Focused ion beam as a tool for graphene technology: Structural study of processing sequence by electron microscopy

FIBID-C features have been patterned on top of 500-nm-thick oxidized-Si wafers. After a simple surface chemical cleaning of the support chips, the patterning is performed using a commercially available dual beam system. The Ga⁺ beam conditions are 30 kV and beam current 300 pA, irradiation being done under normal incidence. The carbon precursor is phenanthrene (C₁₄H₁₀) molecules, vaporized at 65 °C and injected close (300 µm) to the incidence point of the (electron and) ion beam(s) upon the sample surface via a gas injection system (GIS) nozzle. The whole fabrication process for metal-induced FIBID-C crystallization has been previously described in Ref. 25. Additional data will be published in Ref. 31. More detailed explanation of the FIBID-C deposition and its processing, as well as FIBID technique related aspects, can be found there.

All samples used for the TEM studies, called Case 1, Case 2, and Case 3, have been patterned under the same conditions. FIBID-C nominal thickness is 10 nm for all three cases, which is controlled and can be determined by the deposition time at a fixed aperture/beam current. Only post-deposition processing differs. Respectively, Case 1 corresponds to as-deposited FIBID-C; Case 2 is heat-treated FIBID-C; Case 3 is analogous to Ref. 25, thus, Ni foil-assisted thermal treatment of the FIBID-C ultrathin, patterned layers. For Case 2 and Case 3, the heating conditions are 975 °C lasting for 30 min. An IR lamp furnace operated in vacuum is used for the thermal treatment. The Ni catalyst is a polycrystalline high purity material, structured as a 10-µm-thick foil.

For the TEM specimen preparation, conventional FIB-based fabrication is applied, once fully processed, for each sample type, thus, after thermal treatment if applied (Cases 2 and 3), not for Case 1. A Pt layer is deposited onto the sample surface to protect the FIBID-C patterns during the milling routine of fabricating the cross-sectional TEM-transparent specimens. Lamellas were prepared by a Hitachi FB2100 FIB system. FIBID-C specimens were roughly milled at 40 kV and finally, thinned at 10 kV.

EM-based characterization includes high resolution TEM (HRTEM), EELS, and elemental maps by scanning TEM energy-dispersive X-ray spectroscopy (EDX). The microscope is a JEOL JEM-ARM-200F. Electron beam is operated at 80 kV to prevent knock-on atomic damage of crystalline C nanomaterials, as reported in Ref. 10, and to reduce eventual electron beam induced phenomena on the FIBID-C (metastable) materials, particularly Case 1. The EELS spectra were acquired using TEM mode with an energy resolution of 0.6 eV. Some diffraction analysis had been applied as well, but the characteristics of the samples, the transversal view of the ultrathin layers and imperfect crystallinity, were not favorable to obtaining clear diffraction patterns (hence, not shown).

Instrument and measurement conditions for micro-Raman scattering are equivalent to previously reported.²⁵⁾ Laser excitation wavelength of 532 nm and a spot size of >2 µm have been used.

Figure 1 is a HRTEM micrograph of an as-deposited FIBID-C layer (Case 1). The cross sectional view of the FIBID-C exposes the evenness and smoothness of the as-deposited C layer. No impurities are observed (nor unambiguously detected by EDX, see more details below). The most important aspect to note is that the uniform granular texture, no contrast domains are seen, does not show any structural ordering within the solid layer. This corroborates that the precursor molecules, phenanthrene, do not keep their integrity as expected by the use of FIBID, a FIB-based growth process rooted on the ion-molecule collisions²³⁾ and do not pile up in an ordered or self-assembling manner. The resulting material is constituted by an amorphous material, where chemical bonding and composition changes of original precursor molecules are assumed and reported.^32,33) In summary, as-deposited FIBID-C is an atomically disordered compact material resulting from the crystalline distortion and certain bonding of the precursor molecules. The mechanical vibrational properties of pillar structures fabricated with an equivalent method and analogous precursor indicate that the as-deposited FIBID-C material is amorphous (non-crystalline), where diamond-like carbon (DLC) (sp³) bonds dominate.³⁴⁾ Based on this, as-deposited FIBID-C may be denser than common (porous) amorphous carbon materials and, instead, show more similarities with thin layer DLC materials³⁵⁾ deposited by ion beams³⁶⁾ or plasma CVD.³⁷⁾ Although the SiO₂–C interface mixing caused by the use of the ion beam for the amorphous layer deposition is apparent, (the thickness of) the C layer can be distinguished as to be able to determine its value. Being approximately of 10 nm, therefore, matches the nominal thickness, which is based on the processing conditions (here, deposition time and beam aperture) and as it has been determined by an atomic force microscopy-based calibration of the deposition rate.³¹⁾

Zoom In Zoom Out Reset image size

The transformation resulting from the metal-assisted thermal treatment of the FIBID-C layers (Case 3) is illustrated in Fig. 2. This HRTEM compound image is a zoom of the area marked in the inset TEM image. The amorphous as-deposited FIBID-C layer shows now an obvious layered morphology (>10 layers) which is oriented roughly parallel to the sample support, the SiO₂ substrate. Therefore, observation of graphitization confirms the results of Raman scattering characterization presented in Ref. 25. Although the existence of inclusions (Ni, see EDX discussion below) is clear, its direct and unambiguous correlation to the evidenced graphitization cannot be stated as the crystallization is also found apart from those inclusions. This point can be appreciated for example in the image right side area, but more examples of graphitized materials are included in Fig. 3, Note 1. Another indication related to the Ni contribution is the fact that, quite the opposite, evidence of layering formation does not always occur on top of the inclusions, as it can be seen in the left side edge of the Fig. 2 micrograph. No evidences have been found to attribute the observed layered structure to CNT formation, confirmed by the absence of tiny Ni catalyst nanoparticles.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In the right half image region, another feature can also be observed. Ultimately, incomplete graphitization of the FIBID-C is found next to the SiO₂–C interface. While the SiO₂ layer and C layer can be here easily distinguished, no extended atomic carbon ordering, i.e., graphitic carbon formation, is observed. However, ∼1 nm long chains are visible in this region (red arrows in Fig. 2) which may be antecedents of extended graphitization.³⁸⁾

Zoom out TEM observation and inspection of other areas (not shown) present similar characteristics: formation of graphitic carbon, existence of inclusions, heterogeneous and partial C crystallization. The existence of contrast changes in TEM are a good measure of defects and non-uniformities on crystalline structures.³⁹⁾ In addition to a certain increase of the oddness of the FIBID-C surface, a decrease of the layer thickness can be observed and estimated in ∼20–30% thickness reduction.

A similar trend concerning as-deposited FIBID-C thickness reduction is obtained for the thermal treated FIBID-C, Case 2, i.e., thermally treated without using any metal, e.g., Ni foil (Fig. 4). It accounts for a thickness layer decrease of ∼15–25%. The loss can be expected when graphitization occurs (solid state matter), but the fact that it is also found for Case 2 is an indication of a certain elemental composition change, for instance, partial decomposition by H desorption due to thermal treatment as reported in Refs. 33 and 40.

Zoom In Zoom Out Reset image size

Collecting the observations, the most important fact to mention is that the graphitic conversion of FIBID-C, the transformation into a layered C structure, is never observed for merely heat treated samples. It is clear that no crystalline material is obtained and a granular texture appears more visible in Case 2 than in the non-graphitized portions of Fig. 2 layer (Case 3). As studied from the Raman scattering spectra of thermal annealed FIBID-C,²⁵⁾ in spite of a certain formation of aromatic rings as understood from the distinguishability of D peak and the shrinking of G peak, the broadness of the D and G bands indicates a very low graphitization degree and only incomplete "crystal" relaxation.^41,42) Another manifest feature is that the SiO₂ layer top surface is recovered after thermal treatment, similarly to the previous Case 3, as the C–SiO₂ interface can be easily seen. Interface minor oddness and that of the C layer can probably be correlated. TEM image contrast is fairly uniform throughout the layer.

The morphological and structural direct observation by HRTEM has been complemented with EELS analysis. EELS low energy range informs about the atomic bonding.⁴³⁾ Separate low loss EELS spectra for Case 1, Case 2, and Case 3 are shown in Fig. 5. Characteristic plasmon excitations, corresponding to both bulk and surface modes of the valence electrons, can be observed for the different graphitic materials.⁴⁴⁾ As depicted in Fig. 5, a peak resulting from the π-excitation plasmon loss is clearly promoted for the metal-assisted thermally treated FIBID-C (Case 3), whereas is absent for the as-deposited FIBID-C (Case 1) and significantly lower for heat treated FIBID-C (Case 2). Respect to the signature value of graphene, ∼4.7 eV, the Case 3 peak position located at ∼7 eV is shifted towards higher energies, which is a characteristic property of graphite.⁴⁴⁾ This result is in total agreement with the typical Case 3 Raman spectrum²⁵⁾ and the observed multilayer structure, shown in Fig. 2. For predominantly-sp³ materials, such as DLC or diamond, the lack of this peak is anticipated.⁴⁴⁾

Zoom In Zoom Out Reset image size

The evolution of the plasmon loss peak for π + σ is different. A broad peak between 20–40 eV is recorded for the three cases, like commonly found in the literature for graphite, tetrahedral amorphous C (ta-C) or diamond.⁴⁵⁾ A plasmon loss centered about 24 eV is an indication of the sp³ carbon bonds, consistent with recorded spectra of Case 1 and Case 2 (Fig. 5), but not always observed for so-called amorphous carbons.⁴⁶⁾ Comparing again with single and few layer graphene, the energy shift toward higher energies observed for Case 3, peak at 24 eV plus peak at 27 eV, is an indication of the formation of graphitic carbon,³⁷⁾ while single layer graphene would be typically observed by a peak at ∼15 eV.^39,44) The convoluted peak can be explained by the coexistence of graphitized and non-graphitized FIBID-C, as seen in Fig. 2. Easier appreciation of the shift indicative of graphite tendency for Case 3 material can be found in Fig. 6, where the three spectra are shown all in one, using a π + σ peak-intensity normalization.

Zoom In Zoom Out Reset image size

Not only for completeness, but because a few interesting features have arose, EDX analysis has been applied to the three types of FIBID-C materials discussed in this paper. Some exemplary mapping pictograph (Case 3) is shown in Fig. 7. The chemical identification of the plate-shaped inclusions, similar to the one observed in Fig. 2, indicates that they are composed of mainly Ni, which might be a precipitate from the Ni foil. Interestingly, the Ni plates, with sizes up to 100 nm in length and an average thickness of ∼25 nm, penetrate into the SiO₂. In addition, they are located below graphitized FIBID-C, which suggests that some Ni moves towards the SiO₂–C interface and precipitates. Thus, in some way the dissolved material due to thermal treatment is the Ni. In a strict sense, this feature shows incompatibilities, or in other words suggests particularities, respect the conventional graphene on Ni growth mechanism based on the C dissolution inside the Ni and precipitation upon cooling.⁴⁷⁾ Probably, only by in situ TEM, simultaneous TEM-heat treatment, could be elucidated the precise origin and dynamics of present phenomenon.⁴⁸⁾

Zoom In Zoom Out Reset image size

An additional kind of inclusions is occasionally found (not present in Fig. 7). They consist in ∼150 nm in diameter ball-like particles, which are embedded deep inside the dielectric SiO₂ support. From EDX analysis they appear to be a nickel silicide compound. Some crystalline structure could even be observed. We hypothesize that nickel silicide formation results from a complex reaction of molten Ni and SiO₂ perhaps promoted by some residual Ga atoms from the ion beam and the same presence of C. Again, conclusive thermodynamics involved in the metal-assisted thermal treatment of FIBID-C have not yet been determined and further investigation would be undertaken.⁴⁹⁾

Derived from the profuse characterization and its analysis, two relevant aspects can be understood. First, considering the classification of amorphous carbons based on the ternary phase diagram in Refs. 37 and 49, Case 1, Case 2, and Case 3 materials can be classified. Case 1, as-deposited FIBID-C is a diamond-like carbon hydrogenated material (DLCH): high amount of sp³ bonding, significant H contents and disordered atomic distribution. Case 2, thermally treated FIBID-C would correspond to graphite like carbon with the possibility of remnant H contents [GLC(H)]: increase of sp²/sp³ ratio, partial decomposition, but no apparent structural atomic ordering. Finally, Case 3, metal-assisted heat treated FIBID-C are ultrathin layers of graphite-like or graphitic carbon, whose crystalline structure depends locally on the degree and depth extend of the graphitization. Nanographene nomenclature²⁵⁾ is thus valid.

More details on the variety of amorphous C materials and the necessity to identify them, as they present radically distinctive properties arising from structural (chemical bonding and atomic arrangement) and elemental contents (percentage of H, N, etc.) can be consulted in Refs. 36 and 49. Briefly, in Ref. 49 amorphous carbon family is understood based on a ternary phase diagram whose corners correspond to diamond, graphite and hydrocarbons. Respectively, this classification accounts for sp³, sp², and H material contents. The intermediate materials make up the variety of the amorphous carbon materials, including DLCH, ta-C, GLCH, and so forth. The change from perfect sp² configuration (graphite) can be described by three amorphization stages, as a compound non-linear dependence of 1) decreasing cluster size and 2) increasing percentage of sp³ bonds. Specifically applied to FIBID-C, some additional discussion of amorphousness and chemical bonding for the hereby discussed materials has been included in Ref. 50.

Secondly, the material transformation processes are identified and elucidations on the thermodynamics mechanisms involved can be inferred. The as-deposited FIBID-C layers are metastable under mid-high temperature treatments^32,33,40) and its transformation includes two processes. Carbonization process is found, as a result of H decomposition of the deposited-amorphous phenanthrene molecules. Carbonization not only leads to a purer C material, but additionally implies the break of sp³ chemical bonds, which favors the formation of sp² bonds, the C=C covalent bonding.³⁸⁾

Additionally, the crystallization mechanism, the graphitization process, occurs only when metal-assisted heat treatments are applied. Although our FIBID-C graphitization method has not yet been uniquely clarified, but evidenced, strong arguments for considering surface metal-induced crystallization⁵¹⁾ are indicated. This mechanism applies instead of, typically assumed, C dissolution into Ni and graphene formation by precipitation onto the Ni surface during cooling.⁴⁷⁾ We specifically attribute this particular behavior to the strength of the DLC bonds resulting from the deposition technique (FIBID) of the precursor material, the FIBID-C.^33,35,36) The use of metal foil might also play a significant role as the direct C–Ni contact similar to the one obtained when using carbon and nickel layer deposition⁵²⁾ is not guaranteed in the present case. These two conditions may additionally prevent the activation of a massive C dissolution in Ni and precipitation mechanism.⁴⁷⁾ As an alternative, an actual metal-induced crystallization⁵¹⁾ of the most superficial FIBID-C atoms implicitly provides the subsequently inner FIBID-C atoms with a higher degree of freedom, therefore lowering the energy required for their graphitization. As a result, an in-depth progressive crystallization mechanism could be occurring. The limited mobility is also determined by the already existing strong C=C bonds within the as-deposited FIBID-C.

The remnant stress, or incomplete relaxation, of graphitized FIBID-C is most illustrated by an eventual reduced interlayer space of the graphitic carbon, ranging from ∼2.7 Å to, the standard, ∼3.35 Å for graphite. Four measurement examples are displayed in Fig. 3, Note 2. Compressed graphite, the graphitic carbon of reduced interlayer distance,⁵³⁾ as reported in the literature based on X-ray diffraction data, shows a similar reduction, for example, as a result of pressurization in the GPa range.⁵⁴⁾ A permanent lattice compression of graphitic C is attributed to the fractional conversion of π bonds (sp²) into σ bonds (sp³) bridging (thus, coexisting with) the original graphitic carbon layers.⁵⁵⁾ This aspect directly correlates and can be extrapolated to the EELS results shown in Figs. 5 and 6. The polycrystalline structure of Ni would also be a factor for a local non-uniform catalysis of the graphitization in a nanometer scale, which crystallization heterogeneity and material stress in this scale order are identifiable from the Raman spectra, as mentioned in Refs. 41 and 42 and provided in Fig. 8.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size