Journal of Nanoparticle Research 1: 293–303, 1999. © 1999 Kluwer Academic Publishers. Printed in the Netherlands.
Silicon carbide nanoparticles produced by CO2 laser pyrolysis of SiH4 /C2 H2 gas mixtures in a flow reactor F. Huisken1 , B. Kohn1 , R. Alexandrescu2 , S. Cojocaru2 , A. Crunteanu2 , G. Ledoux3 and C. Reynaud3 Max-Planck-Institut f¨ur Str¨omungsforschung, Bunsenstr. 10, D-37073 G¨ottingen, Germany (E-mail:
[email protected]); 2 National Institute for Lasers, Plasma and Radiation Physics, P.O. Box MG-36, R-76900 Bucharest, Romania; 3 Service des Photons, Atomes et Mol´ecules, CEA-Saclay, F-91191 Gif/Yvette Cedex, France
1
Received 1 December 1998; accepted in revised form 25 March 1999
Key words: nanoclusters, powder synthesis, FTIR spectroscopy, laser prolysis
Abstract Pulsed CO2 -laser-induced decomposition of different mixtures of SiH4 and C2 H2 in a flow reactor has been employed to produce silicon carbide clusters and nanoparticles with varying content of carbon. The as-synthesized species were extracted from the reaction zone by a conical nozzle and expanded into the source chamber of a cluster beam apparatus where, after having traversed a differential chamber, they were analyzed with a time-of-flight mass spectrometer. Thin films of silicon carbide nanoclusters were produced by depositing the clusters at low energy on potassium bromide and sapphire windows mounted into the differential chamber. At the same time, Si and SiC nanoparticles were collected in a filter placed into the exhaust line of the flow reactor. Both beam and powder samples were characterized by FTIR spectroscopy. The close resemblance of the spectra suggests that the composition of the beam and powder particles obtained during the same run is nearly identical. XRD spectroscopy could only be employed for the investigation of the powders. It was found that CO2 laser pyrolysis is ideally suited to produce silicon carbide nanoparticles with a high degree of crystallinity. Nanopowders produced from the pyrolysis of a stoichiometric (2 : 1) mixture of SiH4 /C2 H2 were found to contain particles or domains of pure silicon. The characteristic silicon features in the FTIR and XRD spectra, however, disappeared when C2 H2 was applied in excess. PACS codes: 61.46.+w, 61.10.-i, 78.30.-j, 82.80.Ms 1. Introduction Due to various technological applications, nanoparticles of silicon carbide have attracted the interest of scientists and engineers. For example, ceramic parts made of silicon carbide by sintering SiC nanoparticles offer enhanced performance as far as their mechanical and thermal properties are concerned. Another important feature of nanosized SiC is its wide band gap and expected ability to exhibit photoluminescence (PL) at shorter wavelength than silicon nanoparticles, making it a promising material for optoelectronic devices. This expectation is sustained by the observation of blue light emission from electrochemical formed porous SiC [1].
An efficient technique to fabricate nanosized particles of silicon carbide is based on the pyrolysis of silicon- and carbon-containing gas phase precursors, such as silane and hydrocarbons, respectively. This technique, commonly referred to as chemical vapor deposition (CVD), can be used to produce very fine powders and thin films. From the many possibilities to dissociate the precursor molecules, CO2 -laser-induced decomposition of SiH4 /C2 H2 or similar gas mixtures in a flow reactor has been shown to be particularly useful for the production of ultraclean silicon carbide particles in the nanometer size regime [2–7]. However, with typical diameters above 10 nm, the particles are too large for many applications.
294 In order to produce nanoparticles from various materials in the size regime between 2 and 10 nm, we have recently developed a nanocluster source which combines the laser-driven CVD reactor with a supersonic expansion of the nascent clusters into a high vacuum molecular beam apparatus [8]. It could be shown by time-of-flight mass spectrometry (TOFMS) that the source is capable of producing carbon [9,10], silicon [10,11], and iron [12] nanoclusters in the desired size regime. In case of silicon, the nanoclusters were also deposited on various substrates which were then analyzed by conventional transmission electron microscopy (TEM) [13], high resolution electron microscopy (HREM) [14], Raman [15,16], and photoluminescence [16–18] spectroscopy. The silicon studies revealed that (i) low energy cluster beam deposition yields coatings of loosely stacked nanoparticles with the same diameters as determined in the gas phase with the mass spectrometer, (ii) that the particles have perfect monocrystalline structure and are surrounded by a 1–2 nm thick oxide layer, and (iii) that silicon nanoclusters with diameters around 4 nm exhibit very strong photoluminescence in the red. In the present study we have employed the same technique to produce silicon carbide nanoclusters and ultrafine powders of SiC by pyrolyzing a mixture of SiH4 and C2 H2 , at varying content of C2 H2 , with a pulsed CO2 laser. Nanoclusters have been extracted from the reaction zone and transferred into a freely propagating ‘molecular beam’. They were deposited on KBr and sapphire substrates, and their size distribution was
Figure 1. Schematic view of the experimental setup.
determined in situ by gas phase mass spectrometry. During each experiment, ultrafine nanoscaled powder was collected in the exhaust line of the flow reactor. The beam-deposited nanoclusters as well as the powders were characterized by Fourier transform infrared spectroscopy (FTIR). In addition, X-ray diffraction (XRD) was used to characterize the powders. 2. Experimental The experiments have been carried out in a cluster beam apparatus that has been described in detail earlier [8–11]. For the present purpose it is sufficient to concentrate on the most important aspects that will be discussed along with the schematic view given in Figure 1. As is shown in the enlarged inset, a conical nozzle with a 200-µm-diameter opening projects from the right into the stream of the reactant gas, either pure SiH4 or a mixture of SiH4 and C2 H2 , coming from the top through a 3-mm-diameter inlet tube. The radiation of a pulsed CO2 laser is focused (f = 200 mm) into the gas stream a little bit above the nozzle tip. While most of the clusters and nanoparticles produced during the CO2 laser pulse are pumped away through the funnel and collected in a filter outside of the apparatus, a minor part is expanded through the nozzle into the evacuated source chamber. A conical skimmer with 0.6-mm-diameter opening is then used to produce a well-defined cluster beam and to transfer the nanoclusters into the high-vacuum (5 × 10−5 mbar) of the differential chamber. Here they are deposited on a room
295 Table 1. Flow rates of reaction gases. The crosses (×) indicate that the respective samples were prepared and analyzed Run #
SiC0 SiC1 SiC2 SiC3 SiC4 SiC0x SiC3x
SiH4 (sccm)
C2 H2 (sccm)
25 25 25 25 25 25 25
0 3 6 12.5 25 0 12.5
Beam Deposition on KBr
on Sapphire
– – – – – × ×
× × × × – – –
Powder
× × × × × × ×
temperature substrate (KBr or sapphire window) that can be moved into the cluster beam. Typical deposition times were 2 h. If the substrate holder is not in place, the nanoclusters pass through a 3–mm–diameter pinhole into the third ultra-high-vacuum chamber that enhouses the TOFMS. Here the clusters are ionized by the radiation of an ArF excimer laser (λ = 193 nm) and massanalyzed in a simple Wiley and McLaren TOFMS [11]. The flow reactor has been operated with various mixtures of SiH4 and C2 H2 as is indicated in Table 1. The cluster beam deposition on the different substrates has been performed in two separate but successive experiments. In order to permit comparison with earlier experiments, the first run of each experiment (SiC0 and SiC0x) has been made without acetylene. In runs SiC3 and SiC3x, the flow rates correspond to a stoichiometric mixture in which each Si atom can find its partner C provided that all molecules are pyrolyzed. Finally, in run SiC4, carbon was supplied in excess. The other flow reactor conditions were as follows: flow rate of the buffer gas (helium): 1100 sccm; total pressure in the flow reactor: 330 mbar; pulse energy of the CO2 laser: 45 mJ; laser line: 10µP30 (934.9 cm−1 ). As already mentioned, the Si and SiC nanoclusters were deposited in the beam on either KBr or sapphire. Deposition on KBr – the material that is more suitable for infrared absorption spectroscopy, i.e. the purpose of the present study – has been carried out for run SiC0x and SiC3x, respectively. For photoluminescence studies, which were also planned, the sapphire substrate seemed more appropriate. The powder was collected in all experiments. In each experiment, the cluster beam was first characterized in situ with the time-of-flight mass spectrometer. For this purpose, the entire particle pulse ejected from the flow reactor was analyzed by varying the delay between the CO2 laser and the interrogating excimer
laser in small time steps of 10 µs [16]. The final mass distribution averaged over the entire cluster pulse is then obtained by summing up all TOF spectra. Except for SiC0 (and SiC0x) it is difficult to convert the mass distribution into a size distribution since the mass spectra do not reveal the details of the nanocluster composition. Another point which is important to make is the fact that the nanoclusters collected in the beam need not necessarily be identical with the particles constituting the powder collected in the filter. Since the molecular beam particles are extracted from the reaction zone before their growth is completed they are usually smaller than the powder particles. In addition, as a result of their supersonic expansion in the helium carrier gas, they are cooled much more rapidly. This may also have a strong influence on the crystalline structure of the particles. The IR spectroscopy of the fine powders was studied with a Fourier transform infrared (FTIR) spectrometer (Nicolet, model 550). For this purpose, the powders were pressed into KBr pellets. The spectra covering the spectral range from 400 to 4000 cm−1 were obtained by integrating over 64 individual scans. The crystalline structure of the powder samples was determined by means of X-ray diffraction (XRD) using a Philips PW1400 diffractometer with a Cu-Kα radiation source (λ = 0.154 nm). Unfortunately the beam-deposited samples were too thin to give sufficient XRD signal. So this technique could only be employed to characterize the fine powders.
3. Results 3.1. Nanoclusters 3.1.1. Mass spectrometry Time-of-flight mass spectroscopy has been used for in situ characterization of the nanoparticles in the gas phase before and after the preparation of each sample. In order to probe the entire cluster pulse, the delay 1t between the firing of the CO2 and the probing with the excimer laser was varied from 320 to 450 µs in steps of 10 µs (for details see Refs. [11] and [16]). Figure 2 shows the set of mass spectra obtained in this way during the production of sample SiC3. In this figure we have plotted the individual time-of-flight distributions of the nanoclusters sampled in the different parts of the cluster pulse. It is clearly seen that the mass distribution shifts to larger masses when the cluster pulse is probed
296
Figure 3. Normalized mass spectra of Si and SiC nanoclusters measured in the beam. Table 2. Parameters of the mass spectra displayed in Figure 3
Figure 2. Time-of-flight mass spectra of SiC nanoclusters produced under the conditions of experiment SiC3. The spectra are plotted as a function of the delay between CO2 and excimer laser.
at a later time. The same behavior has been observed before in our silicon nanocluster study [11,16] . It is indicative of a mass separation during the evolution of the cluster pulse and can be exploited to achieve a significant reduction of the size dispersion of the neutral clusters. The total mass distribution is obtained by summing over all individual spectra, fitting a Gauss distribution to the resulting spectrum, and numerically converting the fitted curve to a mass distribution [11,16]. The result of this pocedure is shown in Figure 3 by the dashed curve. Also included in Figure 3 are the results for pure silicon clusters (run SiC0; solid curve) and the experiment in which a 1 : 1 silane/acetylene mixture was used (SiC4; dotted curve). In order to determine the average diameters of the nanoparticles from the mass distributions, both their composition and crystalline phase must be known. This information is only available for the pure silicon nanoparticles which were shown to have perfect monocrystalline structure of the diamond cubic type [14] (Table 2). Given the density of bulk silicon
Run #
Average mass [amu]
FWHM [amu]
Mean diameter [nm]
FWHM [nm]
SiC0 SiC3 SiC4
29550 23660 33750
37800 29680 33900
3.4 3.0∗ 3.2
1.4 1.3∗ 1.2
∗
estimated values
(ρ = 2.33 g/cm3 [19]), the mass distribution shown in Figure 3 can be converted into a size distribution with a mean diameter of 3.4 nm and a full width at half maximum (FWHM) of 1.4 nm. In anticipation of the results of the XRD studies which reveal that the SiC4 powder consists almost exclusively of silicon carbide in its β-phase (and assuming that the beam particles have the same crystalline structure with a density of ρ = 3.21 g/cm3 [19]) we calculate a mean diameter of 3.2 nm and a FWHM of 1.2 nm for the size distribution of the SiC nanoclusters in the beam. The XRD analysis of the SiC3 powder shows contributions from crystalline SiC as well as crystalline Si. To have at least an estimate, we assumed an average density ρ = 2.77 g/cm3 and obtained a mean diameter of 3.0 nm and a width of 1.3 nm. To avoid fragmentation of the ionized nanoclusters as well as their multiple ionization, the mass spectra were measured at very moderate fluence of the excimer laser (∼100 µJ/cm2 ). As is known from our detailed mass spectrometric study on pure Si clusters [11], this fluence value is low enough to exclude such artifacts. Indeed, as can be clearly seen in Figure 2,
297 fragmentation (which would give rise to enhanced signal at lower masses) and double ionization (which √ would result in additional peaks or shoulders at tmax / 2 when tmax is the maximum position for the singly ionized nanoclusters) did not occur. 3.1.2. FTIR spectroscopy The silicon and silicon carbide nanoclusters produced in run SiC0x and SiC3x, respectively, have been deposited in the beam on KBr windows. Approximately 4 weeks later, the deposits were analyzed with a FTIR spectrometer. The corresponding spectra are displayed in Figure 4 by the solid line. The dashed curves represent the IR spectra that were recorded somewhat later from the powder collected during the same runs. For the present purpose it was sufficient to reproduce only the spectral range between 3200 and 300 cm−1 . At first we must state that the IR absorption behavior of the beam-deposited particles does not differ very much from that of the respective powders. Moreover, the two spectra corresponding to run SiC0x resemble very much an infrared absorption spectrum of oxidized silicon nanoparticles published very recently
Figure 4. FTIR spectra of pure silicon (SiC0x) and silicon carbide nanoclusters (SiC3x) deposited in the beam on KBr windows. The dashed curves represent the IR spectra recorded from the powder obtained during the same runs. The upwards-directed arrows refer to Si–O–Si vibrations while the downwards-directed arrows are associated with Si–H bond vibrations.
by Sercel and coworkers [20]. It is dominated by a band at 1080 cm−1 with a shoulder on the blue side at 1140 cm−1 . This doublet is immediately assigned to the asymmetric Si–O–Si stretch [20]. The much smaller sharp peak at 880 cm−1 is attributed to the Si–O–Si bending mode [21] or a Si–O–Si vibration in a (SiO)n ring configuration [22] while the broad feature at 450 cm−1 is generally accepted to belong to an out-of-plane rocking mode [20,23]. The vibrations just discussed are related to Si–O bonds and indicated in the figure by upwards-directed arrows. In contrast, all features labeled with downwardsdirected arrows are attributed to vibrations involving a Si–H bonding group. While the band at 2100 cm−1 has previously been assigned to the stretching vibrations of –SiHx (x = 1−3) bonded to Si [24,25], the 2245 cm−1 band can be associated with the stretch of a Si–H group back-bonded by one or two oxygen atoms [24,25]. The Si–H wagging mode gives rise to a weak feature at 640 cm−1 [22–24]. The smaller feature observed at 800 cm−1 in the powder spectrum could be due to a coupled mode involving O and H displacements in the Si–O–Si–H local bonding group [26]. It should be emphasized that all Si–H features are much weaker than the bands assigned to Si–O–Si, indicating that hydrogenation of the silicon nanoparticles in the flow reactor is not a very important process. The group of weak bands between 1200 and 1800 cm−1 as well as the aliphatic CH stretching bands around 2900 cm−1 are due to contaminations either in the FTIR apparatus or the substrate. They can be due to C=O bonds (1713 cm−1 ), aliphatic CH deformations (1456 and 1400 cm−1 ), and to C–O single bond absorptions at 1280 cm−1 . Many of the features just discussed are also reproduced in spectrum SiC3x. However, pronounced differences are found between 700 and 1200 cm−1 . The silicon oxide peak at 1080 cm−1 is somewhat reduced (68% of its original amplitude) and, in addition, a new strong peak has emerged at 840 cm−1 . This peak resembles very much the major absorption feature of nanoscale SiC powders observed by Cauchetier and coworkers [3,7] in their systematic study on the continuous wave (cw) laser pyrolysis of SiH4 /C2 H2 mixtures. Therefore, it can be assigned without any doubt to SiC. Summarizing the FTIR spectra of beam-deposited nanoparticles, we would like to point out the following results. In the experiment SiC0x, in which pure silicon nanoclusters were produced, silicon is identified by the absorption of its oxide at 1080 cm−1 . When a
298 stoichiometric mixture of SiH4 and C2 H2 is used (experiment SiC3x) the IR spectrum of the deposited nanoparticles is characterized by essentially two absorptions, one being due to SiC and the other belonging to SiO. The SiC peak proves that indeed silicon carbide is formed. In contrast, the Si–O peak is difficult to interpret at the present stage of data analysis. It could be that the Si–O absorption just arises from oxidized SiC; but it could also be that the rather strong Si–O absorption results from the presence of pure silicon nanoparticles (which were oxidized later) or from silicon carbide particles with domains of pure silicon at the surface. The presence of initially pure silicon can easily be rationalized if one assumes that, in the flow reactor, a major part of acetylene is not or incompletely decomposed. For a 2 : 1 silane/acetylene mixture (SiC3x), an incomplete acetylene pyrolysis will result in a deficit of carbon and an abundance of silicon. Finally, the FTIR spectra reveal that the SiC/Si ratio of the nanoparticles collected in the beam is nearly the same as that of the fine powder. 3.2. Powders 3.2.1. FTIR spectroscopy For a systematic characterization of the fine powders collected in the exhaust line of the flow reactor as a function of the flow rate of C2 H2 , we have analyzed the samples SiC0–SiC4 with the FTIR spectrometer. For this purpose the powder was pressed into KBr pellets. The set of infrared spectra obtained in this way is displayed in Figure 5. The uppermost curve, corresponding to pure silicon powder (SiC0), features two strong absorptions at 1080 and 1180 cm−1 due to the excitation of the asymmetric Si–O–Si stretching vibrations. Comparing with the powder spectrum SiC0x that is plotted by the dashed curve, it is noticed that there is almost no difference in the spectral region of interest, i.e. below 1500 cm−1 . If acetylene is added to the reaction gas the Si–O band decreases and a new band appears at 850 cm−1 , between the small 800 and 880 cm−1 features already discussed. This band, which is due to the absorption of Si–C, rises at the expense of the Si–O peak if the concentration of acetylene is further increased. For the stoichiometric mixture of SiH4 and C2 H2 both Si–O and Si–C bands are observed (spectrum SiC3). Comparison with the earlier result of sample SiC3x, which is plotted by the dotted curve, reveals a significant difference as far as the ratio of the Si–C and Si–O bands is concerned.
Figure 5. FTIR spectra of fine powders of silicon or silicon carbide collected in the exhaust line of the flow reactor during runs SiC0–SiC4 (solid lines). The dotted curves represent the results obtained for the powders SiC0x and SiC3x (same as in Figure 4).
Thus, it appears that spectrum SiC3x compares more favorably with the spectrum SiC2. The reason for this discrepancy is not yet understood. It could be that the two powders SiC3 and SiC3x display different degrees of oxidation; but more likely explanations are that the power of the CO2 laser and/or the actual concentration of C2 H2 were somewhat higher in the SiC3 experiment. The bottom spectrum (SiC4) has been measured for the powder being prepared with the highest concentration of acetylene (ratio of Si to C atoms = 1 : 2). This spectrum features a strong Si–C band at 842 cm−1 and only a rather weak contribution from Si–O, which is probably due to oxidized SiC. Hence, we can conclude that, under the conditions of experiment SiC4, the nanoparticles collected in the flow reactor consist predominantly of SiC. 3.2.2. XRD analysis While the IR spectroscopy only yields information on the composition of the nanoparticles, X-ray diffraction can be used to analyze their crystalline structure. Unfortunately, the nanoparticle deposits obtained in the beam are too thin to give sufficient diffraction signal. Thus, we could only use the XRD technique to analyze the
299 Table 3. XRD data for sample SiC0 SiC0
ASTM 27-1402
22 ( )
d (Å)
I /Imax (%)
22 (◦ )
I /Imax (%)
28.8 35.0 47.5 56.5 69.5 76.2
3.09 2.56 1.91 1.62 1.35 1.25
100 11 37 19.5 5 7
28.44 — 47.30 56.12 69.13 76.38
100 — 55 30 6 11
◦
Table 4. XRD data for sample SiC4 SiC4
Figure 6. XRD patterns of fine powders of silicon or silicon carbide collected in the exhaust line of the flow reactor.
fine powders collected in the flow reactor. The corresponding diffraction spectra are plotted in Figure 6 in the same order as the FTIR spectra in Figure 5. During the experiment SiC0, only silane was used as reactant gas so that we expect only silicon peaks in the XRD spectrum. Indeed, if we compare the experimental peak positions with the data base values [27], as is done in Table 3, we note rather good agreement and conclude that the corresponding nanopowder consists of silicon with a high degree of crystallinity. The peak at 22 = 35◦ is probably due to SiO (see discussion below), and the strong increase in intensity for scattering angles smaller than 22 = 27◦ is ascribed to the filter paper which supported the fine powder. This has been verified by analyzing the paper alone. The contribution of the filter paper to the XRD spectrum was not known before this study. Since the signal was very strong and the spectra were not recorded digitally we did not try to subtract the filter contribution. When acetylene is introduced into the flow reactor and when its flow rate is steadily increased, the diffraction peaks of silicon continually decrease and
ASTM 42-1091
22 ( )
d (Å)
I /Imax (%)
22 (◦ )
I /Imax (%)
35.5 42.0 47.0 60.0 71.5
2.52 2.15 1.95 1.54 1.32
100 11 10 21 14
35.8 41.3 — 59.8 71.8
100 20 — 63 50
◦
new peaks appear at 22 = 35.5◦ , 60◦ , and 71.5◦ . As is shown in Table 4, where the literature values [28] for silicon carbide are given, these lines can unambiguously be attributed to crystalline SiC in its β-phase. Using the stoichiometric mixture (SiC3), we still observe crystalline silicon, and only if acetylene is applied in excess (SiC4) the major silicon peak at 22 = 28.8◦ disappears. This observation is in perfect agreement with the results of the FTIR studies discussed before. The SiC0 spectrum shown in Figure 6 features a diffraction peak at 22 = 35◦ that cannot be attributed to SiC since no carbon was introduced into the flow reactor. In addition, it is clearly seen that its position is slightly below that of SiC at 22 = 35.5◦ . The origin of this peak is not completely clear but we believe that it could be due to silicon oxide. Two arguments are in favor of this assignment. First, the FTIR spectrum of pure silicon (SiC0) shows a strong absorption band due to silicon oxide, indicating that silicon oxide is present even though it will mostly be amorphous. Second, the XRD literature reports a form of silicon oxide having its three most intense diffraction lines in the right order at d = 4.11 Å (22 = 21.6◦ ), d = 2.52 Å (22 = 35.6◦ ), and d = 1.64 Å (22 = 56.0◦ ) [29] and d = 4.15 Å (22 = 21.4◦ ), d = 2.53 Å (22 = 35.4◦ ) and d = 1.64 Å (22 = 56.0◦ ) [30]. While the first line appears in the region of the strong diffraction peaks of
300
Figure 7. Relative intensities of the main Si and SiC diffraction peaks as a function of the C2 H2 flow rate after subtraction of the SiO contribution. The data points are connected by straight lines.
the filter and is therefore not seen, the third (and much weaker) line may be hidden under the Si(311) diffraction peak (see Figure 6). In Figure 7 we have plotted the dependencies of the relative intensities of the major Si and SiC peaks (at 22 = 28.8◦ and 22 = 35.5◦ , respectively) on the flow rate of the carbon donor acetylene. For the SiC curve the contribution of SiO has been subtracted, assuming that it is always 20% of the main Si diffraction peak. The figure nicely visualizes that the content of crystalline silicon continually decreases while the portion of silicon carbide increases almost linearly. It is also seen that, for a stoichiometric gas mixture (SiC3; 12.5 sccm C2 H2 ), the SiC character dominates but that there is still a significant contribution from silicon. 4. Discussion An important aspect of the present study was the investigation of the nanocluster beam produced by early extraction of the laser-induced reaction products from the flow reactor. Unfortunately, we could not apply all analysis tools for the structural characterization of the beam particles; but we obtained important information from the analysis of the powder collected during the same experiments. Another aim of our investigations was to continuously follow the formation of silicon carbide particles, starting from pure silicon and introducing more and more carbon. The mass spectrometric characterization of the beam particles revealed average diameters between 3 and
3.5 nm with typical FWHMs of their size distributions around 1.3 nm. It should be emphasized that the beam particles are considerably smaller than the particles collected in the flow reactor. Though not measured in the present study, this information is deduced from the results of other investigators who found typical mean diameters above 20 nm for SiC particles produced by laser pyrolysis under similar conditions [2,6,7]. In our case, this difference is explained with the early extraction of the beam particles from the reaction zone which prevents the particles from further growth, in strict contrast to the situation in the flow reactor. The same observation has been made in our detailed studies on silicon nanoparticles [11,13,16]. Since the nanoclusters are extracted through a 200-µm nozzle, the particle flux in the beam is considerably smaller than the flux in the flow reactor. However, our experiments on silicon nanoclusters [13] have shown that, with deposition rates around 7 nm cm−2 min−1 , our cluster source is superior to most laser vaporization sources (0.1–6 nm cm−2 min−1 [31]). Even higher fluxes could be obtained by using a continuous-wave CO2 laser (instead of the pulsed laser) or a larger nozzle (which, however, would require larger pump facilities for the source chamber). Let us now discuss the IR spectra which have been obtained for both species, beam particles and powder. As is seen in Figure 4, the spectra obtained for pure silicon are very similar. Both spectra feature a strong absorption band due to the asymmetric Si–O–Si vibration [20]. It should be mentioned that the overall shape of the absorption band is difficult to compare with existing literature data, obviously due to different methods of sample preparation, particle size, degree of crystallinity, and surface passivation. The high frequency shoulder appears much more pronounced in the powder spectrum than in the beam particle spectrum. In addition, it seems somewhat shifted to higher frequencies. This observation could indicate a possible nonstoichiometry in the oxide shell of the powder, in the sense of a higher content of oxygen [22] although, at first glance, this seems in contradiction with the higher specific surface area of the (smaller) beam particles. However, we have to remember that the smaller silicon nanocrystals encountered in the beam have a thinner oxid shell [14] and that the powder may contain a higher portion of amorphous silicon which can be oxidized more easily. Following Refs. [21,22], we assigned in Section 3.1.2 the sharp peak at 880 cm−1 to a Si–O–Si
301 vibration. In contrast, Lowe-Webb et al. attributed this peak to the Si–H bending mode in the (O3 )Si–H group [20]. Moreover, these authors observed a correlation between the height of the 880 cm−1 peak and the photoluminescence (PL) properties of the material. In all samples exhibiting PL, the peak at 880 cm−1 was rather pronounced (but always a factor of 4 weaker than the main Si–O feature at 1080 cm−1 ) [20]. This correlation seems to be in agreement with our observation. While the beam-deposited silicon nanoparticles exhibit strong visible photoluminescence, the nanopowder PL is very weak (see discussion below). In agreement with this finding and the results of Lowe-Webb et al. [20], the absorption at 880 cm−1 is clearly more pronounced in the IR spectrum of the beam-deposited sample. The two FTIR spectra of samples SiC3x both show two strong absorption bands due to Si–O (1080 cm−1 ) and Si–C (850 cm−1 ), proving that silicon carbide is present in both samples. The good agreement between the two spectra indicates that the composition of the nanoclusters collected in the beam is almost identical to that of the powder particles, specifically regarding the Si/SiC ratio. Now we want to discuss the evolution of the SiC peak in the FTIR spectra of the powder samples (SiC0 – SiC4) shown in Figure 5. For SiC1 the SiC band is very weak and asymmetric owing to the two silicon features at 800 and 880 cm−1 which are superimposed. These features are still noticeable in spectrum SiC2 but seem to disappear for sample SiC3. The most prominent SiC feature appears for sample SiC4 as a rather narrow band peaking at 855 cm−1 . The shape of the band and the position of its maximum could indicate a high degree of crystallinity and a lack of Si–H bonds [32], which is confirmed by the absence of the Si–H features at 2100 and 2245 cm−1 . A closer look to the shape of the Si–C band reveals a slight asymmetry suggesting a shoulder on the high frequency side. This observation is in line with the results of Cauchetier et al. [3] who studied the formation of SiC powders as a function of the kind of hydrocarbon added to SiH4 . For C2 H2 , these authors found a double-peak structure with an additional peak at 950 cm−1 . Going over C2 H4 to C2 H6 , this secondary peak degenerated to a shoulder. The double-peak structure was related to the higher temperature measured for the SiH4 /C2 H2 mixture. The close resemblance of our SiC band with the one observed by Cauchetier et al. for the SiH4 /C2 H6 mixture suggests a lower temperature in our pulsed pyrolysis experiment, namely 1400◦ C, compared to 1800◦ C in the cw laser experiment [3].
Our FTIR and XRD studies show consistently that C2 H2 must be supplied in excess to consume all Si for SiC particle production. A stoichiometric SiH4 /C2 H2 mixture, as used for the preparation of sample SiC3, results in a powder that contains, besides SiC, also pure silicon nanoparticles or silicon carbide particles with domains of crystalline silicon. This is evidenced by the XRD analysis. A simple explanation could be based on an incomplete decomposition of the C2 H2 molecules, which would be plausible since acetylene itself does not absorb the CO2 laser radiation. Another explanation is based on a possible mechanism for SiC particle formation that we would like to discuss now. Rao et al. [33] studied the properties of SiC nanoparticles produced by dissociating a mixture of SiCl4 /CH4 in a dc arc plasma and quenching the hot gases in a subsonic expansion. They always found some free silicon in their SiC powders except at very high levels of excess methane. To explain their observation, Rao et al. proposed a mechanism in which silicon particle nucleation is the first step in the synthesis of silicon carbide nanoparticles. SiC is then formed by heterogeneous chemical reaction on the surface of the silicon core, followed by a fairly rapid diffusion of carbon into the volume of the particle. If only short residence times are available for the carburization process, a considerable amount of excess hydrocarbon may be needed to obtain stoichiometric silicon carbide. This mechanism, proposed by Rao et al. [33], would satisfactorily explain our observation of crystalline silicon in the powder produced under stoichiometric conditions. However, from the close resemblance of the nanocluster and powder spectra displayed in Figure 4 and the fact that the nanoclusters had much shorter residence time in the flow reactor than the powder particles, we conclude that, in our experiment, the major reason for an incomplete carburization was a deficit of free carbon. Assuming an average flow velocity of 400 cm s−1 [34], the residence time of the nanoclusters before extraction is estimated to be ∼0.5 ms whereas the residence time of the powder particles is at least a factor of 10 longer. If acetylene is supplied in excess one could imagine that the synthesized particles contain domains of amorphous carbon. Unfortunately, it is not possible to identify amorphous carbon with the present techniques of analysis. But, if carbon were present, we would expect to observe C–H vibrations in the IR spectrum around 2900 cm−1 , as we did in our study on carbon nanoparticles produced by laser pyrolysis of photosensitized C2 H2 [35]. The absence of these vibrational
302 bands indicates that the content of amorphous carbon cannot be very high in our sample SiC4. However, to definitely answer this question, further experiments are necessary which allow us to determine the complete stoichiometry of the powders, for example by Rutherford back scattering. In this context it is also interesting to mention the very recent study of M´elinon et al. [36] who investigated thin films of SiC clusters prepared by laser vaporization of polycrystalline SiC in combination with low-energy cluster beam deposition. They found indication of a partial phase separation in single nanoclusters, giving rise to silicon- and carbonrich local phases. During the course of the preparation of this publication we have reexamined the FTIR spectra of the beam-deposited samples SiC0x and SiC3x. This reinvestigation was carried out seven months after the first characterization, which took place four weeks after the sample preparation and the results of which are shown in Figure 4. Whereas the IR spectrum of the beamdeposited sample SiC0x is practically unchanged, sample SiC3x shows significant differences in that the Si–O peak at 1080 cm−1 is now 40% higher while the SiC band has remained constant. This observation suggests that the oxidation of SiC nanoparticles is slower than for pure silicon particles. Finally we would like to make a few comments on the photoluminescence behavior of the samples characterized in the present study by mass spectrometry, FTIR, and XRD. For this purpose, we had deposited the Si and SiC nanoclusters in the beam (SiC0, SiC1, SiC2, and SiC3) on a sapphire window. For the pure silicon sample (SiC0) we observed strong PL in the red, clearly visible with the naked eye [18]. In contrast, no PL could be discerned for the as-prepared samples SiC1, SiC2, and SiC3, at least not with the naked eye. The fine powders, collected in the flow reactor, did not show any PL, as well. The latter observation can be easily understood in the frame of the quantum confinement model which gives the result that silicon particles must be smaller than 5 nm in diameter to exhibit visible luminescence [16,37,38]. Further studies are underway to quantitatively determine the PL yields of the beamdeposited Si/SiC samples. 5. Summary CO2 -laser-induced decomposition of SiH4 /C2 H2 in a flow reactor, with varying concentration of the hydrocarbon, has been employed to produce silicon and
silicon carbide nanoparticles. Using a conical nozzle, the nascent nanoclusters were extracted from the reaction zone into a high-vacuum molecular beam apparatus where they were analyzed with a time-offlight mass spectrometer. The nanoclusters, having diameters between 3 and 3.5 nm were deposited at low energy on various substrates. During the same experiment, significant larger but still nanosized particles were collected in the exhaust line of the flow reactor. The samples were analyzed by FTIR and XRD. Both methods consistently reveal that, with increasing flow rate of acetylene, the concentration of SiC in the nanoparticles also increases. A stoichiometric SiH4 /C2 H2 mixture yields SiC nanoparticles still containing a substantial portion of free crystalline silicon. Pure silicon carbide nanoparticles are only synthesized if the hydrocarbon is supplied in excess. Although the nanoclusters have significantly reduced residence time in the reaction volume they are found to have equal SiC content as the powder particles. This finding suggests that the major reason of incomplete carburization is a deficit of free carbon in the reaction zone. The XRD studies revealed that the powder particles are composed of crystalline SiC in its β-phase. Due to the close resemblance of the corresponding FTIR spectra, we are tempted to ascribe the same property to the nanoclusters in the beam.
Acknowledgements This work was supported by the Deutsche Forschungsgemeinschaft in the frame of its Schwerpunktprogramm Fine Solid Particles and by bilateratal cooperations between FRG and Romania (WTZ RUM008-97) and FRG and France (PROCOPE).
References 1. Matsumoto T., Takahashi J., Tamaki T., Futagi T., Mimura H. and Kanemitsu Y., 1994. Appl. Phys. Lett. 64, 226. 2. Cannon W. R., Danforth S.C., Haggerty J.S. and Marra R.A., 1986. J. Am. Ceramic Soc. 65, 330. 3. Cauchetier M., Croix O. and Luce M., 1988. Adv. Ceram. Mater. 3, 548. 4. Borsella E., Caneve L., Fantoni R., Piccirillo S., Basili N. and Enzo S., 1989. Appl. Surf. Sci. 36, 213. 5. Suzuki M., Nakata Y., Okutani T. and Kato A., 1992. J. Mater. Sci. 27, 6091. 6. Suzuki M., Maniette Y., Nakata Y. and Okutani T., 1993. Ceramics Intern. 19, 407.
303 7. Tougne P., Hommel H., Legrand A.P., Herlin N., Luce M. and Cauchetier M., 1993. Diamond Relat. Mater. 2, 486. 8. Ehbrecht M., Ferkel H., Smirnov V.V., Stelmakh O.M., Zhang W. and Huisken F., 1995. Rev. Sci. Instrum. 66, 3833. 9. Ehbrecht M., Faerber M., Rohmund F., Smirnov V.V., Stelmakh O.M. and Huisken F., 1993. Chem. Phys. Lett. 214, 34. 10. Ehbrecht M., Ferkel H., Smirnov V.V., Stelmakh O.M., Zhang W. and Huisken F., 1996. Surf. Rev. Lett. 3, 807. 11. Ehbrecht M. and Huisken F., 1999. Phys. Rev. B 59, 2975. 12. Huisken F., Kohn B., Alexandrescu R. and Morjan I., 1999. Mass spectrometric characterization of iron clusters produced by laser pyrolysis and photolysis of Fe(CO)5 in a flow reactor. Europ. Phys. J. D (in press). 13. Ehbrecht M., Ferkel H. and Huisken F., 1997. Z. Phys. D 40, 88. 14. Hofmeister H., Huisken F. and Kohn B., 1999. Lattice contraction of nanosized silicon particles produced by laser pyrolysis of silane. Europ. Phys. J. D (in press). 15. Ehbrecht M., Ferkel H., Huisken F., Holz L., Polivanov Y. N., Smirnov V. V., Stelmakh O. M. and Schmidt R., 1995. J. Appl. Phys. 78, 5302. 16. Ehbrecht M., Kohn B., Huisken F., Laguna M. A. and Paillard V., 1997. Phys. Rev. B 56, 6958. 17. Ledoux G., Ehbrecht M., Guillois O., Huisken F., Kohn B., Laguna M.A., Nenner I., Paillard V., Papoular R., Porterat D. and Reynaud C., 1998. Astron. Astrophys. 333, L39. 18. Huisken F. and Kohn B., 1999. Structured films of lightemitting silicon nanoparticles produced by cluster beam deposition. Appl. Phys. Lett. (in press). 19. Lidel D.R. (ed.), 1991, CRC Handbook of Chemistry and Physics. CRC Press, Boca Raton. 20. Lowe-Webb R.R., Lee H., Ewing J.B., Collins S.R., Yang W., and Sercel P.C., 1998. J. Appl. Phys. 83, 2815. 21. Hayashi S., Kawata S., Kim H.M. and Yamamoto K., 1993. Jpn. J. Appl. Phys. 32, 4870. 22. R¨ubel H., Schr¨oder B., Fuhs W., Krauskopf J., Rupp T. and Bethge K., 1987. Physica Status Solidi B 139, 131.
23. Zacharias M., Dimova-Malinovska D. and Stutzmann M., 1996. Philosophical Magazine B 73, 799. 24. Xie Y.H., Wilson W.L., Ross F.M., Mucha J.A., Fitzgerald E.A., Macaulay J.M. and Harris T.D., 1992. J. Appl. Phys. 71, 2403. 25. Tischler M.A., Collins R.T., Stathis J.H. and Tsang J.C., 1992. Appl. Phys. Lett. 60, 639. 26. Lucovsky G. and Pollard W.B., 1983. J. Vac. Sci. Technol. A 1, 313. 27. ASTM 27-1402. 28. ASTM 42-1091. 29. ASTM 27-0605. 30. ASTM 04-0359. 31. M´elinon P., Paillard V., Dupuis V., Perez A., Jensen P., Hoareau A., Perez J.P., Tuallion J., Broyer M., Vialle J.L., Pellarin M., Baguenard B. and Lerme J., 1995. Int. J. Mod. Phys. B 9, 339. 32. De Cesare G., La Monica S., Maiello G., Masini G., Proverbio E., Ferrari A., Chitica N., Dinescu M., Alexandrescu R., Morjan I. and Rotiu E., 1996. Appl. Surface Sci. 106, 193. 33. Rao N., Micheel B., Hansen D., Fandrey C., Bench M., Girshick S., Heberlein J. and McMurry P., 1995. J. Mater. Res. 10, 2073. 34. Cannon W.R., Danforth S.C., Flint J.H., Haggerty J.S. and Marra R.A., 1986. J. Am. Ceramic Soc. 65, 324. 35. Schnaiter M., Henning T., Mutschke H., Kohn B., Ehbrecht M. and Huisken F., 1999. Infrared spectroscopy of nanosized carbon grains produced by laser pyrolysis of acetylene — Analogue materials for interstellar grains. Astrophys. J. (in press). 36. M´elinon P., K´egh´elian P., Perez A., Ray C., Lerm´e J., Pellarin M., Broyer M., Boudelle M., Champagnon B. and Rousset J. L., 1998. Phys. Rev. B 58, 16481. 37. Delerue C., Allan G. and Lannoo M., 1993. Phys. Rev. B 48, 11024. 38. Hill N.A. and Whaley K.B., 1996. J. Electron. Mat. 25, 269.
