2.1. Metal Nano/Micro-Oxides
Materials obtained after HTC of various metal salts in the presence of Mimosa tannin, as well as after calcination in air, are presented in
Figure 1,
Figure 2,
Figure 3 and
Figure 4. Depending on the particle sizes, TEM or SEM pictures are presented, which led to the following, important conclusions.
(i) Depending on the nature of the metal salt, different particle morphologies and sizes can be observed after HTC. In the conditions detailed above, vanadium led to the smallest carbonaceous particles (<0.5 µm), themselves appearing to be aggregates of even smaller grains (<20 nm) as seen in
Figure 1. Chromium led to very similar materials, as seen in
Figure 2, just presenting slightly bigger aggregates (~1 µm). All these metal-doped materials after HTC presented small, poorly spherical carbonaceous particles aggregated together in a nodular, gel-like structure. In contrast, nickel and iron produced bigger (~5 µm) and much more spherical particles showing an apparently solid structure, see
Figure 3 and
Figure 4, respectively.
(ii) As a result, the materials after calcination in air were also quite different. Those produced from V and Cr were nanocrystals of typical size close to 50 nm in the smallest direction. On the contrary, those derived from Ni and Fe were spherical oxide particles having maintained the rounded shape of their carbonaceous precursors, but with a smaller diameter due to pyrolysis-induced shrinkage.
Such extreme differences in the behaviour of V and Cr on the one hand, and of Ni and Fe on the other hand, should be explained by the way the corresponding ions bind to tannin during the hydrothermal process. It is indeed well known that tannins dissolved in water are good metal chelators, and this property has been used for centuries for preparing dyes and inks. Considering that nano-oxides were obtained with V and Cr, it can be conjectured that the corresponding ions were chelated everywhere in the tannin-based polymer network. As a result, the metals were highly dispersed in the carbonaceous structure after HTC, leading to extremely small grains after calcination during which recrystallization could occur, as shown by the needles observed in the right part of
Figure 1. The corresponding reaction scheme is given in
Figure 5, in which a typical oligomer structure is shown (more details can be found elsewhere, e.g., in [
48,
49] and refs. therein).
The case of Ni and Fe is different.
Figure 3 and
Figure 4 indeed clearly show that the materials directly obtained after HTC present the typical solid spherical particles known for pure Mimosa tannin treated in hot pressurised water at 180 °C for 24 h [
50], whereas such classical morphological features were lost by using V and Cr salts (see again
Figure 1 and
Figure 2). Homogeneous hollow spherical particles of diameter around 1 μm were thus recovered after calcination in air at 550 °C. At the beginning of the synthesis, a black solution was obtained immediately after mixing tannin with either ammonium nickel sulphate or ammonium iron(II) sulphate, which was converted into a dark suspension in a reddish solution. However, unlike the former chelation mechanism for which metals were homogeneously dispersed in the bulk of the tannin-based crosslinked network and whose polymerisation was forced by the HTC conditions, it is more likely that both Ni and Fe ions were adsorbed from the solution by the moieties at the surface of carbonaceous tannin microspheres that formed in the autoclave. The scheme presented in
Figure 5 is therefore no longer expected to hold in the bulk of the materials but only at their extreme surface. As a consequence, the metal is concentrated in the outer part of the microspheres, leading to the formation of hollow oxide spheres after the carbon was burnt in air.
This mechanism is further supported by the dumbbell-like structure presented by most hollow oxide capsules. Carbonaceous microspheres were indeed not individualised when recovered from the autoclave, but instead were stuck to each other through broad necks. Such a structure is very well seen in the left part of
Figure 3 and
Figure 4, showing the materials after HTC, as well as the several double-hollow oxide spheres that can be seen in the right part of
Figure 3 and
Figure 4. TEM pictures of iron and nickel oxide particles are given in
Figure 6 and
Figure 7 at different magnifications, showing how they are bonded to each other and how thin is their shell. It can also be seen that, for many of them, double-shelled hollow spheres were obtained. However, the mechanism behind the formation of such structures is still unclear. It suggests the chelation of the metal in two steps at the surface of carbonaceous particles growing from the liquid medium submitted to HTC: the smallest one may correspond to the metallic shell condensed on the tannin-based nuclei whereas the biggest one corresponds to the shell that condensed after the particles finished growing.
Metal oxide nanoparticles have been already synthesised through the HTC technique based on starch [
35,
36] or glucose [
2,
51]. Titirici et al. [
51] also obtained hollow crystalline iron oxide spheres with diameters of 1 µm, for which both the surface area and the thickness of the shell could be varied depending on the experimental ratio carbohydrate/metal salt. For molar ratios from 5:1 up to 15:1, the surface area and the nanoparticles size increased from 22.0 to 83.5 m
2·g
−1 and from 16 to 22 nm, respectively. When saccharides were used as precursors, such hollow structures were assumed to form due to the strong complexation of metals at the more hydrophilic surface of the carbonaceous particles growing in the autoclave, whereas their inner part was more hydrophobic.
The surface area and the pore volumes of the present nano/micro-oxides are gathered in
Table 1, and their values are related to their dispersion state: the higher is the surface area, the smaller is the particle size in the case of non-porous grains and/or the more dispersed is the material in the case of porous grains. However, whether the particles were porous or not, the whole oxides obtained this way were all porous materials simply due to their powdery character. Only the nickel oxide presented a high enough surface area allowing its investigation by adsorption of nitrogen, the other ones were evaluated by adsorption of krypton. As the latter isotherms are not very informative and are limited to a relative pressure of 0.2, only the nitrogen adsorption isotherm of nickel oxide is presented in
Figure 8. The latter shows the typical features of a non-porous material (from the point of view of adsorption), in other words, it is characterised by a surface area which is only external, and the hysteresis loop observed at high relative pressure corresponds to capillary condensation in the mesoporous empty spaces between grains of nanometre size. As a result, only Ni oxide presents a non-negligible pore volume, see
Table 1.
The BET surface area measured on calcined nickel oxide, 161 m
2·g
−1, was much higher than the values reported by Titirici et al. [
51], whereas those measured for the present iron, vanadium and chromium oxides were of the same order of magnitude: 29, 6.5 and 21 m
2·g
−1, respectively. Assuming that the grains are spherical (which is obviously not correct, especially for V oxide) and non-porous, the following equation roughly allows determining an equivalent diameter,
d (µm), for a material having a specific weight
ρ (g·cm
−3) and an external surface area
S (m
2·g
−1):
About 280 and 70 nm were found for V and Cr oxides, respectively, using density values of 3.27 and 4.27 as measured by helium pycnometry, respectively (see
Table 1). Those values of equivalent diameter were consistent with the particle sizes observed in the right parts of
Figure 1 and
Figure 2, respectively. The same calculation can be done for hollow microspheres of Fe oxide, using a density of 5.24 and a measured surface area of 29 m
2·g
−1, leading to particle size of about 40 nm. The latter is in excellent agreement with what can be seen in
Figure 6f. As for NiO hollow microspheres, of density and surface area of 6.10 and 161 m
2·g
−1, respectively, the calculated particle size was close to 6 nm. Again, such order of magnitude perfectly agrees with what can be seen in high-resolution TEM pictures of
Figure 7e,f. Interestingly, hollow NiO microspheres having similar diameters were also produced by ultrasonic spray atomisation but their surface area was lower (11–30 m
2·g
−1, depending on the synthesis conditions) because made of bigger nanoparticles (average size 35–64 nm) [
52].
The XRD spectra of nano/micro-oxides derived from HTC of V, Cr, Ni and Fe salts in the presence of tannins followed by calcination in air are presented in
Figure 9. Based on them, the corresponding major crystalline phases were positively identified as shcherbinaite, eskolaite, bunsenite and hematite, respectively. A few other minor phases were also detected, such as K
2V
18O
45 (explained by the presence of potassium as a major inorganic impurity in tannin), hydrated chromium oxides, and β-Fe
2O
3. The presence of the main minerals is, however, in agreement with the densities determined by helium pycnometry, as the specific weight with can be calculated from the cell parameters and the cell unit atom content of each phase is 3.37, 5.23, 6.81 and 5.28 g·cm
−3 for shcherbinaite, eskolaite, bunsenite, and hematite, respectively. Only the chromium oxide presented a significantly lower measured value of density (see again
Table 1), attributed to the presence of hydrated phases. Finally, application of the well-known Scherrer’s law to the data of
Figure 9 allowed determining the crystallite sizes. Their values, associated with an uncertainty of ±2 nm, are given in
Table 1 and can be compared with the particle sizes also presented in the same Table. The agreement between the two sets of values is very good, except for the biggest particles, but this finding is readily explained by the fact that XRD only accounts for crystallites, whose sizes are lower than—or equal to—those of particles.
2.2. Iron/Carbon Hybrid Materials
When pyrolysed at 900 °C in inert atmosphere instead of being calcined in air at 550 °C, the iron–carbon composite material obtained after HTC led to a mixture of iron and magnetite dispersed in graphitic carbon, as suggested by XRD results given in
Figure 10.
Figure 11a,b show TEM pictures of such materials prepared with 2% and 20% of iron, respectively. The partial graphitisation can be clearly identified for both, and typical capsule-like structures made of oriented carbon layers can indeed be seen. Different mechanisms are reported in the literature to explain the graphitisation of carbon catalysed by transition metals such as iron, and leading to this particular kind of hollow carbon morphology. Thus, such capsules might correspond to the carbon formerly coating Fe nanoparticles and having been converted into a graphite-like material at their contact. Such a kind of graphitisation reaction indeed requires that the metal is directly in contact with carbon [
53]. Moreover, the rest of the material is mainly based on a highly disordered carbon matrix, as already observed in other carbons pyrolysed in the presence of transition metals (see [
53,
54] and refs. therein). The size of graphitic capsules was about 10–20 nm, compatible with that of metal nanoparticles prepared in these conditions and that can indeed be observed in
Figure 11 in the form of dark spots. Moreover, the determination of the crystallite size by application of Scherrer’s equation to the data of
Figure 10 led to values of 20, 17, 14, 17 and 21 nm for materials containing 1%, 2%, 5%, 10% and 20% of iron. The uncertainty on those values of size, all in agreement with TEM observations, was ±2 nm. However, it is worth mentioning that Glatzel et al. [
55] reported another mechanism based on the formation of liquid Fe–C eutectic nanodroplets during pyrolysis, which dissolve amorphous carbon and leave behind during their movement a trail of crystallized graphitic carbon shells. Such a mechanism cannot be completely discarded in the present case.
The materials pyrolysed at 900 °C and containing 1%, 2%, 5%, 10% and 20% of iron presented skeletal densities of 2.25, 2.29, 2.32, 2.42 and 2.67, respectively. Such a trend is consistent with the increasing loading of the materials by iron, expected to produce an increasing level of graphitisation. This phenomenon was checked by Raman spectroscopy, and the corresponding spectra are shown in
Figure 12. Although the spectra all looked similar and presented the typical morphology of highly disordered carbon, the relative intensities of the D and G bands located at 1350 ± 5 cm
−1 and 1595 ± 5 cm
−1, respectively, clearly changed with the amount of Fe. As expected, more iron produced materials with more intense G bands, which directly correspond to more graphitised carbon. The ratio of the D to G band intensities is shown in the inset of
Figure 12 and clearly shows such a trend. Such a characterisation method therefore seems much more accurate than XRD, for which calculated values of carbon “crystallite” size were all around 4–4.1 nm, whatever the sample. Likewise, no trend was observed when the interplanar distances d002 of the Fe-loaded carbons were calculated from the patterns of
Figure 10, all having values within the very narrow range 0.337–0.338 (±0.002) nm. Raman spectroscopy therefore appears to be a much preferred and more sensitive investigation tool in the present case.
Figure 13a,b show N
2 and CO
2 adsorption isotherms at −196 and 0°C, respectively, of most Fe–carbon hybrid materials. The one doped with 2% of iron presented curves and corresponding pore volumes that were almost identical to those of the material loaded at 1% Fe, so this sample was not shown for clarity. All N
2 isotherms were a combination of types I and IV, characteristic of micro-mesoporous solids according to the IUPAC classification [
56]. The hysteresis loop indicates the existence of mesopores, whose volume
Vm (See
Table 2) was calculated as the difference between the total pore volume
V0.97 and the micropore volume calculated by application of the Non-Linear Density Functional Theory (NLDFT) model,
Vμ,NLDFT.
Figure 13c shows the PSD calculated from both N
2 (
Figure 13a) and CO
2 isotherms (
Figure 13b). Two maxima were observed in the PSDs of all materials: one related to ultramicropores (<0.7 nm), and another one corresponding to mesopores (between 2 and 10 nm). The carbon materials presented higher ultramicropore volumes and lower mesopore volumes when increasing the iron content. Thus, the mesopore volume decreased from 25% to 12% with respect to the total pore volume (
V0.97).
Figure 13d shows the changes of BET surface areas and pore volumes by increasing the amount of iron. The BET surface area increased moderately with the iron content, and so did the surface area determined by the NLDFT method.
SNLDFT was found to be significantly higher than
SBET as the former indeed takes into account the narrow microporosity determined from CO
2 isotherms. The increasing surface area of the materials at increasing iron content might appear illogic as (i) Fe is heavier than C, and (ii) Fe is not expected to contribute to the surface area except through the external surface of its (nano)particles. However, increasing (NH
4)
2Fe(SO
4)
2·6H
2O addition produced a decrease of the hydrochar yield (i.e., the weight fraction of solid recovered after HTC of the tannin-metal salt solution) from 65% to 30% at 1 and 20 wt % of Fe, respectively, whereas the carbonisation yield (i.e., the weight fraction of solid recovered after pyrolysis of the hydrochar) was around 51%, irrespective to the hydrochar considered. The total yield to iron-loaded carbon materials based on the initial amount of tannin (i.e., the HTC yield multiplied by the carbonisation yield), thus varied from 33% to 15%. The yield to carbon was much lower for the Fe-richest carbon materials. Although Ni catalysts have been more frequently used to increase the hydrothermal gasification of biomass [
57], it is clear that Fe ions also have a catalytic effect on the production of gas, most probably leading to the activation of the carbon with a resultant higher surface area related to the corresponding production of narrow porosity.
The catalytic effect of Fe on tar yield reduction during dry biomass gasification is indeed well-known [
58], and Fe is usually added to Ni catalysts to increase the selectivity to hydrogen production [
59]. Therefore, the higher evolution of gases, essentially H
2 and CO
2, during the hydrothermal process [
60] is expected to favour the higher porosity and surface area after carbonisation of the iron-loaded carbon materials.
Vμ,N2 was always lower than the one determined by DR method applied to the CO
2 isotherm,
Vμ,CO2, meaning that the materials had an important fraction of narrow microporosity that was not accessible to N
2 at −196 °C, due to the low N
2 diffusion kinetics at this very low temperature.
Table 2 also presents the micropore volumes calculated by the application of the NLDFT model to both N
2 and CO
2 isotherms considered simultaneously. Differences between the micropore volumes determined by DR and NLDFT models are attributed to the overestimation of the micropore volume when applying the DR method.
All the materials listed in
Table 2 were ferromagnetic, as expected from their iron content and as it can be easily observed by approaching a rare-earth magnet to the glass vials containing them. Magnetic porous carbons have been already prepared and suggested for a number of adsorption applications in the liquid phase [
61,
62,
63]. Iron-containing activated carbons can indeed be more easily recovered out of the effluents after some targeted pollutants are removed. The present materials probably have surface areas and porous volumes that still are too low for this kind of application, but they should become perfectly suitable after a short physical activation step. The corresponding gasification of the carbon is indeed known to be catalysed by metals, and therefore the surface area and the pore volumes are expected to develop quickly. As a consequence, the adsorption capacity will be improved and the iron content will increase further, making those materials even more magnetic per unit weight. The method presented here is then quite simple and cost-effective for preparing iron-loaded porous carbons from virtually any precursor soluble in water.