1 Introduction

Amongst the most important contributions made by Professor Michael Spencer to the field of catalysis are several focused on the nature of the metal nanoparticles that form the active component of many heterogeneous catalysts. In 1986 for example, Spencer discussed surface atom mobility on iron, copper and platinum nanoparticles [1] proposing that this mobility allows the nanoparticles to reach their own stable states and this might account for the surprising reproducibility of catalytic activity in many systems despite varying preparation conditions. This work anticipated, in some respects, the scanning probe experiments a decade later that showed how important atom mobility can be in surface reaction mechanisms [2, 3]. In another example, Spencer considered [4] the SMSI effect in the Pt/TiO2 system from a thermodynamic perspective and discussed the possible roles of water and potassium as surface species modifying the behaviour of the nanoparticle/substrate interface. Whereas Spencer was primarily concerned with the nature of the active particle during catalysis, our focus has recently been on the initial formation of such nanoparticles at the surface of carbon catalyst supports and in particular, the role of the functional groups on the support in their nucleation and stabilisation [5,6,7]. The importance of surface species in these systems is widely recognised [8,9,10,11,12], however, the complex nature of the carbons typically used as supports for catalysts has hindered attempts to explore the role of specific species. We have addressed this issue by selecting highly ordered pyrolytic graphite (HOPG), as an idealised carbon substrate to functionalise. Whilst the sp2 nature of a graphite surface does not reflect the majority sp3 species present in most carbons used as catalyst supports, graphite’s uniform surface chemistry allows us to investigate the role of individual functional groups in isolation [5, 6, 13]. We have adopted an experimental protocol that mirrors that used in many catalyst preparation routines and have begun to be able to comment directly on the role of specific functional groups. Acid treatment of graphite surfaces generates OH or C=O groups, the former largely transforming into the latter on heating. DFT calculations [5] show that both types of oxygen interact with gold and suggest that nucleation should occur in both cases. This was confirmed experimentally [7], and gold deposition from an aqueous solution at ~ pH 5–6 was shown to require the oxygen functional groups for nucleation to occur at all on HOPG. OH(a) groups reduced Au3+ ions to a metallic state whereas the C=O groups result in gold deposition in a Au3+ state.

In the present paper, we expand our previous investigations to look at different preparation methods for the surface hydroxides, different functional groups, and a different cation. Specifically we examine graphite surfaces treated with hydroxides produced from a basic solution instead of an acidic one, sulfur containing functional groups (as used in the recently commercialised gold/carbon catalyst for vinyl chloride synthesis [14]) and the effect of surface hydroxyl groups on palladium nucleation.

2 Experimental

HOPG samples (~ 10 mm square, ZYH quality, supplied by Scanwel Ltd, UK) were cleaned before each experiment by exfoliation of the top few graphene monolayers using adhesive tape. Treatment involved the placement of a 100 μl droplet of fresh diluted (see below) acid or base on the HOPG surface, allowing the system to stand for 30 min before drying under a pure helium or nitrogen stream. Aqueous solutions wet the HOPG surfaces effectively. For metal deposition, 100 µl droplet of precursor solution (HAuCl4·3H2O, Sigma–Aldrich, 99.999% or acidified palladium (II) chloride, Sigm–Aldrich ReagentPlus®, 99%), was placed on the pre-treated sample and left for 1 min. The sample was subsequently dried under a stream of helium or nitrogen for 2 min. The pH of the 1 × 10−5 M and 2 × 10−6 M gold solutions were measured to be 5.26 and 5.63 before contact with the HOPG surface. To treat the HOPG surfaces with ammonium thiosulfate ((NH4)2(S2O3)) a 100 μl droplet of 0.2 M, (NH4)2(S2O3) was added to the surface of clean HOPG and left to stand in air for 30 min. The sample was then dried under a stream of He for 15 s. Where indicated, selected samples were heated for 1 h after surface treatment and allowed to cool to ambient temperature before analysis.

Ultra-pure water (“UPW”, Millipore, resistivity ca 19 MΩ cm) was used for dilution of the acids and the gold solution before surface treatment and for control experiments that replicated dosing and analysis conditions. We discovered that these experiments are extremely prone to contamination of the surface by silicon oxides which may have originated from the tape used for strip cleaning the HOPG. Despite precautions, silicon was a frequent contaminant seen in the XPS of the samples and resulting in a dominant SiOx peak in the O(1s) region of the spectrum. In these cases, the data was abandoned, and the samples cleaned again.

X-ray photoelectron spectroscopy (XPS) was performed on a Thermo Fisher Scientific K-alpha+ spectrometer using a micro-focused monochromatic Al x-ray source (72 W) over an area of approximately 400 μM. Pass energies of 150 eV and 40 eV were used for survey and high resolution scans respectively with 1 eV and 0.1 eV step sizes. Samples were mounted on AFM stubs for ease of handling which necessitated the use of charge neutralisation with a combination of low energy electrons and argon ions.

The XPS data was analysed using CasaXPS v2.3.23 using Scofield sensitivity factors and an energy compensation factors to account of the electron inelastic mean free path, which is approximated using KEn, where n = −0.6. All binding energies were referenced to the C(1s) peak taken to be 284.5 eV. Curve fits were made with spin orbit splitting components for the Au(4f), Pd(3d), S(2p) and Cl(2p) spectra fixed at 3.7, 5.3, 1.16 and 1.6 eV, and relative areas of 4:3, 3:2, 1.96:1 and 2:1, respectively. Au(4f) and Pd(3d) spectra were fitted using CasaXPS’ Finite Lorentzian (“LF”) lineshape, which is a Voigt type function with a dampening factor to drive the Lorentzian tail down to the baseline to minimise over estimation of peak areas. Fitting of the Au and Pd regions were made using models derived from the respective bulk metal foils, the FWHM of the species were therefore determined by the 4f7/2 and 3d5/2 peaks of the foils, but allowed to vary by ca. 0.3 eV. The extent of metal reduction under x-ray irradiation was assessed by recording the Au(4f) region at both the start and completion of a set of XP spectra. Typically, no increase in the proportion of Au0 was evident after 30–40 min of irradiation. It is worth noting here, that the experimental protocol of treating samples externally to the spectrometer makes quantitative estimates between samples difficult and thus in the present paper, quantification is only used to compare concentrations of surface components on individual samples at the same treatment stage. However, the observed binding energies of the components within the spectra were reproducible from sample to sample and general trends can be deduced.

AFM images were acquired on a Bruker Veeco Multimode system in tapping mode using silicon nitride tips. Multiple images were recorded of several different sites on the surface and on several different samples and images were analysed using WSxM software [15].

3 Results and Discussion

3.1 Nucleation and Deposition of Palladium from Palladium Chloride

Palladium concentrations of the same order as previously used for gold deposition (1 × 10−6 M) resulted in no adsorption at the HOPG surface but at 1 × 10−5 M palladium was observed in the XP spectra (Fig. 1). On the surface pre-treated with ultra-pure water the expected Pd (3d) doublet is present with the 3d5/2 peak at 338 eV approximately 0.2 eV higher than is typical of PdCl2 (Figure S1 and References [16,17,18]). There is also a single chlorine state present with the Cl(2p3/2) peak at 198.4 eV corresponding to a metal coordinated chloride and this is the first clue that palladium chloride behaves differently to gold. Deposition of gold on HOPG from chloroauric acid solutions usually gives rise to two chlorine states with 2p3/2 peaks at 198 eV and 200 eV [19] assigned to chlorides bound to the metal and carbon respectively (see Fig. 3 for an example). In the present case, there is no evidence of chlorine associated with the carbon support. The Cl:Pd ratio of the adsorbed species is lower than expected for PdCl2, averaging ~ 1.3:1. This discrepancy is not accounted for by the very small degree of reduction of palladium which is evident when comparing spectra before and after exposure to the x-rays (not shown) but a contributing factor could be the replacement of chloride ions by hydroxides in solution (note Pd(OH)2 is not expected to occur in solution [20] below a pH of ~ 10 [21]). Figure. 1 shows that pre-treating the HOPG surface with higher concentrations of acid (and thereby generating higher concentrations of OH(a)) does not lead to any chemical reduction of the deposited palladium chloride species, as was the case with gold, but it does have the effect of decreasing the extent of deposition of palladium species (Fig. 1e). The O(1s) peak at 532.5 eV can be attributed to OH (a) which is unaffected by the adsorption of the palladium salt, the weak peak at ~ 535.5 eV due to Pd (3p3/2), confirming the small extent of adsorption. In the solutions we are using, the aqueous palladium–chloride species is likely to be negatively charged [20] and therefore increasing the number of adsorbed OH groups at the surface may increase the negative charge on the surface and thus repel the ion [22].

Fig. 1
figure 1

a–c, XP spectra of HOPG surface pre-treated with a range of concentrations of HNO3 and, after drying, exposed to 1 × 10−5 M aqueous solutions of PdCl2; (a) Pd adsorption at a surface pre-treated with pure water; (b) pre-treated with 1.0 M HNO3; (c) 2.0 M HNO3; (d) 5.0 M HNO3; d Quantification of the palladium and chlorine surface concentrations

Full size image

AFM images of the HOPG surfaces after exposure to palladium chloride solutions also reflect very different behaviour between the palladium and gold chlorides: PdCl2 deposited on the ultra pure water treated surface results in a number of regular features in the AFM images, approximately 3 nm in height with diameter of ~ 30 nm (Fig. 2a, b, c). These features are not present on clean graphite and we tentatively assign them to clusters of adsorbed PdCl2. There are far fewer features present after palladium chloride is dosed onto nitric acid treated HOPG samples. A small number of ~ 3 nm tall features are evident near step edges but the majority of the surface looks very flat, Fig. 2 AFM images of treated HOPG surfaces; (a) treated with 0.25 M HNO3; (b) 1 × 10−5 M aqueous solution of PdCl2 after treatment withultra pure water; (c) 1 × 10−5 M aqueous solution of PdCl2 after treatment 2 M HNO3;. (d) Line profiles from the AFM images (c). As Fig. 2a, shows and from previous work [5, 6, 13], nitric acid oxidises the HOPG surface leading to areas of local delamination, typically 100–150 nm wide and 3–10 nm high. In this case, and similar to the effect of heating [5, 6], the exposure of the surface to palladium chloride solution appears to remove some of these features.

Fig. 2
figure 2

AFM images of treated HOPG surfaces; a treated with 0.25 M HNO3; b 1 × 10−5 M aqueous solution of PdCl2 after treatment withultra pure water; c 1 × 10−5 M aqueous solution of PdCl2 after treatment 2 M HNO3. d Line profiles from the AFM images

Full size image

3.2 Functionalisation of a HOPG Surface with Aqueous Hydroxide Ions

We have shown that the oxidation of HOPG surfaces by low concentrations of mineral acids leads to functionalisation with OH(a) groups [13]. Since the energy required to break one of the C–C bonds is considerable, we have proposed that this reaction is occurring at existing defects on the surface [5, 6]. The question we wished to examine was whether OH species could be formed directly at the surface from a basic solution. Initial experiments with sodium hydroxide showed oxygen functionalisation of the surface in the XP spectra and the generation of typical 3–4 nm high OH related features in the AFM images but the XPS also showed sodium deposition which creates a possible complicating factor for the deposition of the gold or other cations. Instead, we have explored functionalisation of the surface with ammonium hydroxide solutions and as Fig. 3 shows, a functionalisation of the surface similar to that observed with the weak acids does occur. Furthermore, the OH groups generated at the surface also induce nucleation of gold from a chloro auric acid solution. The initial deposition is of a reduced gold state but at higher concentrations of the gold solution, Au3+ dominates together with the deposition of an associated chloride counter ion. It’s worth noting that whilst achieving the same effect as treating the HOPG with acids, a concentration of at least 5 M ammonium hydroxide was required to generate significant functionalisation of the carbon compared to only 0.2 M of the acids investigated.

Fig. 3
figure 3

NH4OH treatment of HOPG surfaces results in the oxidation of the surface to give primarily hydroxide species. These enable the nucleation of gold at the surface from HAuCl4 solutions. a–c, XP spectra of the O(1s), Au(4f) and Cl(2p) regions: a the presence of OH(a) after treatment with 5 M NH4OH. b After exposure of (a) to 5.0 × 10−6 M HAuCl4; c After exposure of (a) to 1.0 × 10−5 M HAuCl4; At low concentrations of the gold solution, the initial adsorption is gold metal but higher concentration solutions result in extensive Au3+ deposition with the associated chloride counter ion. The AFM images in d, show similar features on the surface after NH4OH treatment as we observed with weak acids

Full size image

3.3 Functionalisation of HOPG Surface with Aqueous Ammonium Thiosulfate Solutions

Johnston et al. have demonstrated [14, 23] that soft donors such as thiosulfates are an effective method of stabilising gold nanoparticles in a cationic form on carbon supports, but how the thiosulfates interact with the carbon is not known and to start to unpick this phenomenon we have studied the adsorption of thiosulfate onto HOPG substrates and their subsequent effect on gold adsorption from solution. To our knowledge, there has not been a study of the adsorption of ammonium thiosulfate on either clean or functionalised HOPG surfaces.

Reference spectra of two different thiosulfates are shown in the supplementary data Figure S2, and show good agreement to data in the literature [24, 25]. Gold sodium thiosulfate solid gives a spectrum with two components in the S(2p) region, characterised by S(2p3/2) peaks at 163.6 eV and ~ 169.1 eV, the latter attributed unambiguously to the SO3 group. The two states have a concentration ratio of 1:1.2 possibly reflecting the surface termination of the crystals. The Au+ 4f peak is at 84.6 eV, and a bit less than half the concentration of the 163.6 eV sulfur peak consistent with the molecular stoichiometry. A reference sample of ammonium thiosulfate was obtained by allowing a 0.2 M solution to dry on a HOPG surface giving a precipitate that was visible to the eye. For this sample, the S(2p) spectra show a third component in addition to those that can be assigned to the two sulfur states in the thiosulfate molecule. The third state has a S(2p3/2) component at 164.2 eV and can be assigned to atomic sulfur at the HOPG surface, implying the dissociation of approximately half the thiosulfate. Interestingly, the SO3 peak at 168.4 eV, has approximately the same surface concentration as the other two peaks combined, indicating that it consists of a combination of SO3(a) and S–SO3(a) and indicating that dissociation has little effect on the binding energy of this already very positively charged sulfur. Furthermore, the stoichiometry suggests that all the sulfur stays on the surface on dissociation of the ammonium thiosulfate. The ammonium ion is also present with a peak in the N(1 s) region at 402.1 eV and a surface coverage similar to that of the sulfur peak at 182.4 eV, a little more than half of that of the SO3 group.

At a clean HOPG surface, XP spectra show no adsorption of any sulfur containing species after treatment with 0.2 M ammonium thiosulfate (Fig. 4a), although there is a small increase in the oxygen species on the surface; the latter is to be expected after exposure to an aqueous solution. Adsorption does occur if the surface is first functionalised with hydroxide groups by acid pre-treatment (Fig. 4b). At least two sulfur states are now evident on the surface, characterised by S(2p3/2) components at ~ 164 eV and ~ 169.0 eV. From the reference spectra in Figure S2, the higher binding energy component can be unambiguously attributed to SO3(a) with the associated oxygen contributing to the O(1s) peak at 532.5 eV. The S:O ratio for the S 2p and O 1 s signals assigned to SO3(a) is higher than expected for − SO3, at 1:4.4, but OH(a) would also be expected to be formed in the experiment and appears near 532.7 eV [5, 6] possibly accounting for the extra peak area. The lower binding energy peak in the S(2p) region can be attributed to adsorbed sulfur showing almost complete dissociation of the thiosulfate, although there is some evidence for more than one sulfur state being present in the lower binding energy peak envelope. The adsorbed components are stable on heating to 373 K but are beginning to desorb by 473 K.

Fig. 4
figure 4

XP spectra of ammonium thiosulfate adsorption (0.2 M) on clean and functionalised surfaces and the effect on gold adsorption. In the S(2p) region, the 2p3/2 and 2p1/2 fitted components are combined to show the three separate states of sulfur more clearly. (a) (NH4)2S2O3 on clean HOPG; (b) (NH4)2S2O3 on HCl treated HOPG; c after heating (b) to 373 K; (d) after heating (c) to 473 K; (e) Exposure of (b) to 10−6 M HAuCl4 solution. (f) Surface in (b) treated with 0.5 M HNO3

Full size image

We conclude that the adsorption of ammonium thiosulfate onto hydroxyl covered graphitic surfaces from aqueous solution leads primarily to dissociative adsorption forming SO3(a) and S(a) with approximately 20% of the adsorbate remaining in the molecular form.

Exposure of the acid/thiol treated surface to an aqueous solution of chloroauric acid results in the selective loss of the SO3(a) species but no change in the lower binding energy state (Fig. 4). There is an associated deposition of gold at the surface similar to that expected on an acid treated HOPG surface with a final gold surface concentration approximately ¼ that of the sulfur concentration. The Au(4f) binding energy of 84.6 eV is ~ 0.6 eV above that expected for the metallic state and was not affected by exposure to the x-ray beam for > 30 min. There is no adsorption of any chloride counter ion or any change in the oxygen spectra to suggest the presence of gold hydroxides. Interestingly, the binding energy of 84.6 eV is identical to that expected for a Au+ state but the stability of the peak may suggest the presence of very small nanoparticles of gold rather than Au+. Freund et al. have shown that a complete attribution of the charge state of gold is best obtained from a modified Auger parameter [26, 27]. We have not been able to access the relevant gold Auger peaks with the current experiments but will explore it in future work. Treatment with 0.5 M HNO3 solution had a similar effect to the aqueous gold solution as shown in (Fig. 4f), where the 168 eV doublet is mostly removed after treatment with the acid. Interestingly, the resolution of the 164 eV doublet is also improved by the acid treatment suggesting a reduction in the range of environments of the S(a) perhaps indicating that there were some residual interactions between the adsorbed sulfur and the sulfite even though the S–S bond was broken.

We cannot definitively rule out the adsorbed thiosulfate having any effect on gold deposition from solution, but there is no direct evidence (other than the unusual Au(4f) binding energy) that this is the case; the behaviour of the gold is very similar to the case of adsorption on a surface functionalised with a weak acid.

The AFM data is less clear cut than the XPS, (Fig. 5). Treatment of the clean HOPG surfaces with thiosulfate solutions leads to no significant changes, consistent with the XP data showing no adsorption. Furthermore, thiosulfate adsorption after a pre-treatment with HCl, shows similar delamination effects to those observed when the surface is only treated with the weak acids. Finally, AFM images of the thiosulfate treated surface exposed to gold solution do not show any features that can be attributed to the adsorption of gold. This would correlate with the interpretation of the 84.6 eV peak in the XP data as resulting from nanoparticles of gold that are too small to be imaged at this resolution but cannot be considered definitive. Further conclusions require imaging at much higher resolution.

Fig. 5
figure 5

AFM images of acid treated HOPG surfaces treated with ammonium thiosulfate and gold solutions. a HOPG surface after treatment with 0.2 M HCl followed by direct exposure to 0.2 M (NH4)2(S2O3). b Exposure of (a) to 10−6 M HAuCl4 solution. c Line profiles from (a) and (b)

Full size image

4 Conclusions

In this paper we have extended our studies of the effects of the functionalisation of graphite surfaces on the deposition of active metal components. The experiments are relevant not only to the catalysis community but to a number of other areas in which metal nanoparticles are associated with carbon allotropes such as graphene and carbon nanotubes. Whereas there is good evidence that gold can be reduced by hydroxides at the carbon surface, we have found that this does not occur when palladium is adsorbed from palladium chloride solutions. Furthermore the adsorption of palladium ions at the surface is actively hindered by the presence of OH groups at the surface. Freund et al. [28] saw similar effects with deposition of palladium on a magnesium oxide surface from an alkaline solution of palladium chloride. At the pH used in those studies, the palladium was expected to be present in solution as a hydroxide and this is the species that was observed at the surface after deposition. Reduction of the palladium to metal nanoparticles was only achieved by heating to ~ 700 K. In our experiments at lower pH, Pd would be expected to be present predominantly as the chloride in solution [21] whereas Au will form partial hydroxides ([AuCl2(OH)2]− and [AuCl(OH)3]−) [29] These observations reflect the higher affinity of Pd for chloride and so may explain the differences seen in the deposition of the metals from their chloride salts on the HOPG surface.

Thiosulfate adsorption from solution is predominantly dissociative leading to the formation of adsorbed sulfite and sulfur, characterised by S(2p) binding energies of 168.9 and 163.9 eV respectively. Adsorption of thiosulfate from solutions with higher concentrations results in some molecular adsorption of the thiosulfate indicated by a S(2p) peak at ~ 162.9 eV. Subsequent exposure of the surface to an acidic aqueous solution leads to the desorption of the sulfite groups but no discernible change in the concentration of the adsorbed sulfur atoms. However, a sharpening of the S(2p) spectrum after acid treatment may indicate a reduction in the variety of sulfur adsorption environments and hints that whilst the S–S bond is broken when the thiosulfate adsorbs, there remains a degree of interaction. Thiosulfates have been exploited commercially [14] to create stable, gold based catalysts on carbon supports, our experiments show that thiosulfates adsorbed directly onto the carbon surface interact only very weakly, if at all, with gold subsequently deposited from solution. The adsorbed sulfite groups desorb as was the case with treatment with an acid solution and the adsorbed sulfur atoms appear unaffected by the presence of the gold. However, there is a significant and reproducible ~ 0.5 eV shift upwards in XP binding energy of the deposited gold from when it is adsorbed in the presence of a hydroxide alone. This could signify the formation of a predominantly small nanoparticles of gold at the surface or it may reflect a degree of coordination of the deposited gold to the adsorbed sulfur. The AFM images are not sufficiently highly resolved to discriminate between these different scenarios. Commercial catalyst preparation involves complexation of the gold ion in solution and this will form the subject of future investigations into this system.