transition in gold nanoparticles has been linked to enhanced catalytic activity [10]. Further, gold particles, as well as some other metals like iron and copper [11], exhibit a rise in their redox potential as their size decreases [12] where a higher reduction potential will give rise to higher catalytic activity [13]. Given this trend, small gold nanoclusters could display higher catalytic activity than nanoparticles for some reactions. However, gold nanoparticles < 1 nm in diameter are not very stable [14], which would manifest as a negative effect on their activity [15].

As research into catalytic gold emerged, so did research into small uniform catalysts that could deliver consistent and reproducible results. This opened the door to nanocluster catalysis, where there are a specified number of atoms in each gold particle, whereas nanoparticles typically incorporate a range of sizes. Thiolate-protected gold nanoparticles, first proposed by Brust and Schiffrin [16], has become the building block to obtaining a well-defined Aun(SR)m formula using specific Au-to-thiol ratios. Here, the gold cluster with number of atoms, n, corresponds to a certain number of thiolate groups, m, encompassing a sulphur (S) atom and an organic rest group (R), which creates a monolayer surrounding the gold cluster [17, 18].

The Au144 nanocluster has been recognized as one of the larger, relatively stable, gold clusters [19,20,21,22,23]. Its structure has been established experimentally, utilizing techniques such as electrospray ionization (ESI), MALDI-MS, STEM, and X-ray spectroscopy [21, 24,25,26]. Weissker et al. performed time-dependent density-functional theory (TD-DFT) calculations to show individual peaks representing discrete levels of the structures localized electronic states [27]. This means there is a discrete energy band structure that develops for small clusters, hence there is a metal-to-nonmetal transition, similar to that of semiconductors [28]. That being stated, gold nanoparticles by themselves tend to sinter easily, losing their catalytic ability. The smaller the particles, the more this sintering becomes a prominent issue. This is, in large part, why very small gold nanoparticles and clusters are a difficult catalyst to effectively utilize. To stabilize gold nanoparticles, at the very least, they are loaded onto a support. The support is generally a metal oxide [29], but others have reported using other support types, such as carbon-based or titanium(IV) chloride supports [30, 31]. Even with different support structures, small gold nanoparticles (< 10 nm in diameter) sinter above room temperature. In the interest of overcoming this issue, core–shell structures have been investigated [32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47], where the gold nanoparticle/cluster is the core with a protective, often metal oxide, shell. This allows the gold to retain its size and shape, preserving its catalytic ability under thermal and mechanical stress. These core–shell structures have not yet been investigated for small gold clusters, given the cluster’s delicate nature. Given the large increase in surface-to-volume ratio for cluster-sized core–shell structures, they are a potential effective catalyst, exhibiting more catalytic sites compared to larger particles of the same volume. For this reason, research into their stabilization is key. This paper seeks to investigate the stability, catalytic properties, and effect of the thickness of an iron oxide shell surrounding Au144 clusters.

2 Experimental

2.1 Materials

The chemical reagents used were hydrogen tetrachloroaurate trihydrate (HAuCl4·3H2O), 1-hexanethiol (SHC6H13, 98%), tetraoctylammonium bromide (TOABr, 98%), sodium borohydride (NaBH4, 99%), iron pentacarbonyl (Fe(CO)5), oleylamine (reagent grade), oleic acid (90%), and Cab-O-Sil (fumed silica). All solvents used were purchased from Fischer Scientific, save deionized water, and were used as received without further purification unless specified otherwise; the list of solvents is as follows: methanol (certified ACS, ≥ 99.8%), toluene (ACS Grade, ≥ 99.5%, LabChem™), acetone (Certified ACS, ≥ 99.5%), ethanol (200 proof (100%), USP, Decon™ Labs), dichloromethane (99.5% min., ACS, EMD Millipore), hexane (certified ACS, ≥ 98.5%), and diphenyl ether (99%, ACROS Organics™).

2.2 Catalyst Preparations

In preparing the Au144 clusters, Qian’s methods were utilized [19]. In ambient conditions, 0.70 mmol of HAuCl4·3H2O was added with 0.70 mmol TOABr in methanol in a round bottom flask. The solution changed color from yellow to red, indicating the formation of Oct4N+ AuBr4−. Then, 1.505 mmol of 1-hexanethiol were added to the solution at room temperature and the color of the reaction mixture rapidly turned white. The thiol/Au ratio was adjusted to 4.3:1 based on another groups’ modification [27]. After 1 h, a fresh solution of NaBH4 (3.5 mmol) in cold DI water was rapidly added to the solution under vigorous stirring. The color of solution immediately turned black and produced Au clusters, which were then precipitated out of the methanol solution. The reaction continued to stir for 5 h to completely equilibrate the reaction. The black precipitates were then collected by centrifugation (5 min at 5000 rpm) and decantation. The black precipitates were washed with excess methanol and collected by centrifugation several times to completely remove small-molecule and electrolyte residue, including excess free thiol residues. Then, toluene was used to separate the Au clusters from Au(I)–SR polymers, which are poorly soluble in most solvents. The as-obtained solution only contains Au144(SR)60 (major product) and Au25(SR)18 (minor product), determined via MALDI and ESI [19, 27, 48]. Acetone was used to separate the Au144(SR)60 and Au25(SR)18 clusters, where the Au144(SR)60 clusters separate out during centrifuge and Au25(SR)18 clusters remain dispersed in the acetone solution. The Au144(SR)60 clusters can be dispersed into hexane, toluene, or dichloromethane depending on characterization. To determine yield, an empty vial was weighed, a Au144 solution was added, and then the solution was dried to determine the weight of clusters. Afterwards, the clusters were dispersed into a determined amount of solution, creating a known concentration. In this way, 20 mg of the clusters can be extracted into another vial for use in the second step to this experiment. To ensure an exact amount was used, this solution was re-weighed out in the same fashion.

In coating the cluster with an iron oxide shell, methods from Yin et al. were applied [33]. In a 4-neck 100 mL round-bottom flask, 50 mL of diphenyl ether, 0.15 mL of an oleylamine, and 0.15 mL oleic acid were added and stirred for 5 min. Then, 20 mg of Au144 clusters dispersed in 2 mL of toluene was added to the mixture. The solution was set to stir under nitrogen flow and heated at 80 °C for 30 min. The reaction was put under a nitrogen blanket before adding the iron pentacarbonyl in varying amounts. The flask was slowly heated to 200 °C and left to stir for 30 min. The reaction was then left to cool to approximately 100 °C and then the N2 environment was removed and air was flowed through the flask for a minimum of 10 min to allow the iron to oxidize. Subsequently, the reaction flask was allowed to cool to room temperature. The solution was washed with ethanol and centrifuged and the core–shell precipitate was then dispersed in hexane.

To analyze and test the core–shell structures, they were deposited onto fumed silica (SiO2) using ~ 1.5% gold loading. Fumed silica support was utilized because of its inert properties, given its weak metal support interaction (WMSI) and ability to adsorb cations, as opposed to other metal oxide supports that have strong metal support interactions (SMSI’s). The silica (Cab-O-Sil) was added to the colloidal mixture of Au144@Fe2O3 core-shell particles in hexane, sonicated, and stirred for 2 h before slowly evaporating off the solvent, using rotary evaporation. After removing the solvent, the powder was dried and calcined at 300 °C for 2 h.

2.3 Characterizations

The Au144 clusters were characterized by utilizing a Bruker matrix assisted laser desorption ionization time-of-flight mass spectrometer (MALDI-TOF-MS) and X-ray diffraction (XRD). The mass spectrum was collected in a linear positive ion mode. A dichloromethane solution of the Au144(SR)60 particles (5 mg/mL) were mixed with the matrix, a chloroform solution of trans-2-[3-(4-tert-butylphenyl)-2-methyl-2-propenylidene]malononitrile (DCTB), applied to the sample plate, and air dried. XRD data of the clusters was gathered by studying thin films of the particles. At room temperature, the clusters were analyzed using a PANalytical Empyrean diffractometer over the range from 10° to 80° 2θ using Cu Kα at wavelength 1.540958 Å radiation at 45 kV and 40 mA.

The Au144@Fe2O3 core–shell structures were characterized by XRD, for which the data of the core–shell structures were collected by studying powder samples of the silica loaded systems and analyzed using the same instrumentation and parameters as those used for the Au144 thin films. The nitrogen physisorption isotherms were measured at 77 K using a Gemini 2375 Brunauer–Emmett–Teller (BET) surface area analyzer. Elemental analysis was done by inductively coupled plasma optical emission spectroscopy (ICP-OES), using an Optima 2100 DV optical emission spectrometer (PerkinElmer Corporation). Visual measurements were operated on Hitachi HD-2000 scanning transmission electron microscopy (STEM) and with high-angle annular dark-field (HAADF) techniques with the aid of ImageJ, the samples were dispersed in ethanol or hexanes assisted by ultrasonic technique. Oxidation states of the metals were determined via X-ray photoelectron spectroscopy (XPS). All XPS data was attained using a Phi 3056 spectrometer equipped with an aluminum anode source operated at 15 kV, an applied power of 350 W, and a pass energy of 93.5 eV.

The catalytic activity of the core–shell structures were tested by analyzing its ability to oxidize CO to CO2. 20 mg of the Au144@Fe2O3 catalyst was packed into a quartz tube (inner diameter = 4 mm) sealed on either side by quartz wool, a gas stream of 1% CO (balance air) flowed through the catalyst at a rate of 10 mL/min, and the exiting stream was analyzed by a gas chromatograph equipped with a dual molecular sieve/porous polymer column, a thermal conductivity detector, and a gas hourly space velocity (GHSV) of 30,000 mL (h gcat)−1. The reaction temperature was controlled using a voltage transformer attached to a furnace. In addition, Fourier transform infrared (FTIR) spectroscopy was able to determine CO adsorption on the structure. With the core–shell structures loaded onto fumed silica and calcined at a certain temperature, FTIR spectroscopy was obtained at a temperature of 5 °C. A spectrum was taken after 10 min of 2% CO flowing over the loaded structure, with any spectral contribution from gaseous CO subtracted. Then, spectra were taken after purging the sample with 2% He for 3 and 10 min. Similar research done by Yin et al. was completed using gold nanoparticles 2–5 nm in diameter with varied iron oxide shell thicknesses tested for catalytic activity using CO oxidation. While many of these authors share authorship with Yin et al. [33], access to the previous instrument was not possible, so a one-to-one comparison was not possible. Regardless, the different diameters of both the Au core and Fe2O3 shell as well as the loading of each to the substrate (silica) will undoubtedly yield different results. Also, of note is that with bigger particles Yin et al. were able to get higher Au loading (~ 3.5%), while we were unable to get above 2% Au loading with smaller particles.

3 Results and Discussion

3.1 Au144(SC6H13)60 Nanocluster

The MALDI-TOF mass spectrum of the Au144 clusters with 1-hexanethiol capping is expected to show a broad peak at approximately 32,500 Da [19, 27], as shown in Fig. 1, even though the total structure itself is calculated to be 35,397 Da. The peak position shift of roughly 2900 Da is due to partial loss and fragmentation of the thiol–gold (–RS–Au–SR–) staple motifs from the laser intensity under the MALDI conditions [19], where each staple motif contributes ~ 430 Da to the total mass of the structure. Each Au144(SR)60 cluster has been previously calculated by other groups to have 30 staple motifs, which include 30 of the 144 Au atoms in the capping structure [27]. For further characterization of the clusters, thin film XRD measurements were taken (Fig. 2). The small broad peak, with FWHM being 5.16° 2θ and centered at 38.58° 2θ, indicates the Au cluster has a domain of approximately 1.7 nm in diameter, via the Scherrer equation [49]. This correlates well with DFT estimations of ~ 1.6 nm, calculated by Weissker et al. [27].

Fig. 1
figure 1

MALDI-TOF-MS of Au144(SC6H13)60 with peak at 32,500 m/z (Da)

Full size image
Fig. 2
figure 2

XRD of Au144(SC6H13)60 thin film with broad peak at 38.58° 2θ indicative of small Au domain

Full size image

3.2 Au144@Fe2O3: Effect of Shell Thickness

The thickness of the iron oxide shell was controlled by varying the amount of iron precursor added during the reaction. The different Au144@Fe2O3 core–shell structures are referred to by their Au:Fe ratios by mass. The structures studied in this paper were 1:2, 1:4, and 1:6. HAADF–STEM images of the 1:4 structures, prior to being loaded and before calcination, show a small Au144 particle surrounded by an iron oxide shell (Fig. 3). There is a small gap at the core–shell interface, which is believed to be indicative of the thiol capping still attached to the gold cluster.

Fig. 3
figure 3

a, b HAADF of 1:4 Au144@Fe2O3 structure as synthesized, as well as the c Fe (blue) and d O (red) areal density of structure (b)

Full size image

Once calcined to 300 °C, the thiol capping and other organic residuals, which includes any solvent surfactants that may still be in solution, are removed with the purpose of activating the gold catalysts. The structures that were fully covered reveal a yolk–shell like structure, as has been seen in other literature [41, 50,51,52,53,54], rather than strictly a core–shell design (Fig. 4b, c). This is due to the removal of thiol capping agent on the gold cluster, which winds up leaving a gap, or void, between the core and shell.

Fig. 4
figure 4

STEM of Au144@Fe2O3/SiO2 with varying Fe2O3 content for the a 1:2, b 1:4, and c 1:6 structures

Full size image

In our studies, a threshold limit was discovered for an optimal metal oxide shell of this particular morphology. When too little of the iron precursor (Fe(CO)5) was added, the metal oxide was unable to cover the cluster completely, which led to sintering under calcination, as was the case with the Au:Fe 1:2 structure, and revealed a dumbbell-like structure (Fig. 4a), as has been seen in other literature [37, 55, 56]. The threshold amount may be different with a different thiol capping. For the 1:4 structure, the Fe2O3 shell was approximately 1.5 nm thick, while the 1:6 structure was approximately 2.5 nm thick, estimated via STEM. According to BET results, the surface area is larger for the 1:4 versus the 1:6 structure (Table 1).

Table 1 BET surface area and ICP-OES weight percent of core–shell structure loaded onto silica
Full size table

Bulk gold has a face-centered-cubic (fcc) crystal structure, and when analyzed with XRD, there will be a predominant peak at 38.2° 2θ and less predominant peaks at 44.4°, 64.5°, and 77.5° 2θ representing the (111), (200), (220), and (311) surface planes, respectively (Joint Committee on Powder Diffraction Standards-International Center for Diffraction Data (JCPDS-ICDD) file #04-0784). However, very small particles (< 2 nm) can be quite difficult to detect and XRD is no exception [57]. Some other factors to consider in these measurements is that the iron oxide shell is amorphous in nature, as well as the weight fraction of gold being very small. With the Au144 clusters, there is barely a hump at the prominent angle at 38° 2θ, with the exception of the 1:2 structure (Fig. 5). After calcination, it is clear that the 1:2 structures sintered as that peak becomes sharper and more defined, representative of larger gold particles. This corroborates with the STEM images, that the iron oxide shell was unable to cover the particle completely, and, without the protective shell, the Au144 clusters coalesced. For the 1:4 and 1:6 structures, no sintering was visible from the XRD spectra, which is indicative of full coverage of the amorphous iron oxide shell, as the STEM images indicate.

Fig. 5
figure 5

XRD of Au144@Fe2O3/SiO2 with varying Fe2O3 content for a 1:2, b 1:4, and c 1:6 structures, with a line at 38.2° 2θ where bulk gold peaks are prominent

Full size image

XPS of the 1:4 structure, which was loaded onto fumed silica and calcined at 300 °C, provided a survey spectrum, or wide-scan showing all elements present (Fig. 6a), with subsequent high resolution (core level) spectra of each element identified (Fig. 6b–f), as well as the overall percentage surface composition that each element contributes (Table 2). The binding energy position of the main Au 4f peak is at 83.8 eV, indicative of a metallic Au species [58]. This is merely a slight shift of − 0.2 eV in the binding energy for the Au 4f7/2 peak from 84.0 eV, which is typically seen for bulk gold [59,60,61,62]. The binding energy position of the Fe 2p3/2 peak is at 711.8 eV, indicative of a primary 3+ valence [63], which means the iron oxide shell is in the Fe2O3 structure, with a + 0.4 eV shift from that typically measured at 711.4 eV [64,65,66,67,68]. The O 1s orbital is primarily associated with the Si 2p orbital from the SiO2 support. The C 1s signal was relatively low and is predominantly adsorbed carbonaceous material during analysis.

Fig. 6
figure 6

XPS of surface composition of Au144@Fe2O3 1:4 structure showing a the survey spectrum and the high-resolution spectra of b gold, c iron, d oxygen, e silicon, and f carbon

Full size image
Table 2 XPS—surface composition (at.%): Au144@Fe2O3
Full size table

3.3 Synergistic Effects of Core–Shell System

Using control experiments of Au144 nanoclusters and Fe2O3 nanoparticles loaded onto silica, (Au144/SiO2 and Fe2O3/SiO2, respectively) we were able to compare with the Au144@Fe2O3/SiO2 core–shell particles. Using CO oxidation tests, the synergistic effects of the core–shell structures are apparent by comparison of each structures catalytic abilities (Fig. 7). For the core–shell structures, full oxidation of CO for the 1:2, 1:4 and 1:6 structures occurred at 211, 75, and 111 °C, respectively. Au144 nanoclusters by themselves showed some oxidation at room temperature, but did not exhibit full oxidation of CO until ~ 500 °C. Whereas, Fe2O3 nanoparticles by themselves never reached full CO oxidation, up to 500 °C. The Au144 nanoclusters were determined to have sintered into larger particles, and more evidence that sintering occurred was that the catalyzed Au144 particles visibly appeared a red/purple color after testing was completed.

Fig. 7
figure 7

CO conversion on Au144@Fe2O3/SiO2 catalysts with varying Fe2O3 shell thickness, as well as Fe2O3/SiO2 and Au144/SiO2

Full size image

To further investigate the behavior of CO adsorption onto the Au144@Fe2O3 particle and obtain more information about the Fe2O3 surface, FTIR spectroscopy of the 1:4 core–shell structure, loaded onto fumed silica and calcined at 300 °C, was obtained at 5 °C (Fig. 8). The top spectrum was taken after 10 min of 2% CO flow. The middle and bottom spectra were taken after purging the sample with 2% He for 3 and 10 min, respectively. There was only one distinct peak at 2139 cm−1, from the top spectrum, which can be attributed to CO being adsorbed onto a positively charged Au species (Auδ+, 0 < δ < 1) [69, 70]. The middle and bottom spectra showed no distinct peaks, which signifies that the CO was readily desorbed from the catalyst surface. The positively charged Au species is attributed to the iron oxide layer that is surrounding the Au core. Only during the process of CO oxidation, while CO is adsorbed, is this positive Au species observed.

Fig. 8
figure 8

FTIR of CO adsorption onto 1:4 structure Au144@Fe2O3/SiO2 at 5 °C (gaseous CO contribution subtracted)

Full size image

3.4 Thermal Treatments of 1:4 Au144@Fe2O3 Structure

The 1:4 structure, with the best CO oxidation results, was put through varying thermal treatments to test the structure’s optimal activity. The structure was tested at pretreatments of 300, 500, and 700 °C. Results for the 300 °C pretreatments showed the best activity with full conversion at 75 °C (Fig. 9a). It should be noted, however, that the 1:4 structure showed full conversion at 95 °C for both pretreatments at 500 and 700 °C. Although the higher temperature pretreatments do not display the best catalytic activity, it is still active below 100 °C, which is more efficient than the 1:6 structure pretreated at 300 °C. This indicates thermal stability at these higher calcination temperatures. However, the XRD for these 1:4 samples (Fig. 9b) show some sintering as the pretreatment temperature is increased, where the peaks become more defined, particularly at 38.2° 2θ, where bulk gold is typically represented.

Fig. 9
figure 9

The 1:4 Au144@Fe2O3/SiO2 structure calcined at temperatures 300, 500 and 700 °C showing a CO conversion and b XRD spectra

Full size image

4 Conclusions

Au144@Fe2O3 core–shell structures were synthesized with varying shell thickness by varying the metal oxide precursor used. These structures were loaded onto a fumed silica support via colloidal deposition, calcined at 300 °C, and the structure’s catalytic activities were compared based on the efficiency of CO oxidation. Between the Au144@Fe2O3/SiO2 structures 1:2, 1:4, and 1:6 structures, the 1:4 structure exhibited 100% oxidation of CO at the lowest temperature, 75 °C, and the 1:6 structure, with a thicker iron oxide shell, reached full oxidation of CO at a slightly higher temperature, 111 °C. The 1:2 structure did not result in full core coverage, which is why it exhibited poor CO oxidation properties compared to the 1:4 and 1:6 structures and represents a minimum threshold of Fe2O3 shell thickness. Based on these results, the 1:4 structure was studied under thermal treatments of 300, 500, and 700 °C, followed by CO oxidation. The results showed that the best results for the 1:4 structure was at 300 °C pretreatment, however, higher temperature pretreatments still showed promising results.