Formation of hollow carbon nanoshells from thiol stabilised silver nanoparticles via heat treatment

. Uniform, < 10 nm sized, hollow carbon nano-shells (HCNS) have been prepared via a single-step, thermal treatment of alkanethiol stabilised Ag nanoparticles (TS-AgNP). Direct evidence for the formation of spherical HCNS from TS-AgNP is provided by in situ MEMS heating on Si 3 N 4 supports within a TEM, and ex situ thermal processing of TS-AgNP on carbon nanotube supports. A mechanism is proposed for the thermally driven, templated formation of HCNS from the TS-AgNP stabilising layer, with Ag catalysing the graphitisation of carbon in advance of thermally induced AgNP template removal. This facile processing route provides for excellent size control of the HCNS product via appropriate AgNP template selection. However, a rapid rate of heating was found to be crucial for the formation of well-defined HCNS, whilst a slow heating rate gave a much more disrupted product, comprising predominantly lacy carbon with decreased levels of graphitic ordering, reflecting a competition between the thermal transformation of the TS-layer and the rate of removal of the AgNP template.


Introduction
Carbon is abundant, non-toxic and forms a diverse range of allotropes on the nanoscale, including graphene, nanofibres, nanotubes, quantum dots and fullerenes [1]. In particular, hollow, nanostructured carbons provide for a wide variety of applications including catalysis [2], gas-storage [3] and for drug delivery systems [4]. However, fullerenes have limited volumes for use as containers, whilst quasi one dimensional (1D) carbon nanotubes (CNT) and nanofibers (NF) with larger volumes have limited solubilities and tend to bundle [5].
Hollow carbon nanoshell (HCNS) structures, of size ~ 200 nm, were documented originally as by-products from the carbon-arc processing of CNT [6]. More recently, HCNS structures with sizes ranging from ~ 70 -800 nm have been produced using a variety of sacrificial templating and annealing processes [7][8][9][10]. The generic approach is that of hydrocarbon templating of mesoporous shell SiO2 nanoparticles (NP), [11], thermal processing (at ~ 450 -1000 °C in an inert atmosphere) to create graphitic C, followed by template removal (using dilute HF or NaOH) to create bespoke HCNS structures. As part of this processing route, it is noted that a processing temperature of ~ 600°C is considered necessary for the effective graphitisation of amorphous carbon [12]. Alternative strategies for creating HCNS structures (50 -150 nm) include soft-templating, involving micelles or droplets [13], or template-free methods of production [14], however, these still present significant challenges, e.g. a lack of scalability [15].
Here, we demonstrate the novel use of thiol-stabilised Ag nanoparticles (TS-AgNP) to synthesise hollow carbon nanoshell (HCNS) structures of < 10 nm size. This self-contained process may be performed under vacuum, with the thiol stabilising layer providing the source of carbon and the AgNP template dictating the morphology of the resultant carbon shell, whilst thermal processing at ~ 600 -900°C catalyses HCNS production, in advance of thermally induced template removal.

Experimental
Silver nanoparticles (AgNP) stabilised with 1-dodecanethiol (DDM) were synthesised via a modified Brust-Schiffrin reduction [16] as follows: To a solution of silver nitrate (0.3 mmol / 30 mL ethanol), DDM (0.37 mmol) was added slowly with vigorous stirring for 15 minutes. This was followed by dropwise addition of excess sodium borohydride (saturated / 60 mL ethanol) and the solution stirred vigorously for 2 h. The product was then precipitated through the addition of 250 mL ethanol and refrigeration at -15°C for 24 h. The TS-AgNP product was isolated by vacuum filtration using a 0.2 μm PTFE membrane, washed with 10 mL toluene, followed by washing with 200 mL acetone and drying under vacuum. All reagents were used as purchased from Sigma-Aldrich UK without further purification and syntheses were performed at room temperature (RT) using standard glassware and laboratory equipment. TS-AgNP samples were processed thermally either in situ within a TEM on a micro-electro-mechanical systems (MEMS) chip or ex situ within a furnace on multi-walled carbon nanotube (MWCNT) supports.
Samples for in situ TEM investigation were dissolved in CHCl3 (0.5 mg / mL) and drop-cast onto a MEMS-based EMheaterchip TM (Si3N4 support film; DENSolutions), with drying at RT.
TEM with a heating holder (DENSolutions; SH30) was performed using a JEOL 2100 Plus with a LaB6 source operating at 80 kV, with a Gatan Ultrascan 1000XP camera, or on a JEOL 2100F operating at 100 kV, with a Gatan Orius SC1000 camera; with Oxford Instruments energy dispersive X-ray spectrometers (X-Max 100 TLE, 2100 Plus and X-Max 80T, 2100F, respectively). Two different heating protocols were developed to elucidate the effect of heating rate on the formation of HCNS structures. Firstly, rapid heating from RT (23°C) to 850°C (at 92°C/s), followed by a hold time of 80 minutes, then cool down to RT within 30 s (Thermal Protocol 1 (TP-1)). Secondly, slow heating from RT to 650°C (at 1.5°C/s), followed by a hold time of 2 minutes, then slow heating to 900°C (at 2°C/s) and a further hold time of 15 minutes, followed by cool down to RT within 30 s (Thermal Protocol 2 (TP-2)). TS-AgNP and resultant HCNS sizes and size distributions were determined from TEM micrographs using ImageJ 1.48v [17], Java 1.6.0-20 and OriginPro 8 software.
Micro Raman spectroscopy was performed using a Horiba Jobin Yvon LabRAM HR Raman spectrometer. Spectra were acquired using a 532 nm laser at 0.3 mW power, to avoid local heating of the sample, a 100x objective lens and a 50 µm confocal pinhole. The spectral resolution in this configuration is better than ~1.6 cm -1 . To simultaneously scan a range of Raman shifts, a 600 lines mm -1 rotatable diffraction grating along a path length of 800 mm was employed. Spectra were acquired using a Synapse CCD detector (1024 pixels), thermoelectrically cooled to −60°C. Before the collection of spectra, the instrument was calibrated using the zero-order line and a standard Si(100) reference band at 520.7 cm -1 . Point spectra were acquired over the range 800-2000 cm -1 with an acquisition time of 60 s and 16 accumulations to improve signal to noise ratio. The DuoScan TM functionality was employed to confer an effective laser spot size of 20x5x4 m in the xyz directions, respectively. As such, the obtained spectra represent an average of all HCNS structures (produced in situ) present within a TEM grid window. Spectra were baseline-corrected using a third-order polynomial subtraction and fit with a five-band model analogous to that described by Sadezky et al. [18] using Labspec 6.4.3 software. The G and D2 bands are described by Lorenzian peak shapes, whereas the D1, D3 and D4 bands are Gaussian.
Samples for ex situ thermal processing were prepared by mixing TS-AgNP (3.6 mg in 1.5 ml cyclohexane) with MWCNT (3.9 mg; Nanothinx S. A.), via brief sonication then shaking for 15 min. After isolation by filtration (0.2 μm PTFE membrane) and washing with cyclohexane, ethanol and acetone (10 mL), the resultant solid (TS-AgNP@CNT) was dried for 16 h. The TS-AgNP@CNT (3 mg) was then sealed inside a quartz ampoule under vacuum (10 -5 mbar) and placed inside a pre-heated furnace at 850°C for 180 minutes (Nabertherm GmbH P330 muffle furnace), before cooling naturally, outside the furnace, to RT. As-prepared TS-AgNP@CNT samples and ex situ thermally treated products were dissolved in cyclohexane Complementary Fourier transform infrared spectroscopy (FTIR) of the starting TS-AgNP was performed using a Bruker Tensor 27; ~ 0.5 mg of solid sample was ground with 100 mg KBr and pressed to form a disc, whilst for Nuclear Magnetic Resonance (NMR) samples were prepared by dissolution of TS-AgNP in CDCl3 (2 mg / 4.0 ml) and filtering, prior to data collection using a Bruker AV(III) 500 MHz.

Results
As-synthesised TS-AgNP samples were characterised initially using TEM and EDS to evaluate their starting morphologies prior to thermal processing. Figure 1a presents a bright field, diffraction contrast TEM image of TS-AgNP dispersed onto the planar Si3N4 support of a MEMs heating chip, recorded at RT. In this case, the imaging conditions revealed the crystalline AgNP which appeared circular in projection, consistent with spheres of average size 6.6 ± 1.4 nm ( Figure 1c). Conversely, Figure 1b shows the same field of view, imaged at 850°C, following rapid in situ heating (TP-1). In this case, the resultant HCNS structures were      with graphitic spacing 0.34 nm [19]. It is noted that the walls of the resultant hollow carbon shell nanostructure form and graphitise at elevated temperature [20], prior to removal of the AgNP template, consistent with the catalytic activity of Ag during carbon shell formation [21].
The close proximity of templated G-AgNP and hollow G-HCNS in this instance is indicative of some variability in the activation for this transformation process.
Localised variability in the AgNP template removal was also observed depending on particle position on the support during in situ thermal processing, with fast heating rate to 850°C (TP-1). Figure 4a shows a distribution of distinct HCNS on the Si3N4 electron transparent windows of the MEMS chip, whilst an array of encapsulated CNS-AgNP decorate the edge of the support window. Further, it is noted that silver nanoparticles within the CNS-AgNP (23.3 ± 4.2 nm, n = 6), in this instance, were significantly larger than the starting TS-AgNP (Figure 1a), indicative that some AgNP Ostwald ripening had occurred [22]. These observations are again consistent with the suggestion that localised thermal effects mediate the onset and progression of this transformation process.    Table S1 in the Supporting Information for assignment and further discussion). Of note, the spectra are very similar to each other and consistent with that expected from a disordered graphitic nanocarbon. Conversely, complementary NMR and FTIR data demonstrated that the starting TS-AgNP materials were non-graphitic (SI-4 and SI-5, respectively). a b Figure 5. First order Raman spectra (λex = 532 nm) obtained in the range 800-2000 cm -1 from samples thermally processed using either: (a) fast (TP-1); or (b) slow (TP-2) heating rates, respectively, confirming the disordered graphitic nature of the products. Spectra have been normalised to the intensity of the G band, baseline-corrected using a third-order polynomial subtraction and fit using a five-band model analogous to that described previously [18] (χ 2 = 2.3 and 2.4 for (a) and (b), respectively).

Discussion
Discrete, < 10 nm sized HCNS have been formed from the single-step, thermal processing of TS-AgNP, by ex situ heating on a MWCNT support, or as observed in real time via in situ TEM on a Si3N4 MEMS support. The identical location of the HCNS product with the starting dispersion of TS-AgNP, as observed during in situ TEM thermal processing (TP-1), provides direct evidence for the templated formation of HCNS on AgNP (Figure 1). In particular, the size of these HCNS particles (8.9 ± 1.7 nm for the sample set shown here) is defined by, and slightly larger than, the size of the AgNP template (6.6 ± 1.4 nm), as compared with the previous smallest HCNS reported to date of ~70 nm [23]. Nevertheless, the defining features for the product, i.e. hollow interior and < 10 nm size are clearly evident (Figure 3). For the case of structures observable free from background contributions at the edges of MWCNT supports (Figure 4), HCNS interlayer spacings commensurate with graphitic spacings (0.34 nm) [19] were observed, whilst noting that these structures are likely to be disordered [24] because of their very pronounced curvatures.
With regard to the HCNS products processed on planar Si3N4 supports in situ (Figure 2), definitive visual evidence for graphitisation could not be obtained from phase contrast imaging, considered to be compromised by the morphology and thickness of the support, combined again with the very high levels of curvature associated with such < 10 nm sized shells, acting to mask the structural details of the developed shell walls. However, indirect evidence from complementary Raman spectroscopy confirmed the presence of highly disordered graphitic domains for samples processed in situ (Table SI-1; with high intensity peak ratios indicating the presence of both disordered and graphitic carbon (ID1:IG = 3.08) [18]), suggesting that HCNS graphitisation at elevated temperature is not CNT template specific. Whilst a contribution from the support material on graphitisation cannot be discounted, it is considered that the process of graphitisation is thermally induced and catalysed by Ag [20] in this instance.
Further, the ability to form commensurate HCNS structures by ex situ furnace heating of TS-AgNP ( Figure 4) confirms that this thermally mediated process is not governed by electron beam effects. Indeed, once formed, the HCNS appear to be very stable.
Accordingly, it is proposed that the formation of HCNS, from the thiol stabilising layer of TS-AgNP, proceeds via thermally driven dehydrogenation [25], with the crosslinking of alkane chains, the transformation of sp 3 to sp 2 carbon during graphitisation, and finally Ag template removal ( Figure 6). Indeed, the appearance of carbon shells surrounding AgNP (Figures 3d,e and Figure 4b), during both in situ and ex situ heating experiments, demonstrates that the development of carbon shells proceeds at the surfaces of AgNP, in advance of template removal. Hence, it is considered that the AgNP core acts both as a physical template for HCNS and catalyses the formation of carbon shells [26]. The residual presence of small amounts of Ag within some of the developed HCNS, indicative of incomplete processing, is also consistent with the removal of AgNP after shell formation. It is suggested that Ag migrates through the shells via structural defects [27]. With regard to the EDS observation of low S levels within the product, this heteroatom is known to terminate unsaturated vacancies at graphitic domain edges, resulting in thermodynamic stabilisation [28]. perspective, it is recognised that further work is needed to separate out the competing roles of temperature, heating rate, template size and support material on the development of these HCNS products.
The formation of bespoke HCNS (of sizes > 70 nm) has been achieved to date via the use of multi-step processing, e.g. carbon coating of hard template materials, necessitating a synthetic step and thermal treatment, in advance of chemical template removal. Here, we have demonstrated that thiol stabilised AgNP, can be transformed into < 10 nm sized HCNS, using a facile one-step thermal process. This approach provides for excellent size control of the HCNS product via appropriate AgNP template selection.

Conclusion
Discrete, < 10 nm sized HCNS structures have been formed from TS-AgNP via heat induced transformation, using a facile one-step process, by either ex situ heating on a MWCNT support, or in situ heating within a TEM on a Si3N4 MEMS support. Direct evidence for the templated formation of predominantly complete, spherical HCNS on AgNP under rapid heating conditions is presented, with the size of the resultant HCNS being slightly larger than that of the AgNP template. These very stable HCNS exhibited clearly the twin characteristics of hollow interiors and highly disordered, sphere wall graphitisation. It is proposed that the HCNS formation mechanism, from the thiol stabilising layer of TS-AgNP, proceeds via thermally driven dehydrogenation, with crosslinking of alkane chains and transformation of sp 3 to sp 2 during graphitisation, in advance of AgNP template removal. Accordingly, the AgNP core acts both as a physical template for HCNS and catalyses the graphitisation of carbon. Spatial variations in the development of the reaction product reflects predominantly a competition between the thermal transformation of the TS-layer and loss of the AgNP template effect. In particular, a fast heating rate is found to be critical for the formation of well-defined HCNS, whilst a slow heating rate gave a more variable product, comprising lacey, amorphous, and non-homogeneous carbon structures with decreased levels of graphitic ordering. It is considered that one-step thermal processing of thiol stabilised AgNP provides for excellent size control of the HCNS product, via appropriate AgNP template selection. Subtle differences in the first order Raman spectra are noted for carbon shells prepared in situ using TP-1 and TP-2, including a shift in the position of the apparent G band from ~1595 to ~1590 cm -1 , respectively. The difference in the position of this diagnostic band can be tentatively rationalised by either: (a) The presence of residual Ag in the carbon structure afforded, using a slow heating rate, resulting in a red shifted G band due to either plasmonic heating or charge transfer effects; or (b) differences in the effective curvature of the nanostructures obtained. The general peak shapes in the first order spectra ( Figure 5) and weak intensity features in second order spectra associated with combination modes and overtones (data not shown) are both consistent with the formation of highly disordered graphitic carbon.

Raman Spectroscopy
Also noted are differences in the relative intensities of the apparent D and G bands. Whilst the intensity ratio of D and G bands (ID:IG) is often used to quantify disorder in graphitic nanostructures, its direct application to more disordered graphitic structures, such as those obtained here, is often unreliable. This relates to the apparent ambiguity associated with the use of peak heights or areas to define ID:IG [SI-1], the non-linearity in the relationship between ID:IG and inter-defect distance within the graphitic lattice itself , and the understanding that the region between 1000-1800 cm -1 more likely represents a superposition of at least five different spectral contributions, including D and G bands (SI- Table 1), which require careful deconvolution for comparative analysis.
Accordingly, fitting of the spectra obtained here and subsequent evaluation of the associated peak areas yielded some interesting quantitative information into the extent of disorder and amorphous C in the HCNS. The intensity ratio (as peak area) of the D1 and G bands (ID1:IG)a measure of the disorder/order in the graphitic domainsis high, and greater for the carbon obtained with a slower heating rate, TP-2 (ID1:IG = 3.08 and 3.77 for TP-1 and TP-2, respectively). The intensity ratio (as peak area) of the D3 and G bands (ID3:IG)a metric used to quantity the ratio of amorphous carbon to graphitic domainsis also high and significantly higher for the shells obtained with a slower heating rate (ID3:IG = 1.68 and 2.76 for TP-1 and TP-2, respectively). Alkane CH2 symmetric and asymmetric stretching bands usually observed in DDM IR spectra at 2853 and 2924 cm -1 , respectively, are shifted slightly to 2847 and 2916 cm -1 , as observed in published spectra for DDM stabilised AgNP , whilst the expected twisting-rocking and wagging bands of DDM in the 1150-1400 cm -1 region are observed at 1378 and 1468 cm -1 , respectively; the latter CH2 scissoring band position is typical of all trans zig-zag chains of DDM stabilising AgNP . Peaks due to aromatic carbon functional groups around 1600 cm -1 are not observed. Hence, the starting TS-AgNP is non-graphitic.