Microstructure evolution of precursors-derived SiCN ceramics upon thermal treatment between 1000 and 1400 °C

Microstructure evolution of precursors-derived SiCN ceramics upon thermal treatment between 1000 and 1400 °C

Journal of Non-Crystalline Solids 351 (2005) 1393–1402 www.elsevier.com/locate/jnoncrysol Microstructure evolution of precursors-derived SiCN ceramic...

439KB Sizes 0 Downloads 32 Views

Journal of Non-Crystalline Solids 351 (2005) 1393–1402 www.elsevier.com/locate/jnoncrysol

Microstructure evolution of precursors-derived SiCN ceramics upon thermal treatment between 1000 and 1400 C G. Gregori a

a,*

, H.-J. Kleebe b, H. Brequel c, S. Enzo c, G. Ziegler

a

University of Bayreuth, Institute for Materials Research (IMA-I), Ludwig-Thoma Strasse 36b, 95447 Bayreuth, Germany b Colorado School of Mines, Metallurgical and Materials Engineering Department, Golden, CO 80401, USA c University of Sassari, Chemistry Department, Via Vienna 2, 07100 Sassari, Italy Received 4 November 2004; received in revised form 17 February 2005

Abstract The microstructure evolution of different precursors-derived SiCN ceramics upon exposure at temperatures ranging between 1000 and 1400 C has been investigated by means of transmission electron microscopy (TEM), electron energy-loss spectroscopy (EELS), X-ray energy dispersive spectroscopy (XEDS), Raman spectroscopy and Neutron scattering. The results of the present study indicate that these materials remain structurally amorphous and can, therefore, be considered stable against devitrification within this temperature range. However, the amorphous microstructure of these ceramics does undergo a short-range rearrangement. It is important to note that SiCN precursors with very similar chemical composition but different starting precursor architecture result in very similar amorphous network structures after exposure to 1400 C. The presence of few localized crystalline regions, composed of Si3N4 crystals or turbostratic carbon features is explained in terms of gas-phase reactions occurring at residual closed pores between 1000 and 1400 C.  2005 Elsevier B.V. All rights reserved.

1. Introduction Due to their high thermal resistance and chemical stability [1–6], SiCN polymer-derived ceramics (PDCs) present a unique set of properties that make them promising candidates for high-temperature applications. These materials, first processed by Verbeek et al. [7] and Yaijma et al. [8] in the early seventies, are typically synthesized via pyrolysis of organo-silicon compounds at temperatures ranging between 800 and 1000 C yielding the formation of amorphous silicon-based ceramics.

* Corresponding author. Current Address: California NanoSystems Institute, University of California, Santa Barbara, CA 93106-6105, United States. E-mail address: [email protected] (G. Gregori).

0022-3093/$ - see front matter  2005 Elsevier B.V. All rights reserved. doi:10.1016/j.jnoncrysol.2005.03.025

PDCs have gained great interest because of their favorable properties and their innovative processing technology. Compared to more traditional processing methods of ceramic materials, the organic-to-inorganic transition offers the possibility to control the chemical composition and improve specific properties of the final ceramic solids through the selection of properly designed organic precursors. Note, for example, that the use of boron-containing SiCN-precursors enabled the synthesis of PDCs with excellent thermal stability up to 2000 C [9]. In addition, the ability to prepare near net-shaped materials using processing methods typical of the polymer technology as well as the low processing temperatures (800–1000 C) represent two other relevant advantages in comparison to traditional powder processing of high-performance ceramics. These aspects

1394

G. Gregori et al. / Journal of Non-Crystalline Solids 351 (2005) 1393–1402

have made PDCs extremely attractive for a wide range of applications such as powders, fibers, coatings, binders and matrices for ceramic–matrix composites and more recently MEMS applications [5,10–16]. After pyrolysis at 1000 C, the microstructure of these amorphous materials can be generally described as a random Si–C–N network composed of tetrahedral mixed units, [SiCxN(4x)] (1 6 x 6 4), whose relative volume fraction depends on the composition of the starting precursor. Note that if the carbon fraction present in the starting precursor is high enough, then, upon pyrolysis, part of it remain excluded from the Ôsilicon backboneÕ constituting the amorphous network and creating the so-called free carbon phase. Further thermal treatments at temperatures above 1000 C induce a bondings redistribution process yielding the rearrangement of the amorphous network [17–22]. Annealing environment and chemical composition as well as type of side groups connected to the silicon atoms present in the precursors affect this process. The microstructure of these metastable SiCN glasses undergoes major modifications above 1440 C yielding the formation of SiC and Si3N4 crystals as well as graphite-like domains (turbostratic carbon), as described previously [23–27]. The motivation of the present contribution is the lack of information regarding the relationship between microstructural evolution (network rearrangement) below 1440 C and the early stages of crystallization. In this study SiCN ceramics heat-treated at temperatures ranging between 1000 and 1400 C are investigated by means of complementary experimental techniques, i.e., neutron scattering (NS), Raman spectroscopy, transmission electron microscopy (TEM) in conjunction with Xray energy dispersive spectroscopy (XEDS) and electron energy-loss spectroscopy (EELS). Three SiCN precursors namely HVNG, HPS and ABSE were employed in this study. Note that HPS and ABSE having similar composition but different precursor-architecture (Fig. 1, Table 1) were chosen in order to study the effect of the precursor architecture on the microstructure evolution of the metastable glass phase.

2. Experimental SiCN samples were prepared starting from three different silazanes, i.e., HPS, HVNG and ABSE (Fig. 1). HPS was obtained by coammonolysis of dichloromethylsilane and dichloromethylvinylsilane, while the preparation of HVNG was achieved by coammonolysis of dichloromethylsilane and trichlorovinylsilane, as described in Ref. [2]. ABSE was prepared by hydrosilylation and ammonolysis of the same chlorosilanes used for HPS, as illustrated in Ref. [12]. Importantly, although the HPS and ABSE precursors were prepared

starting from the same chemical compounds, the different processing conditions yield the formation of two precursors having the same basic units, which are however differently arranged. Since all these precursors are sensitive to the effect of oxygen and moisture, the entire processing was performed in controlled atmosphere (N2). The liquid HPS and HVNG precursors were subsequently crosslinked at 300 C (heating and cooling rate: 1 C/min) for 3 h in N2-atmosphere by addition of diculmylperoxide as radical initiator. Conversely, since in ABSE a large fraction of silicon atoms is bridged via ethylene groups [12], this precursor was heat-treated at 500 C for 3 h in N2-atmosphere, so to complete the crosslinking process. The pre-ceramics were ball-milled in a teflon container using ZrO2 milling balls and then sieved (<32 lm). The powders were then mixed with the corresponding starting precursors in a 7:3 volume ratio and uniaxially pressed at 10 MPa at temperatures ranging between 130 and 180 C. The soobtained bars (70 · 6 · 4 mm) were pyrolyzed at 1000 C for 1 h in N2-atmosphere. The chemical composition of these compounds upon pyrolysis is given in Table 1. Note that due to the composition of the starting polymers, the resulting amorphous ceramics typically contain a certain amount of carbon not directly bonded to silicon, described as free carbon phase. Part of the specimens underwent a second heat-treatment at 1400 C for 6 h in N2-atmosphere. Neutron scattering (NS) measurements were performed on SiCN samples at ISIS, Rutherford Appleton Laboratory (UK), using the GEM time-of-flight (TOF) diffractometer. Two sets of SiCN bars were investigated in order to detect amorphous network changes occurring between pyrolysis at 1000 C and thermal treatment for 6 h at 1400 C. The experimental results were elaborated as described in Ref. [28] in order to obtain the total pair correlation functions (PCFs). Raman spectroscopy was employed with the aim to investigate the structural rearrangement of the free carbon phase present in the amorphous bulk. The measurements were performed on cross-sectional surfaces of SiCN bulk samples using a System 1000 (Renishaw plc) microprobe employing an Ar-ion laser (514.5 nm). TEM in conjunction with XEDS and EELS analysis required a specific sample preparation procedure. TEM-foils obtained starting from the ceramic samples, was achieved following the standard routine: cutting, ultrasonic drilling, dimpling and Ar-ion thinning to perforation followed by light carbon coating to minimize charging under the electron beam. The samples were investigated using a Philips CM20 field emission gun operating at an acceleration voltage of 200 kV, equipped with a PEELS spectrometer, model 666 (Gatan Inc.), for EELS analysis. All spectra, collected from areas having a diameter of approximately 150 nm, were acquired using the 2 mm PEELS entrance aperture and with an

G. Gregori et al. / Journal of Non-Crystalline Solids 351 (2005) 1393–1402

1395

Fig. 1. Basic structural units of the different SiCN-precursors employed here: (a) HVNG, (b) HPS and (c) ABSE.

Table 1 Chemical composition of the different SiCN compounds after pyrolysis (1000 C). The oxygen content is below 1 wt% Composition (wt%)

HPS HVNG ABSE

Si

C

N

51.4 50.3 51.8

26.5 20.6 27.7

21.9 26.9 18.3

energy dispersion of 0.1 eV/channel. The energy resolution, defined as the full-width half-maximum (FWHM) of the zero-loss peak, was ca. 0.8 eV.

3. Results The experimental data of the NS measurements are shown in Fig. 2. After pyrolysis at 1000 C the paircorrelation functions (PCFs), corresponding to the different compositions, are characterized by the presence of four major peaks located at ca. 1.42, 1.75, 2.48 and ˚ . The position of all peaks in addition to charac2.86 A teristic atomic distances of reference materials are sum˚ is marized in Table 2 [22,29–31]. The first peak at 1.42 A attributed to the nearest-neighbour distance in graphite, ˚ , is assigned to Si–N while the second centered at 1.75 A ˚ is bonds in amorphous Si3N4. The peak at 2.48 A

Stoichiometry

C/Si (at.%)

N/Si (at.%)

SiC0.36N0.85 + C0.85 SiC0.18N1.08 + C0.78 SiC0.45N0.71 + C0.80

1.21 0.96 1.25

0.85 1.08 0.71

referred to C–(C)–C bond distances in graphite, whereas ˚ is most likely a combination of the the fourth at 2.86 A ˚ distance in graphite and the pair distances pres2.84 A ˚ . A closer ent in Si3N4, ranging between 2.80 and 2.94 A analysis of the total pair correlation functions allows the ˚ (see the arrows recognition of a weak signal at ca. 3.1 A in Fig. 2(a)). According to the data listed in Table 2, this may be due to the Si–Si distance in SiC or Si3N4 as well. The heat-treatment at 1400 C induces a network rearrangement in all three samples, as indicated by the PCFs shown in Fig. 2(b). The three systems share a similar evolution given by the appearance of new weak peaks ˚ ) as well as by the increase in the mid-range (2.4–2.9 A of intensity of the already existing peaks (in particular of carbon) in the short-range region. In general, peaks ˚ can be assigned to interlocated at 3.74, 4.27 and 5.05 A atomic distances of graphitic carbon. However, a broad ˚ is also characteristic peak at approximately 4.30–4.34 A

1396

G. Gregori et al. / Journal of Non-Crystalline Solids 351 (2005) 1393–1402

Intensity of the total correlation function [a.u.]

(a) 1000 °C

HVNG

HVNG

HPS

HPS

ABSE

ABSE

1

2

SiCN specimens upon exposure at 1540 C in N2atmosphere. In both ABSE and HVNG spectra, the position of the Si-L3 peak centered at ca. 105 eV is characteristic of Si–N bonding. However, the relative low value of the peak onset (ca. 99.5 eV) indicates that the contribution of more covalent bonding. i.e., Si–C, is also present within the Si-based network. The heat-treatment at 1400 C does not induce significant modifications on the near-edge structure of the Si-L edge. The nitrogen K-edge consists of two peaks at ca. 405 and 420 eV. Their position and their relative intensity do not change with the different starting compositions and appear to be not affected by the second heat-treatment, as shown in Fig. 3(c) and (d). The comparison between SiCN and reference data evidences that the local environment around N-atoms in ABSE and HVNG-derived ceramics is essentially the same of N-atoms in Si3N4. EELS investigations performed on N-doped carbon compounds [32–34] showed that C–N bonds are characterized by p* and r* peaks located at 399 and 408 eV, respectively. None of these characteristic features can be recognized in the present spectra. Therefore, consistently with NS data, it can be concluded that N-atoms are exclusively bonded to silicon and no C–N bondings are present in the Si–C–N network. The near-edge structures of carbon K-edge spectra are depicted in Fig. 3(e) and (f). The data corresponding to samples annealed at 1000 C show the presence of a shoulder at ca. 285 eV, which is the typical ÔfingerprintÕ of p* bonding due to the presence of sp2-hybridized carbon. The broad peak located between 290 and 300 eV is characteristic for amorphous carbon. Upon exposure at 1400 C, the shoulder at 285 eV becomes a well-shaped peak and the broad region between 290 and 300 eV is characterized by a relative sharp peak centered at ca. 292 eV. This arises from r* C-bonding due to the presence of sp3-hybridized carbon. The reference spectra collected from graphite and SiC indicate that, within this

(b) 1400 °C

3 6 4 5 Radial distance [Å]

7

1

2

3 6 4 5 Radial distance [Å]

7

Fig. 2. Pair correlation functions of HVNG, HPS and ABSE samples (a) pyrolyzed at 1000 C and (b) upon heat-treatment at 1400 C.

of the medium-range order of amorphous Si3N4, as reported by Dixmier et al. [29]. The correlation function corresponding to the HVNG composition shows a rather intense background noise, which makes the iden˚ in the second order tification of peaks at 3.73 and 4.27 A range difficult. It is to note that the correlation function of HVNG is clearly different compared to materials derived from the HPS and ABSE precursors. The nature of the chemical bonding present within the amorphous networks was investigated by electron energy-loss spectroscopy (EELS). For this purpose, the near-edge fine structure of energy-loss spectra corresponding to the three atomic species present in these compounds was examined. The near-edge fine structures of representative Si-L edge spectra acquired from ABSE and HVNG systems in addition to spectra of reference materials are shown in Fig. 3(a) and (b). These latter data were recorded from SiC and Si3N4 whiskers grown on the surface of

Table 2 ˚ ] of the pair correlation functions obtained from the different SiCN compounds in addition to typical atomic distances in related Peak positions [A reference materials; the specific atomic pairs are given in parenthesis Graphite

SiC

Si3N4

1000 C HVNG

HPS

ABSE

HVNG

HPS

ABSE

1.71–1.79 (Si–N)

1.42 1.74

1.41 1.75

1.42 1.75

1.43 1.74

1.43 1.74

1.42 1.74

2.47

2.49

2.46

2.44

2.46

2.45

2.86

2.85

2.86

2.85

2.84

2.87

3.73 4.27

3.75 4.27

3.73 4.26

1.42

1400 C

1.88 (Si–C) 2.46 2.58 2.72–3.15 (Si–Si) 2.80–2.94 (N–N)

2.84 3.06 (Si–Si) 3.75 4.26 4.91 5.11

5.04

G. Gregori et al. / Journal of Non-Crystalline Solids 351 (2005) 1393–1402

(b) SiC

SiC

Si3N4

Si3N4

ABSE 1400 °C

HVNG 1400 °C

ABSE 1000 °C

HVNG 1000 °C

100 105 110 115 120 125 130 135

100 105 110 115 120 125 130 135

Energy Loss [eV]

Energy Loss [eV]

(d)

(c)

Intensity [a.u.]

Si3N4

390

400

410

Si3N4

ABSE 1400 °C

HVNG 1400 °C

ABSE 1000 °C

HVNG 1000 °C

420

430

440

390

Energy Loss [eV]

(e)

400

Intensity [a.u.]

420

430

440

(f)

SiC

SiC

σ* π*

410

Energy Loss [eV]

at 285 and 292 eV, respectively. Between 1000 and 1400 C, the HVNG system exhibits an intensity ratio increment of 26%, while ABSE shows an increase of 19%. EELS data collected from ABSE and HVNG do not show remarkable differences, in particular, as far as the Si-L edge is considered. The main reason for these results is correlated to the size of the areas irradiated (150 nm diameter) during the EELS measurements. ˚ -scale are The subtle differences detected by NS at A averaged over this large area monitored during EELS analysis. The evolution of the free carbon phase was investigated also by Raman spectroscopy. The results acquired from ABSE, HPS and HVNG pyrolyzed at 1000 C and then annealed at 1400 C are given in Fig. 4. The spectra are shown after the subtraction of the linear background, which was interpreted in terms of recombination of electron–hole pairs due to the presence of dangling bonds within the amorphous Si–C–N network. All spectra show the typical features of disordered graphite-like carbon, namely the D-peak located at ca. 1350 cm1, and the G-peak centered between 1550 and 1600 cm1 [31,35–39]. The Raman spectra were fitted using two Gaussian curves so to extract the intensity values of the D- and G-modes. The corresponding results are summarized in Table 3. Importantly, their relative distance increases upon exposure at 1400 C, indicating that an ordering process occurs within the free carbon phase during annealing. In addition, the intensity ratio I(D)/I(G) is used to evaluate the lateral size, Ls, of the free carbon domains, as reported by Tuinstra and Koenig [35] for crystalline carbon and more recently by Ferrari and Robertson [36] in the case of amorphous carbon. Since

σ* Graphite

ABSE 1400 °C

ABSE 1000 °C 280 290 300 310 320 330 340

Energy Loss [eV]

π*

Graphite

(a) 1000 °C

(b) 1400 °C

HVNG 1400 °C

HVNG 1000 °C 280 290 300 310 320 330 340

Energy Loss [eV]

Fig. 3. (a) Silicon L-edge spectra of ABSE and HVNG samples pyrolyzed at 1000 and (b) subsequently annealed at 1400 C. (c) Nitrogen K-edge after pyrolysis and (d) heat-treatment at 1400 C. (e) Carbon K-edge after 1000 C and (f) upon thermal treatment. In addition, spectra acquired from SiC and Si3N4 and graphite are shown for comparison.

temperature range, the free carbon present in both system undergoes a graphitization process. Here, the evolution of the sp2/sp3 bonding ratios is estimated considering the p*/r* intensity ratios, which are extracted using energy windows of 3 and 10 eV centered

Raman Intensity [a.u]

Intensity [a.u.]

(a)

1397

HVNG

HVNG

HPS

HPS D

D

G

ABSE

G

ABSE

1200

1400

1600

Raman shift [cm-1]

1800

1200

1400

1600

1800

Raman shift [cm-1]

Fig. 4. Raman spectra of HPS, HVNG and ABSE compositions upon exposure at (a) 1000 C and (b) 1400 C. The gray dashed lines highlight the position of the D- and G-bands.

1398

G. Gregori et al. / Journal of Non-Crystalline Solids 351 (2005) 1393–1402

Table 3 Peak position of D- and G-Raman mode after pyrolysis and upon annealing at 1400 C 1000 C 1

HPS HVNG ABSE

1400 C 1

D (cm )

G (cm )

I(D)/I(G)

D (cm1)

G (cm1)

I(D)/I(G)

1344 1362 1359

1582 1551 1553

1.28 1.86 2.24

1350 1356 1349

1600 1601 1599

1.88 2.00 2.07

in SiCN ceramics the free carbon phase is amorphous upon pyrolysis at 1000 C, the equation proposed by Ferrari and Robertson IðDÞ ¼ CðkÞL2S IðGÞ

ð1Þ

was employed in the present work. The value of C(k) depends on the laser wavelength (514.5 nm) and here ˚ 2. These data are also precorresponds to 0.0055 A sented in Table 3. After pyrolysis at 1000 C, the intensity ratio of HPS is lower than the ratio of HVNG and ABSE, indicating that HPS contains the smallest free ˚ ). Upon annealing at carbon domains (Ls = 15 A 1400 C, the intensity ratio of HPS clearly increases, while it remains nearly constant for the ABSE and

HVNG compositions. According to these data, the final lateral size of the free carbon clusters is nearly the same for all three systems. In particular, the lateral size calcu˚ , while it remains lated for HPS increases from 15 to 18 A ˚ unchanged (19 A) for ABSE and HVNG. TEM analysis performed on samples pyrolyzed at 1000 C revealed the presence of residual closed porosity homogeneously dispersed within the amorphous bulk of all three compositions. A closer observation of these features showed that these pores often contain basic structural units (BSUs) [40] of graphite-like carbon. Upon annealing at 1400 C in N2-atmosphere, the three SiCN compounds still exhibit an almost entirely amorphous microstructure and only local areas estimated to less than 5% of the total volume show the pres-

Fig. 5. TEM micrographs of (a) local crystalline areas upon heat-treatments at 1400 C (b) showing the high-resolution detail of a characteristic aSi3N4 crystals. (c) XEDS spectra acquired from the three regions highlighted in (a) and (d) High-resolution TEM image of isolated turbostratic carbon features in HVNG upon exposure at 1400 C.

G. Gregori et al. / Journal of Non-Crystalline Solids 351 (2005) 1393–1402

ence of crystalline features (see Fig. 5(a) and (b)). Typically, these areas are characterized by the presence of aSi3N4 crystals that, in some regions, are surrounded by turbostratic carbon (no carbothermal reduction). These local features were observed for all three compositions. HVNG exhibited also the presence of localized and isolated turbostratic carbon features, as shown in Fig. 5(d). XEDS analysis was performed on the regions containing Si3N4 crystals in order to acquire information about possible compositional gradients. The spectra shown in Fig. 5(c) were collected from spots corresponding to different microstructures: (1) small crystallites, (2) large a-Si3N4 crystals and (3) the amorphous matrix, as indicated in Fig. 5(a). The XEDS data show that the chemical composition of these regions changes from spot to spot. In particular, while the surrounding amorphous matrix is composed of Si–C–N, as shown by spectrum (3), the region (1), close to the large a-Si3N4 crystals, contains also oxygen. As far as spot (2) is concerned, the presence of carbon shown by the corresponding spectrum is due to the presence of BSUs in close proximity to a-Si3N4. Since analogous results were obtained over several crystalline regions within all samples analyzed, it is concluded that a direct correlation between local crystallization and O-content exists.

4. Discussion Silicon carbonitrides prepared starting from HVNG, HPS and ABSE precursors are completely amorphous after pyrolysis at 1000 C. NS data indicates that their amorphous bulk consists of two main phases: (i) a Si– C–N network and (ii) a free carbon phase. According to the NS results, Si-atoms are bonded to N-atoms while the signal corresponding to the Si–C atomic pair ˚ ) is not clearly observed. However, distance (at ca. 1.88 A since the EELS data of the Si-L ionization edge indicates the presence of both Si–N and Si–C bonding, it is concluded that a fraction of C-atoms is incorporated into the Si-based network. This observation is also supported by previous NMR studies performed on SiCN-powders processed starting from the same precursors considered here [18–21]. NS data collected from the different compounds show ˚ ) differ that the first-order peaks (ca. at 1.42 and 1.74 A in intensity from composition to composition, while the second-order peaks do not exhibit significant variations. With respect to the first-order features, ABSE reveals ˚ , HPS the the most intense peaks at 1.42 and 1.75 A weakest, while HVNG has a rather weak peak at ˚ followed by a strong peak at 1.74 A ˚ . Since the 1.42 A intensity of the PCFs is directly related to the average number of surrounding atoms, this provides information about the atomic coordination within graphite-like

1399

and Si3N4-like domains. A qualitative evaluation of the peak intensities allows to conclude that ABSE presents the highest coordination number (CN) in both graphite-like and silicon nitride domains, HPS the lowest, while HVNG exhibits a low coordination number for the free carbon phase and a high CN for amorphous Si–N. This suggests that the three amorphous networks are characterized by different short-range order. This result is particularly interesting if compared to the stoichiometry data listed in Table 1 and implies that after pyrolysis the short-range order is strongly affected by the different structure of the starting precursors (compare also Fig. 1). The Raman data shown in Fig. 4 indicate the presence of an amorphous graphite-like carbon phase and are consistent with the NS results showing the presence of atomic distances characteristic of low-ordered carbon. The Raman signals arise from the different hybridization states of carbon atoms as well as to their arrangement within the network. The G vibrational mode has E2g symmetry and is generated by in-plane bond stretching of sp2-hybridized carbon pairs [36]. This mode is active at all sp2 sites and not necessarily limited to those arranged in sixfold symmetry. The D-mode has A1g symmetry and corresponds to the breathing mode of aromatic rings [36]. This mode is forbidden in perfect graphite and becomes active only in the presence of local disorder. In general, disorder in a graphite-like phase like the free carbon present in these SiCN compounds can be ascribed to (i) randomly oriented and stretched sixfold carbon rings, (ii) the presence of rings with different symmetry, e.g., fivefold and sevenfold rings, which requires sp3-hybridization and (iii) distorted graphene planes. Deformation of the carbon sheets, e.g. out of plane bending, can be achieved only if some sp2-sites change their hybridization state from sp2 to sp3, as observed during the characterization of carbon nanotubes [41]. In a graphite-like phase, the position of the G-peak is affected by the amount of sp3-sites [36]. For example, increasing content of sp3-sites induces a shift of the G-peak position towards lower vibration frequencies. The peak width also depends on the microstructural configuration of the free carbon, so that, for example, crystalline graphite yields a sharp G-mode, while the presence of distorted rings or rings with fivefold/sevenfold symmetry results in a widening of the D-peak. The shape of the spectra and the relative low value of the Raman shift of the G-mode suggest that the free carbon phase contains a high amount of sp3-hybridized carbon after pyrolysis at 1000 C. This can be due to (i) a high degree of disorder within the network (bent carbon layers), (ii) residual hydrogen atoms that, initially present in the starting precursors, can remain in the bulk up to ca. 1200 C, as reported by Traßl et al. [20] and

1400

G. Gregori et al. / Journal of Non-Crystalline Solids 351 (2005) 1393–1402

(iii) the possible presence of Si–C bondings at the interface between free carbon and the Si–C–N phase. EELS data showing a broad r* peak for both ABSE and HVNG compositions are consistent with these results (Fig. 3(e) and (f)). Exposure at 1400 C in N2-atmosphere induces modifications within the network structure of these compounds, as illustrated by the NS results given in Fig. 2(b). In general, the short-range peaks show increased intensity while new weak signals can be clearly observed in the mid range of the HPS and ABSE samples. This is an indication that both phases (Si-based network and free carbon) undergo a rearrangement/ordering process, which occurs in all three systems, however, with some characteristic differences. In particular, the amorphous bulk of HVNG is differently arranged, as compared to HPS and ABSE. The low intensity of the Si–N peaks suggests that in HVNG the configuration of the Si–Nrich domains is more distorted than in HPS and ABSE. The ABSE composition exhibits more well-defined peaks whose intensity ratio remains unchanged between 1000 C and 1400 C. With respect to the HPS specimen, its PCF is very close to the one recorded from the ABSE sample. After exposure to 1400 C, the HPS/ABSE peaks intensity and, thus, the atomic coordination of the related graphite-like and Si–N-rich domains are very similar. This indicates that precursor-derived systems, which have very close starting composition and different starting polymeric arrangement (HPS and ABSE) reached, after pyrolysis at 1000 C, different metastable configurations. Further heat-treatment at 1400 C activated a rearrangement of the different amorphous configurations that resulted in the formation of similar glass networks. Therefore, it is proposed that these SiCN polymer-derived ceramics (ABSE and HPS) form a ÔcommonÕ metastable glass structure at high temperatures, independently of the starting arrangement of their constituting units formed upon pyrolysis. Another important aspect regarding the rearrangement process is that this proceeds essentially within the two separated phases without apparent interaction between them. Indeed, the Si-L edge remains unchanged upon thermal treatment at 1400 C, suggesting that nature and proportion of chemical bonds (Si–N vs. Si–C) present in the Si-based network is unaffected by annealing at 1400 C. A possible further incorporation of carbon atoms into the silicon-based network did not occur, because otherwise the L3 peak would have shifted towards lower energy-loss values. Similarly, the expulsion of carbon from the Si–C–N network would result in the shift of the edge onset towards higher energy-loss values and therefore is excluded as a possible mechanism here. As far as the C-K ionization edge (EELS) is concerned, the heat-treatment at 1400 C results in a remarkable change of the near-edge structure. This is consistent with the NS results indicating the intensifica-

tion of C–C and C–(C)–C peaks upon annealing, which is related to (i) the elimination of residual hydrogen still present in the bulk upon pyrolysis at 1000 C that contributes to an increase in the fraction of sp2-sites and (ii) the ordering process (graphitization) of the free carbon phase. This assumption is also validated by Raman spectroscopy through the analysis of the evolution of the D- and G-signals upon different thermal treatments. First, the increased intensity of the G-peak is a clear indication of the reduced amount of sp3 sites within the carbon phase and, second, the increased separation between the D- and G-peaks is related to the graphitization process. TEM analysis revealed the presence of closed porosity homogeneously dispersed within the amorphous bulk. In SiCN materials, closed porosity is created during the crosslinking process and, therefore, below 500 C. Thermogravimetric analysis (TGA) coupled to Fourier transform infrared spectroscopy (FTIR) detected the emission of CH4 at temperature ranging between 500 and 800 C [6] in SiCN compounds prepared from the same starting precursors used here. Therefore, the presence of BSU upon pyrolysis within closed pores is referred to gas-phase reactions involving CH4 formation. TEM allowed the detection of local crystallization upon annealing at 1400 C. The volume fraction involved by this localized phenomenon is extremely low compared to the remaining bulk. This evidence in conjunction with the fact that most of these a-Si3N4 crystals are rather large (sub-micron scale compared to ˚ -scale of the ordering process detected by NS) the A led to the conclusion that they do not result from the network rearrangement process but are most likely the effect of local processes. In a previous work, Li et al. [42] investigated the growth of a-Si3N4 whiskers from SiCN powders upon annealing at 1600 C in N2-atmosphere. The authors concluded that the presence of oxygen as impurity on the surface of the SiCN powder yielded the release of SiO gas, which promoted the growth of the Si3N4 crystals via vapor/solid reaction. Here, the concomitant presence of oxygen contamination (see XEDS data) and porosity can promote the local release of SiO and CO gases and, therefore, favor the local growth of large Si3N4 crystals (in presence of nitrogen) according to the following reactions [43] 3SiO þ 3CO þ 2N2 ! Si3 N4 þ 3CO2

ð2Þ

6SiO þ 4N2 ! 2Si3 N4 þ 3O2

ð3Þ

Also the detection of local BSU and isolated turbostratic structures (HVNG) can be explained in terms of gas-phase reactions involving CH4. Interestingly, there was no evidence of the presence of SiC crystals in all the SiCN compositions investigated here. As illustrated

G. Gregori et al. / Journal of Non-Crystalline Solids 351 (2005) 1393–1402

before, although the presence of Si–C bonding was detected by EELS, there are no clear evidences of an ordering process involving Si–C-rich domains upon exposure at 1400 C. Therefore, the formation of SiC precipitates in these compounds is due only to the carbothermal reduction of Si3N4-like regions at temperatures exceeding 1440 C. It is important to highlight here that the formation of these local structures involves only a small volume fraction of these materials while the bulk of these SiCN compounds undergoes a homogeneous ordering process on the atomic scale. These materials remain structurally amorphous and can, therefore, be considered stable against devitrification within this temperature range (1000–1400 C).

5. Conclusions Three different SiCN-compounds, namely HVNG, HPS and ABSE, were investigated upon exposure to temperatures ranging between 1000 and 1400 C, using different experimental techniques, i.e., neutron scattering, Raman spectroscopy, high-resolution TEM in conjunction with XEDS and EELS. The purpose of this work was to investigate the microstructural evolution of these intrinsically amorphous materials (upon pyrolysis) with particular emphasis on the relationship between network rearrangement and early stages of crystallization. The main results obtained in the present work can be summarized as follows: 1. Pair correlation functions obtained from neutron scattering experiments allow the identification of Si– N and C–C atomic pair distances. Six hours annealing at 1400 C (N2, 1 atm) induces short-range rearrangement within the amorphous networks. Note that compounds with similar composition but different starting polymeric architecture (HPS and ABSE) undergo a network reconfiguration process that ultimately results in a very similar amorphous network structure. 2. The bulk SiCN glasses remain predominantly amorphous after heat-treatment at 1400 C and therefore are stable against devitrification within this temperatures range. Crystalline features (Si3N4 crystals and graphitic carbon) could be detected only locally (less than 5 vol.%). Their presence is not related to the short-range rearrangement occurring within the amorphous bulk, but is a result of gas-phase reactions. The emission of CH4 in closed pores induces the formation of graphitic carbon structures while the presence of oxygen can promote the local nucleation and growth of Si3N4 crystals through gas-phase reactions involving the release of SiO and CO (in N2atmosphere).

1401

3. The lack of evidences of the presence of SiC crystals within this temperature range suggests that, for these compounds, the formation of SiC crystals only occurs above the temperature of carbothermal reduction of Si3N4-rich domains (>1440 C).

Acknowledgements The authors wish to thank J. Hacker and Dr G. Motz, University of Bayreuth, Germany, for the precursors synthesis. The authors are grateful to Professor V. Sergo for the use of the Raman Spectroscopy Laboratory at the Applied Chemistry and Materials Engineering Department, University of Trieste. Dr A. Hannon is thanked for the helpful discussions regarding the Neutron Scattering measurements. The European Commission and the German Science Foundation (DFG) are thanked for their financial support.

References [1] R. Riedel, H.-J. Kleebe, H. Scho¨nfelder, F. Aldinger, Nature 374 (1995) 526. [2] J. Lu¨cke, J. Hacker, D. Suttor, G. Ziegler, Appl. Organomet. Chem. 11 (1997) 181. [3] G. Thurn, J. Canel, J. Bill, F. Aldinger, J. Eur. Ceram. Soc. 19 (1999) 2317. [4] S.R. Shah, R. Raj, J. Am. Ceram. Soc. 84 (2001) 2208. [5] J. Hacker, G. Motz, G. Ziegler, in: W. Krenkel, R. Naslain, H. Schneider (Eds.), High Temperature Ceramic Matrix Composites, VCH, Weinheim, 2001, p. 52. [6] W. Weibelzahl, PhD thesis work (in German), University of Bayreuth, 2002. [7] W. Verbeek, Ger. Pat. no. 2218960 (Bayer AG), 8 November 1973 (US Pat. no. 3853567). [8] S. Yajima, Y. Hasegawa, K. Okamura, T. Matsuzawa, Nature 273 (1978) 525. [9] Z.C. Wang, F. Aldinger, R. Riedel, J. Am. Ceram. Soc. 84 (2001) 2179. [10] L.A. Liew, W. Zhang, V.M. Bright, L. An, M.L. Dunn, R. Raj, Sens. Actuators: A 89 (2001) 64. [11] G. Ziegler, I. Richter, D. Suttor, Composites: Part A 30 (1999) 411. [12] G. Motz, J. Hacker, G. Ziegler, in: T. Jessen, E. Ustundag (Eds.), Ceramic Eng. Sci. Proc., Vol.21, 2000, p. 307. [13] G. Motz, G. Ziegler, Key Eng. Mat. 206–213 (2002) 1989. [14] E. Kroke, Y.-L. Li, C. Konetschny, E. Lecomte, C. Fasel, R. Riedel, Mat. Sci. Eng. 26 (2000) 97. [15] J. Bill, J. Seitz, G. Thurn, J. Du¨rr, J. Canel, B.Z. Janos, A. Jalowiecki, D. Sauter, S. Schempp, H.P. Lamparter, J. Mayer, F. Aldinger, Phys. Stat. Sol. (a) 166 (1998) 269. [16] J. Bill, J. Schuhmacher, K. Mu¨ller, S. Schempp, J. Seitz, J. Du¨rr, H.-P. Lamparter, J. Golczewski, J. Peng, H.J. Seifert, F. Aldinger, Z. Metallkd. 91 (2000) 4. [17] J. Seitz, J. Bill, N. Eggert, F. Aldinger, J. Eur. Ceram. Soc. 16 (1996) 885. [18] S. Traßl, D. Suttor, G. Motz, E. Ro¨ssler, G. Ziegler, J. Eur. Ceram. Soc. 20 (2000) 215.

1402

G. Gregori et al. / Journal of Non-Crystalline Solids 351 (2005) 1393–1402

[19] S. Traßl, G. Motz, E. Ro¨ssler, G. Ziegler, J. Non-Cryst. Solids 293–295 (2001) 261. [20] S. Traßl, G. Motz, E. Ro¨ssler, G. Ziegler, J. Am. Ceram. Soc. 85 (2002) 239. [21] S. Traßl, H.-J. Kleebe, H. Sto¨rmer, G. Motz, E. Ro¨ssler, G. Ziegler, J. Am. Ceram. Soc. 85 (2002) 1268. [22] J. Du¨rr, P. Lamparter, J. Bill, S. Steeb, F. Aldinger, J. NonCryst. Solids 232–234 (1998) 155. [23] M. Monthioux, O. Delverdier, J. Eur. Ceram. Soc. 16 (1996) 721. [24] H.-J. Kleebe, D. Suttor, H. Mu¨ller, G. Ziegler, J. Am. Ceram. Soc. 81 (1998) 2971. [25] H.-J. Kleebe, Phys. Stat. Sol. (a) 166 (1998) 297. [26] H. Sto¨rmer, PhD thesis work (in German), University of Bayreuth, 2001. [27] H.-J. Kleebe, H. Sto¨rmer, S. Traßl, G. Ziegler, Appl. Organomet. Chem. 15 (2001) 858. [28] H. Brequel, S. Enzo, G. Gregori, H.-J. Kleebe, A. Hannon, Mater. Sci. Forum 386–388 (2002) 365. [29] J. Dixmier, R. Bellissent, D. Bahloul, P. Goursat, J. Eur. Ceram. Soc. 13 (1994) 293. [30] A. Burian, A. Ratuszna, J.C. Dore, S.W. Howells, Carbon 36 (1998) 1613.

[31] T.W. Zerda, W. Tu, A. Zerda, Y. Zhao, R.B. von Dreele, Carbon 38 (2000) 355. [32] S. Waidmann, M. Knupfer, J. Fink, B. Kleinsorge, J. Robertson, Diam. and Rel. Mater. 9 (2000) 722. [33] S. Trasobares, O. Stephan, C. Colliex, W.K. Hsu, H.W. Kroto, D.R.M. Walton, J. Chem. Phys. 116 (2002) 8966. [34] A.R. Merchant, D.G. McCulloch, R. Brydson, Diam. and Rel. Mater. 7 (1998) 1303. [35] F. Tuinstra, J.L. Koening, J. Chem. Phys. 53 (1970) 1126. [36] A.C. Ferrari, J. Robertson, Phys. Rev. B 61 (2000) 14095. [37] R.E. Shroder, J. Nemanich, J.T. Glass, Phys. Rev. B 41 (1990) 3738. [38] M.A. Tamor, W.C. Vassell, J. Appl. Phys. 76 (1994) 3823. [39] B. Marchon, J. Gui, K. Grannen, G.C. Rauch, IEEE Trans. on Magn. 33 (1997) 3148. [40] A. Oberlin, in: P.A. Thrower (Ed.), Chemistry and Physics of Carbon, Dekker, New York, 1989, p. 22. [41] T.W. Ebbesen, T. Takada, Carbon 33 (1995) 973. [42] Y.-L. Li, Y. Liang, Z.-Q. Hu, J. Mat. Sci. 31 (1996) 2677. [43] O. Delverdier, M. Monthioux, D. Mocaer, R. Pailler, J. Eur. Ceram. Soc. 14 (1994) 313.