The samples were prepared by deposition of a SiO2 layer (thickness 8 nm) by a Double Electron Cyclotron Resonance (DECR) plasma process on a Si(100) sample (Sb n-doped with size). SiO2 was evaporated on Si(100) for two main reasons: it is a protective barrier layer that prevents the formation of transition metal silicide through direct interaction with silicon, and SiO2 is a non-wetting substrate that is convenient for transition metal diffusion and aggregation. However the thickness of the SiO2 layer must be thin enough to allow electron conduction through tunnelling for field emission measurements. The sample SiO2/ Si(100) was then transferred into a stainless steel Ultra High Vacuum (UHV) preparation chamber (base vacuum 10 - 10 mbar) where transition metal (TM Co or Fe or a mixture of them) evaporation was performed without air removal. Co (Fe) of grade 99.995 was evaporated with an OMICRON EFM3 effusive source at a pressure within 7 - 10 × 10⁻¹⁰ mbar on the sample heated at during 30 min. The flux rate at 973 K is estimated to 0.025 nm of equivalent layer per minute from an in situ XPS analysis of the Co2p/Si2p signal. In other cases the transition metal is deposited by sputtering within conditions that have been elsewhere described [39] .

As well, the CNTs growth method using a direct current plasma and hot filaments-enhanced catalytic chemical vapor deposition (DC HF CCVD) process has been fully described elsewhere [37] . After the catalyst evaporation, the samples were further transferred into an UHV CVD chamber for the growth of the carbon nanostructures (base pressure lower than 10⁻⁹ mbar). The gas mixture (100 sccm C2H2:H2:NH3) was thermally activated by hot filaments (up to a power) and kinetic energy-activated by polarisation between tantalum grid electrodes with the cathode grid in front of the sample at. The discharge was ignited and stabilized by the electron emission of the hot filaments. This ensured a high concentration of ionic species as well as activated radicals in front of the sample. A small additional negative extraction voltage was put on the sample, which allowed withdrawing a controlled current of ionic species onto the sample, with extraction power. The temperature (973 K) was controlled and regulated by an independent infrared heater set on the rear side of the sample. A Pt/PtRh thermocouple was contacting the rear side of the sample during the temperature rise. This thermocouple was switched off when the polarisation was started. The contact was then used to monitor the electric current onto the sample due to the discharge. The sequences of deposition were the followings: the sample was first heated under vacuum (10 K/min, 573 K, 10 min), then the temperature was risen to 973 K (10 K/min; 40 min) in a H2 atmosphere at 15 mbar. Acetylene and ammonia were introduced. Subsequently the primary discharge and the extraction discharge onto the sample were adjusted to the desired values. The extraction current Ie was set constant throughout the deposition process.

To stop the CNTs growth, the acetylene feedthrough, the polarisation, the filaments and finally the hydrogen feedthrough were subsequently switched off. The references as well as the main characteristics of the sample preparation are displayed in Table 1. According to the nature, the mode of deposition of the catalyst as well as the pressure of the reactive gas mixture, the temperature, the hot filaments power and the plasma power, different carbon nanostructures were allowed to growth, as listed in the following Table 2. Highly HOPG was purchased from Union Carbide.

TEM observations were performed on a TOPCON 002B microscope operating at 200 kV. The samples were scratched with a diamond tip and the material was directly pulled onto an amorphous carbon membrane drilled with holes for direct observations. SEM observations were performed on an XL30S-FEG PHILIPPS working at 3 kV. The nature of the carbon deposit was probed by Raman spectroscopy on a Renishaw apparatus with a He-Ne light source. More structural and spectroscopic data are reported in [39] .

C K-edge measurements were performed at the Laboratoire pour l’Utilisation du Rayonnement Electromagnétique (LURE, ORSAY, France) on the VUV Super-Aco storage ring. They were carried out on the SACEMOR beam line [12] [40] using a high energy TGM monochromator (1200 lines・mm⁻¹ grating, resolution better than 0.2 eV at the C K-edge). The base vacuum was around 10 - 9 mbar. The carbon surface contamination by the optics

of the line does not exceed 1% of the total signal and can be neglected for materials with a high carbon concentration. The spectra were recorded in the total-electron-yield detection (TEY) and partial electron yield (PEY), the last being expected to be less surface-sensitive. Experiments were carried out in two experimental configurations according to the angle α between the sample and incidence of the light: at normal incidence (α ≈ 0˚ with electric field vector E parallel to the surface, and at grazing incidence (α ≈ 90 - 10)˚ with E nearly normal to the surface. The angular variations of the intensity of the C K edge transitions to the and unoccupied states were monitored at these two incidence angles, respectively (Scheme 1). The step was 0.1 eV. The spectra were normalized to the spectrum of a copper grid freshly coated with gold. The energies were calibrated to the carbon ionisation potential (IP). Owing to this calibration procedure, the error on the C K-edge is expected to be no more than 0.2 eV. The energy step was 0.1 eV. The base pressure in the chamber was Torr. The samples were analyzed as such after introduction within the XAS chamber and after a subsequent in situ cleaning by electron bombardment at 500˚C for 2 hours.

The spectra were first corrected for the background by substracting it on the preedge low energy side with a linear background contribution. Then the spectra were normalized with regard to the preedge intensity on one side, and to the absorption intensity at 320 eV on another side, according to. Above this energy range, oscillations due to extended ray-absorption fine structure (EXAFS) occur. Lineshape analyses of the individual transitions were carried out using a home-made software. The spectra were deconvoluted by a procedure where each feature was broadened by a variable Voigt function including an overall gaussian broadening accounting for the experimental resolution set to 0.75 eV, a variable gaussian broadening accounting for electron-phonon interactions, amorphisation effects and the width of the final states distribution, and a lorentzian broadening accounting for the lifetime of the core hole fixed to 0.21 eV for the C K edge [41] . Heavyside step functions broadened by the gaussian experimental resolution were added to these individual contributions. They were located at the threshold of the and absorption transitions, respectively, and they accounted for absorption transitions to the continuum of empty states. Inadequate energy resolution, uncertain normalization process and non-attributed features result in relevant uncertainties in the determination of the intensities.

As the SEM and TEM images clearly illustrate in Figure 1 and Figure 2, respectively, the carbon nanostructures prepared in this study display widely different morphologies according to some variable parameters of the

catalyst preparation (amount of cartalyst deposited measured by the surface ratio Co/Si, mode of Co deposition, and growth conditions (temperature, plasma power and hot filaments power, pressure)) reported in Table 1. Under conditions where the catalyst is deposited by UHV atomic evaporation at moderate pressure (5 - 15 mbar), it is possible to control the nature of the carbon nanostructures [39] . Carbon nanowalls (CNWs) are prepared at low pressure (5 mbar) (Figure 1(f)).

These are graphene sheets that merge in the direction normal to the surface (Figure 2(f)), when the energy of the ions impinging the surface is rather high. Carbon nanoparticules (CNPs) are prepared when the power of the hot filaments is low (Figure 1(c) and Figure 2(c)). Carbon nanofibers were prepared under different conditions.

When the plasma power is high and the catalyst surface concentration is low, then graphene sheets grow in a direction normal to the surface (Figure 1(b) and Figure 2(b)), forming conical nanostructures with the metal particle on top of it. When the catalyst is prepared by sputtering and the plasma power is rather low, then CNFs can grow with graphene sheets parallel to the surface (Figure 1(a) and Figure 2(a)). Strong adhesion of the catalyst to the substrate and low energy ions can explain this mode of growth. Within medium plasma power, carbon nanotubes can yet be grown with graphitic planes in a parallel direction to the fiber axis can yet be prepared (Figure 2(d)). These samples however display different mutual orientation. Highly oriented films are obtained under optimized conditions (Figure 1(d)). Poorly oriented films are also obtained (Figure 1(e)) and the nanotubes show more defects (Figure 2(e)). Anyway the presence of hot filaments heated around 2200 K must be stressed.

They provide hydrogen radicals that are very reactive towards all kinds of amorphous carbon. This is checked in Raman spectra (Figure 3). The most intense Raman spectrum corresponds to sample V as the etching of carbon by hydrogen radicals is less effective. Thus probably carbon not only surrounds the particle but also is spread onto the surface of the sample. It is beyond the scope of this paper to discuss the Raman spectra of these different carbon nanostructures. We must just underline that whatever the sample the D band due to disordered carbon and the G band due to the main tangential vibrations in graphene sheets or shells are very narrow. This indicates that the carbon deposit is selective

To ascertain the reliability of the analysis of CNTs XAS spectra, the XAS spectra of HOPG sample is first recorded. Graphite, with its layered structure and large interlayer separation, is often modelled as a two-dimen- sional solid. In addition, the knowledge of the properties of graphite is a starting point for understanding the structure and properties of many new carbon nanostructures like nanotubes. The two-dimensional nature of graphite results in a strong directionality of the orbitals: orbitals lie within the basal plane, while orbitals are directed perpendicular to this basal plane along the c axis. By tuning the incidence angle of the polarized synchrotron light on the sample, it is possible to excite final states of specific symmetry at the carbon K edge. According to the Fermi gold rule and the dipole selection rules, when the electric field vector E lies within the basal plane (i.e. perpendicular to the c axis, then final states are selected. When the E vector is perpendicular to the plane (i.e. parallel to the c axis, then final states of -symmetry are excited. Thus the intensity of each transition will strongly depend on the angle between the electric field vector E and the c axis. Using calculations of density of states (DOS), it has been possible to assign the features of the C(K) absorption structure (XAS) to specific states of the Brillouin zone in the theoretical band structure of

graphite [29] -[31] . Moreover, by monitoring the angular dependence of the spectral features of the carbon K edge, the symmetries of the final states were determined. Therefore, in the single-crystal graphite, the intensity of a pure transition at 285 eV is shown to be proportional to (with its maximum at glancing incidence), while the intensity of a pure transition is shown to be proportional to cos2 with its maximum at normal incidence) [29] . As a result of this behaviour, the assignments (energy, symmetry, location in Brillouin-zone regions) of the main features in the carbon K edge spectrum of HOPG graphite are determined in agreement with the assignments reported in [29] [31] (Figure 4) and the results are reported in Table 3. In this fitting procedure, the energy and the width of the lorentzian line, quoted to 0.21 eV, are fixed, whereas gaussian shapes of different width for each contribution were allowed to vary. It is clear that the occurrence of these transitions is related to the unoccupied density of states (UDOS) and to final state effects. Thus the physical meaning of the different contributions introduced is not warranted. Nevertheless it is clear that the full absorption spectra can be divided into three regions. At 285.5 eV the peak named A corresponds to the transition towards unoccupied states near Q in the Brillouin zone at 2 eV above the Fermi level [29] -[31] [42] [43] , strongly dependent on the incident angle. The second region within 286.5 - 292.5 eV corresponds to the free electron like interlayer states [22] [27] [30] [43] and adsorbed or chemisorbed molecular states [22] [23] [27] [30] [32] [36] [43] . They are strongly dependent on the chemical treatment of the sample, as indicated in Figure 5 on sample III before and after an in situ thermal treatment at 500˚C.

In this energy range the contributions of and transitions are due to adsorbed functionalized states such as C-O-, C-H, C = O, C-OH ... They are due to the adsorption of various molecules like water, CO, CO₂ or hydrocarbons. HOPG XAS spectra exhibit a exciton at 291.6 eV corresponding to a C-H* bond resonance [23] [24] [30] [43] . The presence of a exciton at 291.6 eV points to a material with well-formed local bonding configuration or good short-range order. This line, which is an atomic-like transition, is used to ascertain the lorentzian width due to the lifetime of the core-hole. Moreover the small shoulder at 286 - 287 eV may also be attributed to a singularity in the DOS of graphite [29] . Another explanation of these lines could be the presence of free-electron-like interlayer states (FELI) in the graphite with -symmetry [30] . The third region above 291.8 eV corresponds to transitions towards empty states of and symmetry states. The assignations of these states closely agree the features of the conduction band diagram as reported in Figure 4. Moreover the angular dependence allows to distinguish between symmetry from symmetry states. Thus at a grazing incidence, in addition to the peak A, other transitions with a -symmetry are involved in transitions named C and H at 295.5 eV and at 316.5 eV which might correspond to or final states near and to final state near Q in the Brillouin zone, respectively. In agreement with these assignations these contributions are maxima at this incidence angle. Reversely from the normalized carbon K edge spectrum of HOPG at normal incidence (Figure 6(a)), the B, D, E, F and G spectral features at 292.5 eV, 297.8 eV, 303.5 eV, 307,5 eV and 311.4 eV, respectively, are mainly transitions to the unoccupied final states. They are,;;; and states in the Brillouin zone, , and near Q, respectively. Other small contributions due to the resonance of transitions of adsorbed molecules can be present around 300 eV and around 296.5 eV, but they are believed to be negligible after heat treatment. In the high energy range, contributions to multiple and single scattering interfere with the transitions and the analysis have not been continued above 320 eV. Dramatic changes occur with incidence of the polarisation light onto the substrate, as displayed in Figure 4 and Figure 6. At normal incidence, the transition A almost disappears, consistent with the orthogonality of the electric field vector with the orientation of the orbitals.

The small intensity of transition A observed may be explained by incomplete polarisation or by a small sample misalignment. Two parameters and are defined to determine more quantitatively the respective contributions of the and transitions at incidence angle. is deduced from the fitting of the carbon absorption spectra by the ratio of the intensity (surface) of -type features over -type features. It is expected to be maximum and minimum when graphite is recorded at a grazing and a normal incidence, respectively. In agreement, S// and S are quoted to 0.93 and 0.08, respectively (Table 4). is another ratio of the heights of A feature by B feature, characteristic of both symmetries. It is expected this ratio yield a crude estimation of the orientation effect.

We examine now different carbon nanostructures at GI and NI incidences, respectively.