Introduction

Silicon carbide (SiC) is broadly used ceramic and semiconductor materials family, which offers interesting material properties arising from its several polytypes. It possesses a high hardness, wear resistance, Youngs modulus, as well as high thermal and chemical stability. Along with the electronic properties, SiC allows for advantageous replacement of Si by SiC in a wide range of specific application fields. They are ranging from applications as a functional ceramic material [1], in microelectromechanical systems (MEMS) [2, 3] through applications in electronics [4] and sensorics [4,5,6,7,8] to quantum electronics and photonics [9, 10]. Every application field requires specific solutions of the SiC synthesis, deposition or growth as well as the integration into a process flow. Low temperature SiC formation processes are often required to achieve the goal of process integration. To meet the requirements, a variety of fabrication techniques for SiC thin films have been proposed and tested. The most widely used are magnetron sputtering [11], chemical vapour deposition (CVD) [12], as well as molecular beam epitaxy (MBE) [13, 14].

Recently, an alternative method of forming SiC thin films has been proposed [15]. This method is based on the application of self-propagating high-temperature synthesis developed for powder mixtures consisting of pure silicon and carbon crystallites [16,17,18]. One of the first attempts of preparing SiC from pure silicon and carbon powder mixtures was demonstrated in [16]. A detailed analysis of the reaction mechanisms involved in this synthesis was reported in [17]. Here, resistive heating was applied to the powder mixture leading to a combustion reaction at relatively high temperature (2000 °C) resulting in the formation of crystalline 3C-SiC. The reported enthalpy of reaction of − 250 kJ/mol is exceptionally high. This value is ten times higher than the value commonly set out as the lower critical limit for a self-sustained reaction. Later publications, for example [18], reported a much lower value of the enthalpy of reaction (− 69 kJ/mol). Furthermore, these studies showed the formation of at least two different SiC crystalline phases: the α- and the β-SiC depending on the nature of the carbon source. They reported results for graphite and carbon black as carbon sources. Other studies were at ambient conditions in air at 1800 °C [19]. This synthesis method is simple and inexpensive. However, this approach is very sensitive to the partial air pressure and prone to oxidation and nitrification, thus not applicable for MEMS fabrication. In this study the enthalpy of reaction was relatively low (− 72.8 kJ/mol) and in agreement with [18]. To enhance the combustion energy in [20] high energy ball milling (HEBM) was applied to the powder mixture to increase the materials reactivity prior annealing. The HEBM of the Si/C powder mixture led to the formation of pure SiC, after combustion, without secondary phases. The proposed mechanism is based on the formation of nano-crystalline silicon as well as amorphous carbon during the HEBM process. This highly metastable system leads to a polymorphic nucleation and growth of SiC during reaction. In [15] the nano-structural confinement is achieved by depositing alternating carbon and silicon layers with a thickness below 100 nm using magnetron sputtering. This one-dimensional compositional modulation increases the contact area between the reactive constituents Si and C thus enhancing the reactivity of the heterogeneous material system. For this case, the reaction enthalpy was determined to be − 70 kJ/mol and four steps of the structural evolution were identified.

In this work, an investigation is presented which shows the influence of the thickness of the silicon-carbon bilayer stack and the annealing temperature on phase formation and composition. It is demonstrated that the bilayer thickness has to be adjusted to the length scale of the silicon and carbon interdiffusion governing the diffusion limited silicon carbide phase formation.

Materials and methods

The 1 µm thick silicon-carbon bilayer stacks were deposited by magnetron sputtering. Two elemental pure targets were used consisting of high purity silicon and carbon. Alternating layers of Si and C, with a layer thickness of 5 or 50 nm each and a bilayer periodicity λ of 10 or 100 nm were deposited, respectively. The λ = 10 nm and 100 nm samples are abbreviated 10 nm and 100 nm samples throughout the paper, respectively. The Si and C bilayer stack was repeatedly deposited at room temperature onto the substrate until an overall film thickness of 1 µm is achieved. As a substrate, boron doped p-type Si(100) with a 5 mΩ cm specific resistance was used.

The samples were annealed in a JetFirst rapid thermal annealing (RTA) furnace from Jepilec (ECM Group) in pure argon atmosphere at atmospheric pressure and in a temperature range between 500 and 1100 °C with annealing times of 5 min. The heating-up rate was set to 10 K/s. The cooling down rate was kept at 4 K/s. The structural investigations were performed by X-ray diffraction (XRD) using a Bruker D5000 operated in grazing incidence mode. The angle of incidence was set to 3°. The 2θ measurement range was from 25° to 80° with a step size of 0.008°. The integration time was set to 1 s/step. Reflection Fourier transform infrared (FTIR) measurements were performed using a Sentech SE 900 integrated into a Biorad FTS 3000 FTIR spectrometer equipped with a CsI beam splitter. The measurements were carried out at near normal incidence (5° deviation from normal incidence) with a spectral resolution of 8 cm−1 in the spectral range between 300 and 4000 cm−1.

Results and discussion

XRD diffraction measurement

Figure 1 displays the XRD diffraction pattern of the RTA annealed samples. The diffraction pattern of the 10 nm samples are shown in Fig. 1a, whereas the XRD investigations of the 100 nm samples are depicted in Fig. 1b.

Fig. 1
figure 1

XRD diffraction pattern of the as deposited and annealed Si/C multilayer stacks having a total layer thickness of 1 µm: a 10 nm samples, b 100 nm samples. The peak positions are indicated as red solid lines for SiC, black dashed lines for Si diffraction peaks and black dashed-dotted-dashed lines for the XRD amorphous material peak position

Full size image

In both cases the results are compared to the as deposited case. For the 10 nm and 100 nm samples a similarity in the behavior of the obtained diffraction spectra can be noticed. Firstly, no peaks related to carbon are visible in any of the diffraction pattern. Secondly, for annealing temperatures below 700 °C only a very broad peak in the 2θ region around 52° can be obtained. Such peak forms are characteristic of XRD amorphous materials. Here, the deposited Si, C or a formed silicon carbon alloy formed at the interface of the Si and C layers by intermixing due to ion bombardment during layer deposition. The origin is the absence of long range order or grain sizes in the sub 10 nm range. Identical peak positions and forms were found for as deposited amorphous Si, C and Si1− xCx alloys [21, 22] supporting the observed behavior. This peak decreases with increasing annealing temperature and start to vanish with annealing temperatures equal or above 700 °C.

Thirdly, new peaks appear in the diffraction spectra for both layer systems above 700 °C. In the case of the 10 nm samples only 3C-SiC related peaks are observed, whereas for the 100 nm samples Si and SiC related peaks could be identified. The appearance of peaks related to different lattice planes for both materials indicate the formation of a polycrystalline layer. The dominating orientations in the formed polycrystalline SiC layer are 3C-SiC(111), 3C-SiC(220) and 3C-SiC(311). For the highest annealing temperature in the 3C-SiC(111) a weak asymmetry is observed. This asymmetry might stem from the formation of hexagonal SiC or the formation of stacking faults in none [111] oriented SiC grains. This feature is denoted with SiC(00n), where n means the real hexagonal structure of the grain cannot be determined. The observed onset of the SiC formation at 700 °C is in between the SiC formation temperature obtained by investigating the interaction of the elemental C [23] and hydrocarbons [24] with clean Si surfaces at ultra-low vacuum conditions being below 500 °C and 670 °C, respectively; and the critical epitaxial growth temperatures of low temperature CVD [12] and MBE [14] (750 °C).

The striking difference of the temperature dependent behavior of the 10 nm and 100 nm samples in the diffraction pattern is the observation of Si(111), Si(220), Si(311), Si(331) and Si(400) diffraction peaks in case of the 100 nm samples. This peaks are observable before a substantial SiC formation can be detected. Additionally, noticeable SiC peaks are appearing at higher temperature in comparison to the 10 nm samples. Consequently, the occurrence of the Si diffraction peaks allow to conclude that the Si recrystallizes in the multilayer system before SiC can be formed.

Furthermore, the 3C-SiC peak intensities of the 10 nm samples are higher than the peak intensities of the 100 nm samples for identical annealing conditions. A possible explanation for this observation is twofold. Firstly, if we assume that SiC is formed by the diffusion of carbon into the adjacent silicon layer and of the silicon into the adjacent carbon layer. The regions of the silicon and carbon layers where the composition is changed are the interaction zones. They are located at both sides of the Si–C heterojunction. In these interaction zones SiC is formed, if the critical values of the carbon or silicon concentrations in the silicon or carbon layer are exceeded, respectively. The sum of the width of the SiC phase formation zones (transformation zones) is smaller than the sum of the interaction zones, but also limited by the carbon and silicon diffusion coefficients and their diffusion lengths. At these conditions the total amount of the formed 3C-SiC is proportional to the number of interfaces in the multilayer stack. For the 10 nm samples the number of interfaces is ten times higher compared to the 100 nm samples. The absence of the Si related peaks in the 10 nm samples indicate that the Si might be completely consumed in the SiC phase formation reaction driven by the diffusional interaction of Si and C. Therefore, the phase transformation length is larger than 5 nm and smaller than 50 nm, because in the 100 nm samples the Si peaks are still evident even for the sample annealed at the highest annealing temperature. Secondly, the Si recrystallization, observed in the 100 nm samples, reduces the thickness of the interaction zone, because the diffusion coefficient in the recrystallized Si layers is reduced. This in turn reduces the thickness of the transformation zone where SiC is formed.

FTIR reflection measurements

FTIR reflection investigation were conducted to gain complementary knowledge on the composition of the as deposited and reacted layers. The obtained infrared reflection spectra are given in Fig. 2. Figure 2a, b display the spectra of for the 10 nm and 100 nm samples, respectively.

Fig. 2
figure 2

FTIR reflectivity spectra of the as deposited and RTA annealed Si/C multilayer samples with a total layer thickness of 1 µm: a 10 nm samples, b 100 nm samples. The spectra were recorded at near normal incidence (5° off of the normal incidence). The arrows mark the phonon positions of different bonds and materials: 1, 2, 3: SiO2; 4, 8, 9, 10: 3C-SiC; 5, 6: Graphite; 7: CO2; 11: SiO–H; CSi: Carbon on silicon lattice site

Full size image

The FTIR spectra of the 10 nm samples (Fig. 2a) and the 100 nm samples (Fig. 2b) show a complex, but qualitatively similar structure. They consist of two main elements: (1) a long range oscillation stemming from the total thickness of the deposited multilayer Si/C stack and (2) a modulation of the thickness oscillation by phonon peaks. The wavelength of the oscillation is slightly increasing with the annealing temperature. This is attributed to the densification of the constituent Si and C layers and the SiC formation evolving with increasing annealing temperatures. The peaks modulating the thickness oscillation of the reflectivity spectra appear in different wavenumber regions. The most prominent region is located between 400 and 2000 cm−1. These peaks can be grouped and assigned to originate from different materials and bonds. The most prominent peak 4 having the highest intensity is located in the region between 785 and 800 cm−1 in dependence on the annealing temperature and can be assigned to the 3C-SiC TO phonon vibration [25]. The vibration frequency of the TO phonon increases with increasing annealing temperature. Taking into account that the TO phonon peak position of unstressed 3C-SiC is located at 796.5 cm−1 [26], it can be concluded that with increasing annealing temperature the residual stress in the 3C-SiC changes from tensile to compressive stress.

The 3C-SiC TO peak intensity is increasing with increasing annealing temperature as well. Peak intensity is directly correlated with the SiC volume formed in the ongoing temperature dependent Si/C to SiC phase transformation process, i.e. higher peak intensities are indicative for a larger SiC volume. Therefore, higher annealing temperatures intensify the SiC phase formation supporting the XRD investigations. Interestingly, as can be seen in Fig. 2a, b, even at very low annealing temperatures for both layer stacks in the 3C-SiC TO phonon vibration frequency region the reflectivity is changed to higher values. This might originate from Si–C bond and SiC precipitation formation in the near interface region between the Si and C layers [27]. Such possible low transformation temperatures are supported by the research conducted in [23, 24].

The TO peak in Fig. 2a is asymmetric which might originate from the nanocrystalline nature of the formed 3C-SiC having grain sizes in the order of the thickness of the interacting Si and C layers in the bilayer stack. Therefore, SiC grains dimension around 5 nm can be expected for the 10 nm samples. This is reflected by the asymmetric line shape caused by planar defects and phonon confinement [28] as evidenced in Fig. 2a. In the reflectivity spectra 3C-SiC two phonon peaks can be identified. They are the following: (1) 3C-SiC TO + LA peak located at 1310 cm−1 (peak 8 in Fig. 2a), (2) 3C-SiC 2TO phonon vibration at 1525 cm−1 (peak 9 in Fig. 2a), and (3) the TO + LO phonon vibration at 1630 cm−1(peak 10 in Fig. 2a) [25]. Peak 8 and peak 9 are overlapping or close to the D- and G-bands of graphite or graphene [29]. The appearance of this carbon related peaks near the SiC two phonon peaks are indicative for the existence of residual carbon in the reacted multilayer, which might be a hint for a non-proper stoichiometric ratio of the as deposited Si/C bilayers.

Beside the Si, C and SiC related phonon vibrations other groups of peaks were observed. The peaks 1, 2, 3 and 11 can be assigned to the SiO2 rocking (455 cm−1) [27], symmetrical stretching (800 cm−1) [27], asymmetrical stretching (1085 cm−1) [27] and the SiO–H stretching (3736 cm−1 and 3745 cm−1) [30] modes, respectively. The peak 7 at 2345 cm−1 is due to CO2. The appearance of SiO2 related phonon peaks indicate an oxidation of Si before or during the Si/C to SiC phase transformation. This might be due to outgassing of the carbon layers in the Si–C multilayer stack during the processing in pure Ar atmosphere unable to remove water and oxygen adsorption efficiently from the surface and intercalated water and oxygen from the interfaces and the carbon layers. The SiO2 symmetrical stretching mode and the nanocrystallinity are complicating the correct interpretation of the 3C-SiC TO phonon position, because they are contributing to the peak shift and broadening. For this reason a residual stress analysis can only be done qualitatively.

Conclusions

The interdiffusion driven solid state transformation in Si/C multilayers is able to transform the bilayer stack into nanocrystalline silicon carbide if the single layer thickness accounts for the transformation length and the atomic density in the unreacted as deposited silicon and carbon layers. A further prerequisite of the formation of a pure silicon carbide layer is a stoichiometric proportion of carbon and silicon in the bilayer stack. The onset of the transformation process at 700 °C shifts to higher temperatures if recrystallization of the silicon layer occurs.