Structure analysis and properties of Si-C-N ceramics derived from polysilazanes

Article abstract of DOI:10.1002/(SICI)1521-396X(199803)166:1<269::AID-PSSA269>3.0.CO;2-7

Pyrolysis of different polysilazanes has been used to prepare novel covalent amorphous ceramics composed of silicon, carbon, and nitrogen. The formation and structure of the as-pyrolyzed amorphous state and its devitrification into stable crystalline phas

Full text of DOI:10.1002/(SICI)1521-396X(199803)166:1<269::AID-PSSA269>3.0.CO;2-7

J. Bill et al.: Structure Analysis and Properties of Si±C±N Ceramics
269
phys. stat. sol. (a) 166, 269 (1998)
Subject classification: 61.43.±±j; 61.41.‡e; S6; S11
Structure Analysis and Properties
of Si±C±N Ceramics Derived from Polysilazanes

J. Bill (a), J. Seitz (a), G. Thurn (b), J. Durr (b), J. Canel (b),
B. Z. Janos (b), A. Jalowiecki (b), D. Sauter (a), S. Schempp (b),
H. P. Lamparter (b), J. Mayer (b), and F. Aldinger (a, b)
(a) Institut fur Nichtmetallische Anorganische Materialien, Universitat Stuttgart,
(b) Max-Planck-Institut fur Metallforschung,
Pulvermetallurgisches Laboratorium, Heisenbergstr. 5, D-70569 Stuttgart, Germany
(Received January 15, 1998)

Dedicated to Professor Dr. Manfred Ruhle on the occasion of his 60th birthday
Pyrolysis of different polysilazanes has been used to prepare novel covalent amorphous ceramics
composed of silicon, carbon, and nitrogen. The formation and structure of the as-pyrolyzed amor-
phous state and its devitrification into stable crystalline phases have been investigated by means of
nuclear magnetic resonance spectroscopy, infrared spectroscopy, thermogravimetry, mass spectro-
scopy, X-ray and neutron diffraction, and by transmission electron microscopy. Additionally, the
electrical conductivity and the compression creep behaviour of the as-received materials have been
analyzed.
1. Introduction
Silicon nitride, silicon carbide and silicon nitride/carbide-based composites are suitable
candidates for high temperature applications because of their high thermal, chemical
and mechanical stability. One possibility for the preparation of such ceramics is liquid
phase sintering of silicon nitride and/or carbide powders in the presence of oxidic addi-
tives, like Al2O3 and Y2O3 [1 to 3]. Since during the densification process these sintering
aids vitrify and segregate at the grain boundaries they can lead to a deterioration of the
mechanical and chemical stability of sintered materials at elevated temperatures due to
softening and enhanced oxygen diffusion, respectively, provided by their glassy nature.
Therefore, current research activities with respect to high temperature applications of
such materials focus on the development of powder technology routes which allow the
reduction or even the elimination of the fraction of additives.
A proper route of a sinter aid-free preparation of carbide/nitride ceramics is the pyro-
lysis of preceramic polymers [4 to 8]. The main aim of this processing route is the archi-
tectural design of ceramics on the basis of molecular units in order to control the struc-
ture and the properties of such materials on an atomic scale. According to this route the
preceramic polymers are at first transformed into amorphous ceramics by pyrolysis at
temperatures dependent on their chemical composition and molecular structure. With a
subsequent heat treatment at higher temperatures such amorphous solids can be crystal-
lized into thermodynamically stable phases.

270
J. Bill et al.
Synthesis and thermal treatment of the precursor polymers allow the control of their
physical and chemical properties providing processing opportunities for the preparation
of ceramic fibers, films or monoliths without using sintering aids.
This paper describes the synthesis of silicon nitride/carbide-based ceramics by pyroly-
sis of polyhydridomethylsilazane (PHMS) and polyvinylsilazane (PVS). It is the aim of
this paper to combine experimental results of different previous studies [9 to 15] and
recent unpublished data in order to present a comprehensive description of the structure
formation during ceramization and crystallization of these two commercially available
polysilazanes. Additionally, recent results on electrical conductivity measurements [16]
and on compression creep tests [17, 30] of ceramics derived from these precursors are
reported.
2. Experimental
The structure units of both polymers used for the preparation of Si±C±N ceramics are
shown in Figs. 1 and 2. PHMS (NCP 200, Nichimen Corp., Japan) is a copolymer of
methyl- and dimethylsilazane units in the ratio 2:1 (calculated from the corresponding
peak areas in the 29Si NMR spectrum). PVS (VT 50, Hoechst AG, Germany) consists of
vinylsilazane units and about 10%-N(CH3)2 end group units [18].
In order to form preceramic networks which do not melt during pyrolysis the as-re-
ceived polymers were crosslinked at 400 ꢀC for 4.5 h (PHMS) or at 240 ꢀC for 2 h (PVS)
and subsequently milled and sieved (28 mm screen). The resulting powders subsequently
were compacted either by cold-isostatic pressing at 635 MPa or by pressing at elevated
temperature (plastic forming) [9] using a maximum pressure of 47.75 MPa and tempera-
tures of 275 ꢀC (PHMS) or 250 ꢀC (PVS). For pyrolysis the cold or warm compacted
samples were placed in quartz tubes in a slight argon flow and heated with a heating
rate of 25 K/h up to different temperatures (350, 550, 625, 700 or 1050 ꢀC) which were
held for 4 h.
By cold-isostatic pressing and subsequent pyrolysis amorphous ceramics with a poros-
ity of 10 to 15% (determined by mercury porosimetry) were obtained. If plastic forming
was applied the green density could be increased and the porosity of the as-pyrolyzed
material thereby reduced to 3%.
For crystallization studies the as-pyrolyzed materials were heat treated in N2 atmos-
phere for 5 to 50 h at different temperatures in the range of 1200 to 1450 ꢀC (heating
rate: 120 K/h).
The Nuclear Magnetic Resonance Spectroscopy (NMR) experiments were performed
on a Bruker 300 MSL spectrometer [11]. The resonance frequency for 1H was
300.13 MHz, for 13C 75.47 MHz and for 29Si 59.60 MHz. All spectra were acquired using
Fig. 1. Structure units of the PHMS polymer

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
271
Fig. 2. Structure units of the PVS polymer
magic angle spinning (MAS) with a rotation frequency of 4.5 kHz. The excitation of the
observed nucleus was 3 ms for 1H, 4 ms for 13C and 5 ms for 29Si. All spectra were 1H-
decoupled during the data acquisition. The samples were also measured with a crosspo-
larization sequence (CP). The contact time for 13C and 29Si was 1 and 5 ms, respec-
tively. The repetition time was 10 s.
The Infrared Spectroscopy (IR) investigations were performed on a Bruker FT-IR IFS
66 using diffuse reflexion technique (powder samples).
The Thermogravimetry and Mass Spectrometry (TG/MS) investigations were per-
formed on a Netzsch 409/429-403 machine in vacuum. The heating rate was 5 K/min up
ꢀ
to 1200 C.
The X-ray wide angle scattering was done in transmission mode using AgKa radiation
[12].
The neutron wide angle scattering experiments were performed at the ISIS facility of
the Rutherford Appleton Laboratory, UK,ꢁu1sing the SANDALS instrument [12]. There,
ꢀ
a range of wave vector transfer up to 25 A could be measured. This is essential for the
resolution of the radial distribution function in real space. Neutron small angle scatter-
ing was performed at the BENSC facilitiy of the HMI, Berlin, using the V4 instrument.
ꢀ
The neutron wavelength was 8 A [12].
For the Transmission Electron Microscopy (TEM) investigations were performed with
specimens fabricated by standard preparation techniques (mechanically sectioning the
materials, polishing and dimpling the 3 mm discs, which was followed by a final Ar ion
beam thinning). The specimen were characterized by conventional (CTEM), energy fil-
tering (EFTEM) and high resolution TEM (HRTEM). For the studies a Zeiss EM 912
Omega and a JEOL 4000EX were used [10].
Besides various conventional TEM methods two dimensional elemental distribution
images in the electron spectroscopic imaging (ESI) technique have been proven to be an
important tool for the characterization of the chemical microstructure in large specimen
areas (several mm2) at high resolution (2 to 3 nm) within short times [19 to 22].
Electrical conductivity measurements were done with cold compacted and pyrolyzed
samples by means of two point measurements in a temperature range from room tempera-
ture up to 500 ꢀC [16]. The device used consists of a high temperature furnace (GERO), a
measuring cell unit (made of quartz glass), a measuring device for voltage and current
(Keithley 236), a temperature measuring device (THERM 2281-8), and a computer.
For the measurements the as-pyrolyzed samples were cut and ground into rectangular
shaped bars. In order to ensure the electrical contact between the material and the plati-
ꢀ
num electrodes a 3000 A thick platinum film was deposited onto the two surfaces per-

272
J. Bill et al.
pendicular to the longitudinal axis of the samples by sputtering. Subsequently, the sam-
ples were placed into evacuated quartz glass cells. The temperature and its gradient
along the longitudinal axis of the sample were measured by two thermocouples placed at
both ends of the sample. The thermoelectric voltage applied to the sample is read into
the computer through an analog±digital converter.
For the determination of the sample conductivity, a voltage is applied to the sample
and the resulting current read into the computer through the serial IO port. A control
program is used to determine the electrical resistance of the sample and to plot the
electrical conductivity against the sample temperature.
Compression creep studies [17, 30] have been done with material densified by plastic
forming and subsequent pyrolysis. The test specimen were cut and ground into rectangu-
lar bars with a height of 7.5 mm and with a cross sectional area of 1.5 Â 1.5 mm2. The
small cross sectional area was chosen because stresses up to 250 MPa were applied and
the maximum load of the creep apparatus was limited to 1000 N at high temperatures.
The height of the specimen was also limited to avoid buckling of the specimen. A square
cross sectional area was chosen to facilitate the preparation of the specimen.
A spring loaded creep apparatus (Amsler DSM 6101) with a 1600 ꢀC furnace (Maytec
Achtkantofen) was used for the testing. Inside the furnace alumina push rods and silicon
carbide load pads were used to apply the testing force. A spiral spring was used for the
loading. The strain of spring was measured by a linear potentiometer and the testing
force was calculated from the strain of the spring. An accuracy of Æ1 N was achieved for
the force measurements. Two SiC scanning pins placed at the upper load pad and an-
other scanning pin at the lower load pad were used to measure the length change of the
specimen. The scanning pins were connected to an inductive strain gage with a resolu-
tion of 0.1 mm in the cold part of the testing machine.
Testing was done in air at temperatures of 1450, 1500 or 1550 ꢀC. The temperature
was measured with PtRh(10)/Pt thermocouple which was placed at a distance of about
2 mm from the specimen. The specimens were heated up to the testing temperature as
fast as possible using the maximum power of the furnace to avoid oxidation at inter-
mediate temperatures. The heating ramp for tests at 1550 ꢀC is shown in Fig. 3.
Fig. 3. Heating ramp used for creep tests at 1550 ꢀC

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
273
During the heating a clamping force of about 2 N was applied to the specimen. To get
the displacement measuring system into thermal balance after test temperature was
reached, the creep stress was not applied until 30 min. The loading of the specimen took
30 s. The strain measurement was started after reaching the creep stress.
3. Structure Formation
The structure and phase formation during ceramization and crystallization of preceramic
compounds is very much dependent on the composition and molecular structure of the
precursor. Fig. 4 shows the composition of the ceramics obtained by pyrolysis of PHMS
and PVS at 1050 ꢀC within isothermal sections of the Si±C±N phase diagram valid up
to a) 1438 ꢀC and b) between 1438 and 1876 ꢀC.
As can be seen from this figure (crystalline) Si3N4, SiC and C form a three-phase
equilibrium which is stable up to 1438 ꢀC. Above this temperature Si3N4 is unstable
next to C, thus decomposition occurs by loss of nitrogen according to the equation
Si3N4 ‡ 3 C ! 3 SiC ‡ 2 N2 .
(1)
I.e., at temperatures above 1438 ꢀC the ceramics decompose into SiC and C or SiC and
Si3N4 depending on their composition.
According to these phase equilibria the pyrolysis of PVS should result in the forma-
tion of a two-phase material consisting of Si3N4 and C whereas the ceramization of
PHMS should lead to a material composed of Si3N4, SiC and C. At 1438 ꢀC both materi-
als should transform into SiC/C and Si3N4/SiC composites, respectively combined with
a loss of nitrogen. At 1876 ꢀC and ambient pressure Si3N4 dissociates into nitrogen and
liquid Si.
Fig. 4. Isothermal section of the ternary system Si±C±N. a) below 1438 ꢀC and b) in the tempera-
ture range between 1438 and 1876 ꢀC [23]. The compositions of the ceramics obtained by pyrolysis
ꢀ
*
of the polysilazanes PHMS (&) and PVS ( ) at 1050 C are inserted

274
J. Bill et al.
Fig. 5. 13C-NMR spectra of PHMS at dif-
ferent annealing temperatures up to
1050 ꢀC [11]
3.1 Ceramization
The ceramization of the crosslinked
PHMS and PVS, i.e. the structure for-
mation during their pyrolysis has been
studied by NMR, IR, TG/MS, X-ray
and neutron diffraction.
3.1.1 Polysilazane PHMS
The results of the NMR investigations
[11] are summarized in Figs. 5 and 6,
those of the IR investigations in Fig. 7,
and those of the X-ray and neutron dif-
fraction studies [12] in Figs. 8 and Fig. 9.
The results of these investigations
reveal the reaction process of cerami-
zation as shown in Fig. 10.
In detail the results can be explained
as follows: The NMR 13C spectrum of
the polymer (Fig. 5) shows a peak at
5 ppm which is assigned to the Si±CH3
group. The 29Si spectrum of the polymer
(Fig. 6) shows three signals. The peaks
at ±±22 ppm and ±±5 ppm are caused
by SiN2CH and SiN2C2 environments,
respectively, within a six-membered
cyclosilazane ring. The peak at ±±14
ppm is assigned to the bridging SiN3H
sites in the six-membered ring.
The IR spectroscopy of the polymer
(Fig. 7) shows signals which are as-
signed to N±H (3405 cm±±1), C±H
(2960 to 2910 cm±±1) and Si±H (2130 cm±±1) [24] and signals of SiC±H (1265 and
1405 cm±±1), Si2N±H (1180 cm±±1) and Si±CH2±Si (1035 cm±±1) (caused by a small
amount of CH2 bridges obtained by crosslinking reactions described below).
The structure of the polymer based on both the NMR and IR results is shown in
Fig. 11. The structure consists of crosslinked six-membered rings as proven by the re-
sults of the 29Si NMR spectrum of the polymer (see above).
The NMR as well as the IR spectra of the product heated up to 350 ꢀC mainly re-
mained unchanged as compared to those of the as-received polymer.

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
275
Fig. 6. 29Si±NMR spectra of PHMS
at different annealing temperatures
up to 1050 ꢀC [11]
At 550 ꢀC in the NMR 13C spectrum (Fig. 5) an additional peak at 12 ppm is found as
a shoulder. This is attributed to CH2 bridges which were formed by a crosslinking reac-
tion of the methyl groups with the hydrogen of Si±H (2) or C±H (3) groups:
-CH3 ‡ H±Siꢂ ! -CH2±Siꢂ ‡ H2 ,
(2)
(3)
-CH3 ‡ CH3±Siꢂ! -CH2±Siꢂ ‡ CH4 .

276
J. Bill et al.
Fig. 7. IR spectra of PHMS at different annealing temperatures up to 1050 ꢀC
The occurrence of these reactions is supported by means of TG/MS investigations which
show an evaporation of methane in the temperature range between 400 and 800 ꢀC and
of hydrogen between 350 and 1050 ꢀC, respectively. The same is true for the IR investi-
gations which show with increasing temperature decreasing signals assigned to N±H,
C±H, Si±H, SiC±H, Si2N±H, and Si±CH2±Si (Fig. 7).
29
The Si spectrum at 550 ꢀC (Fig. 6) shows also a broadening of the lines. This broad-
ening is due to a modification of the environment of the Si atoms caused by the crosslink-
ing reaction (see Fig. 10). The signals of the SiN2CH and the SiN3H sites combine. The
number of SiN2C2 sites were increased at the expense of the SiN2CH sites which can be
detected by calculating of the appropriate peak areas (the signal was fitted with three
peaks to estimate the several components). The ratio of SiN2C2 and SiN2CH has changed
from 1:2 (polymer) to 1:1 (550 ꢀC product) indicating that the crosslinking reactions men-
tioned above ((2) and (3)) occur preferably via the reaction of Si±H groups with methyl
groups which leads to the formation of H2. This is in accordance with the MS investiga-
tions which reveal that the evaporation of H2 starts at lower temperatures compared to the
formation of methane caused by the reaction of Si±C groups with methyl groups.

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
277
Fig. 8. Total pair correlation functions, a) from X-ray diffraction, b) from neutron diffraction [12]
During heating up to 625 ꢀC the successive increase of the chemical shift in the 13C
spectrum (Fig. 5) continued, caused by the increased coordination of the C atoms with
Si atoms (Si±CH3: 5 ppm; Si±CH2±Si: 12 ppm; Si2±CH±Si: 18 ppm; Si4C: 24 ppm). In
the 29Si spectrum (Fig. 6) a signal at
ꢁ43 ppm is detected which is assigned to
SiN4 sites. These sites were formed by
crosslinking reactions of the NH groups
with Si±H and Si±CH3 groups, respec-
tively (see Fig. 10). Moreover, TG/MS in-
vestigations showed an evaporation of a
small amount of ammonia between 500
and 650 ꢀC. This is caused by the trans-
amination reaction
3 H±NSi2 ! 2 NSi3 ‡ NH3 . (4)
ꢀ
Fig. 9. Gaussian fitting to the peak at 1.74 A
of the RDF(r)-function corresponding to Gn(r)
from Fig. 8 [12]

278
J. Bill et al.
Fig. 10. Mechanisms suggested
for the ceramization of the poly-
silazane PHMS
Thus, the coordination of the Si atoms with N atoms and vice versa increases succes-
sively.
At 1050 ꢀC, i.e. after complete ceramization, the 13C spectrum (Fig. 5) shows a
broad signal with a maximum located at 24 ppm which is assigned to Si4C sites that
can be considered as structural units of silicon carbide. The 29Si spectrum (Fig. 6)
remains in general unchanged compared to the 625 ꢀC product. The crosslinking reac-
tions continued and additional SiN4 units were formed which are the basic structure
units of Si3N4. Broad signals centered at ±±25 and ±±15 ppm as a shoulder are still
found in the 1050 ꢀC products which confirm the existence of Si units containing Si±C
bonds.
The short-range ordering and structural units which build up the amorphous as-pyro-
lyzed ceramic network as have been revealed from NMR studies have been confirmed by
X-ray and neutron scattering [12]. Wide angle scattering did not indicate any traces of
crystalline peaks but distinct information on short-range ordering. Fig. 8 shows the pair
correlation functions G(r) deduced by Fourier transformation from the measured total
X-ray and neutron structure factor of the as-pyrolyzed material.

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
Fig. 11. Structure of the PHMS polymer
279
With X-rays in the pair correlation function (Gx(r)) two distinct peaks at 1.73 and
ꢀ
2.98 A are observed (Fig. 8a), whereas in the neutron pair correlation function Gn(r)
ꢀ
four peaks at 1.37, 1.74, 2.46 and 2.85 A are displayed (Fig. 8b).
ꢀ
The peaks at 1.37 and 2.46 A in Gn(r) belong to C±C and C±(C)±C correlations,
respectively. With X-rays these peaks are not observed due to the much smaller weight-
ing factor. Note that the peak positions are in good agreement with the distances in
ꢀ
graphite at 1.42 and 2.46 A. Thus, it can be concluded that the investigated ceramics
include amorphous graphite-like carbon.
ꢀ
The peak at 1.74 A reflects the Si±N bonds. It appears with about the same height in
ꢀ
Gx(r) and Gn(r). The peak at 2.85 A in Gn(r) is attributed to a N±(Si)±N correlation,
ꢀ
and the one at 2.98 A in Gx(r) to a Si±(N)±Si correlation. All three peaks are in good
agreement with those of the corresponding distances of crystalline Si3N4.
ꢀ
Nevertheless, the positions of the peaks at 1.74 and 2.85 A are shifted slightly to
larger values of r and they are broadened at the right hand side compared to pure
ꢀ
Si3N4. Considering that in crystalline SiC a direct Si±C distance appears at 1.89 A and
ꢀ
an indirect C±(Si)±C distance at 3.06 A this suggests that in the investigated ceramics
there is also a considerable number of Si±C bonds.
ꢀ
Fig. 9 shows the peak at 1.74 A in the total radial distribution function RDF(r) ˆ
4pr2r0 ‡ rGn(r) (r0 is the atomic density) fitted by the sum of two Gaussian curves at
ꢀ
ꢀ
1.74 and 1.83 A. Note, that the position of the peak in this plot is shifted to 1.76 A
caused by the calculation of the RDF.
Fig. 12. TEM bright field image (left part) and diffraction pattern (right part) of siliconcarboni-
tride obtained from a PHMS precursor after pyrolysis at 1050 ꢀC
19 physica (a) 166/1

280
J. Bill et al.
From the areas of the Gaussian curves the average coordination numbers ZSi±N ˆ 3.0
and ZSi±C ˆ 1.0 were obtained. The sum ZSi ˆ ZSi±N ‡ ZSi±C ˆ 4.0 proves that in the
amorphous ceramics the Si atoms are tetrahedrally coordinated which is in good agree-
ment with the NMR results. It should be noted however that on the basis of the average
values of Z it cannot be distinguished between the existence of SiN4 and SiC4 tetrahedra
only and the additional existence of tetrahedra mixed in C and N (Si(C, N)4).
Despite this extented ordering the as-pyrolyzed material is fully amorphous. This is in
accordance not only with wide angle X-ray and neutron diffraction studies but also with
structural and analytical transmission electron microscopy investigations which did not
show any crystalline precipitates. This is shown by bright field images (Fig. 12), selected
area diffraction, and electron spectroscopic imaging (Fig. 13).
Elemental distribution images of the elements Si, C and N did not show any indica-
tion of the occurrence of segregation on a scale above the resolution limit of about 2 nm
Fig. 13. Elemental distribution images for carbon (above, left), nitrogen (above, right) and silicon
(below, left) obtained with the electron spectroscopic imaging (ESI) technique of a PHMS-derived
ceramic prepared at 1050 ꢀC

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
281
Fig. 14. High resolution micrograph of the
amorphous PHMS-derived Si±C±N ceramic
obtained at 1050 ꢀC. The contrast of the
high resolution micrograph shows the typi-
cal appearance known from perfectly amor-
phous materials
(Fig. 13). However, the formation of
small clusters or crystallites with com-
position close to the average composi-
tion of the material could not be ex-
cluded from these studies. Therefore,
high resolution microscopy was em-
ployed which would reveal structural
inhomogeneities on
a subnanometer
scale. Fig. 14 shows the high resolution
micrograph of a 1050 ꢀC sample with
contrast variations typical for an
amorphous material and with no indi-
cation of the formation of crystallites
or crystalline precipitates.
Fig. 15. Amorphous PHMS-derived Si±C±N ceramic; formation of amorphous Si3N4 regions during
annealing. a) volume fraction, b) radius. 1050, 1200, 1300, 1400, ‡1450 ꢀC. The origin of
the time scale refers to the point in the heating curve where the indicated temperature was reached,
i.e. it involves the heating ramp [12]
&
*
~
!
*
19

282
J. Bill et al.
In order to gain more insight into the medium range order of the amorphous materi-
als, small angle scattering has been applied to study fluctuations of the composition and
density which exhibit a size below the scale of the detection limit of the ESI technique.
In this connection small angle neutron scattering (SANS) cross sections show, that the
material contains inhomogeneities which give rise to an observed SANS signal [12]. The
measured SANS effect has to be attributed to an amorphous silicon nitride phase, segre-
gated into a Si±C±N matrix due to the values determined for the difference of the scat-
tering length densities of the segregations and the matrix. Fig. 15 indicates that the
material contains already in the as-
pyrolyzed stage, i.e. after heat treat-
ment at 1050 ꢀC inhomogeneities in
the size range of 1.2 to 1.4 nm.
Keeping in mind the wide angle
scattering results these inhomogene-
ities were suggested to consist of
amorphous Si3N4, SiC and C phases.
3.1.2 Polysilazane PVS
The results of NMR investigations
[13] are summarized in Figs. 16 and
17 and those of the IR investigations
in Fig. 18. From these results the
ceramization process as shown in
Fig. 19 has been deduced.
The results are in detail as fol-
lows: In the 13C spectrum of the
polymer (Fig. 16) beside two spin-
ning side bands at around 200 and
75 ppm three peaks are detected.
The two major peaks at 142 and
132 ppm are caused by the CˆC
bonding (sp2-hybridized carbon).
The signal between 20 and 40 ppm
is assigned to a small amount of sp3-
hybridized carbon. It is caused by
the -N(CH3)2 end groups.
The 29Si spectrum of the polymer
(Fig. 17) shows two peaks: the sig-
Fig. 16. 13C-NMR spectra of PVS at dif-
ferent annealing temperatures up to
1050 ꢀC

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
283
Fig. 17. 29Si±NMR spectra of PVS
at different annealing temperatures
up to 1050 ꢀC [13]
nal at ±±20 ppm is caused by a SiN3C(sp3) environment (polymerized vinyl groups) and
the peak at ±±35 ppm is assigned to a SiN3C(sp2) environment.
For the product heated up to 250 ꢀC the NMR spectra mainly remained unchanged. After
heating up to 350 ꢀC the 13C spectrum (Fig. 16) shows only a small amount of CˆC groups,
i.e. sp2 carbon and a strongly increased number of sp3 carbon. With increased temperature
more and more vinyl groups are polymerized and form CH2 bridges (sp3) (see Fig. 19). Thus,

284
J. Bill et al.
Fig. 18. IR spectra of the ceramization of PVS at different annealing temperatures up to 1050 ꢀC
the peak at 25 ppm resulting from the CH2C2 groups [25] combines with the signal of the
-N(CH3)2 end groups. An additional signal at 5 ppm is found as a shoulder which is due
to the CHC2Si groups [25]. This is confirmed by the corresponding 29Si spectrum (Fig. 17)
which shows an increased signal of SiN3C(sp3) and a decreased peak of SiN3C(sp2).
Compared to the 350 ꢀC product the 13C spectrum of the 550 ꢀC product (Fig. 16) exhi-
bits an increased signal of sp2 carbon and a decreased signal of sp3 carbon. Moreover, this
transformation of the carbon atoms is accompanied by the evaporation of H2, which begins
at 350 ꢀC and can be still detected at 550 ꢀC by MS. Thus, these results indicate that
hydrogen is removed from the hydrocarbon chains leaving behind sp2 (graphite-like) car-
bon (see Fig. 19). Simultaneously, the sp3 carbon signal is split into two peaks centered at
5 and 25 ppm, respectively. They are due to the two different kinds of sp3 carbon men-
tioned above: with and without Si±C bonds, respectively. If the temperature was raised
further to 625 ꢀC these signals completely disappear due to a strong decrease of the num-
ber of Si±C bonds and of the sp3 carbon content. These results are in accordance with the
29
MS investigations and the Si spectra of the 550 and 625 ꢀC products. Owing to the 29Si
spectra the signal attributed to SiN3C(sp3) sites decreases. Moreover, an increasing con-
tent of SiN4 sites (±±42 ppm) can be detected. These results indicate a transformation of

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
285
Fig. 19. Mechanisms suggested for the ceramization of the polysilazane PVS
SiN3C(sp3) into SiN4 sites connected with a separation of carbon and Si/N-containing
units. Additionally, the MS studies reveal the evaporation of methane in the temperature
range between 400 ꢀC and 600 ꢀC which leads to a further decrease of the sp3 carbon con-
tent and can be attributed to the removal of the N(CH3)2 end groups.
If the temperature was raised up to 1050 ꢀC the 13C spectrum (Fig. 16) shows only
the signal of sp2 carbon. The peak in the 29Si spectrum (Fig. 17) centered at ±±40 ppm
is assigned to a SiN4 environment [13]. Furthermore, the evaporation of a small amount
of ammonia between 700 and 800 ꢀC can be detected by MS and can be assigned to the
transamination reaction mentioned above (4) yielding NSi3 units.
The IR investigations (Fig. 18) confirm the NMR results. With increasing temperature
the signals assigned to N±H (3375 cm±±1), to C±H (methyl group: 3045 to 2900 cm±±1
,
vinyl group: 1915, 1405 and 1010 cm±±1) and to CˆC (1595 cm±±1) [24] decrease. The
Si±N signal is found at 980 cm±±1. The SiC±H signal (1225 cm±±1) decreases during cera-
mization. Moreover, at temperatures above 350 ꢀC the signal of Si±H (2150 cm±±1) is

286
J. Bill et al.
detected. As a consequence dehydrogenation between Si±H and N±H groups can occur
leading to an increased coordination of Si with N atoms which is in accordance with the
above-mentioned NMR results as well as with MS studies that show the evaporation of
ꢀ
hydrogen between 350 and 1050 C.
The ceramization process can be summarized as follows (Fig. 19): The first step is the
polymerization of vinyl groups resulting in the formation of carbon chains (sp3 hybri-
dized carbon). In the temperature range between 550 and 625 ꢀC the Si±C bonds disap-
pear and Si±N bonds are formed. As Fig. 19 shows there are two possible ways for these
reactions: a) splitting of carbon chains leaving behind Si±H bonds which can react with
NH groups resulting in the formation of Si±N bonds and evaporation of H2 and b)
direct reaction of the Si±C bonds with NH groups. Reaction a) is confirmed by the IR
spectra which show the presence of Si±H bonds at temperatures above 350 ꢀC.
3.2 Crystallization
3.2.1 PHMS-derived material
During annealing above the pyrolysis temperature the amorphous stage of the as-pyro-
lyzed material is fully maintained. Elemental distribution images still show a completely
Fig. 20. Bright field image of the PHMS-derived Si±C±N ceramic after heat treatment at 1350 ꢀC.
First crystalline segregations (Si3N4) can be detected in the still amorphous Si±C±N matrix

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
287
homogeneous distribution of the elements. However, according to small angle neutron
scattering results, there occurs both an increase of the volume fraction and a coarsening
of the segregations observed in the amorphous as-pyrolyzed stage during annealing at
this temperature and with further temperature increase up to 1400 ꢀC (Fig. 15).The
dashed lines were fitted using a Gibbs-Evetts-Leake type of annealing function [26]. In
the temperature range around 1200 to 1300 ꢀC the increase in volume of the segregation
is faster than at temperatures as high as 1400 ꢀC. It is to note that at this temperature
the state of equilibrium is not attained within the time interval of 15 h.
The temperature±time dependence of the radius R of the scattering regions in Fig.
15b shows an increase from 1.2 nm in the as-pyrolyzed ceramic up to about 2.5 nm after
a 15 h anneal at 1400 ꢀC.
At 1450 ꢀC a decrease of the volume fraction v of the segregated Si3N4 phase can be
detected [14]. It is to note that the decrease of v at 1450 ꢀC in Fig. 15a is not associated
Fig. 21. Elemental distribution images of the material obtained from PHMS after crystallization at
1500 ꢀC

288
J. Bill et al.
with a decrease of R. This behaviour is explained by the facts that on the one hand the decom-
position affects first the smaller Si3N4 regions, whereas on the other hand the size R, as estab-
lished from small angle scattering, generally is associated mainly with the larger regions.
The reduction of the volume fraction of Si3N4 at 1450 ꢀC corresponds completely with
the phase equilibria shown in Fig. 4, since for temperatures above 1438 ꢀC the Si3N4
content of the PHMS-derived ceramic is reduced according to (1).
In transmission electron microscopy studies at annealing temperatures of 1350 ꢀC first
isolated areas containing crystalline a-Si3N4 grains and free carbon were detected (Fig. 20).
An increase in length or temperature of the heat treatment increases the amount of
crystallinity.
Annealing of the material at 1500 ꢀC yields a completely crystalline material consist-
ing of Si3N4 and SiC. The microstructure of which is characterized by nanocrystalline
SiC grains embedded in a microcrystalline matrix of Si3N4 (Fig. 21).
Fig. 22. Amorphous PVS-derived ceramic after annealing treatment of 50 h at 1350 ꢀC. Large scale
contrast variations in the elemental maps are caused by thickness variations of the TEM specimen

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
289
3.2.2 PVS-derived material
According to electron diffraction and electron spectroscopic imaging the microstructure
of PVS-derived material is still completely amorphous after annealing 50 h at 1350 ꢀC in
nitrogen (Fig. 22). The bright field image and the corresponding elemental maps pro-
duced by electron spectroscopic imaging reveal a homogeneous distribution of the ele-
ments.
The diffraction patterns show an amorphous ring system which confirmed that the
material was fully amorphous. However, the SAD rings show intensity deviations which
trace back to the advanced state of ordering in the amorphous 1350 ꢀC product.
If the annealing temperature is raised to 1500 ꢀC crystallization occurs. Elemental distri-
bution images reveal in conformity with the phase diagram (Fig. 4) the existence of SiC and
C (Fig. 23). In the bright field image the crystallized matrix can be seen. The elemental
Fig. 23. Bright field image and elemental distribution images of a PVS-derived ceramic after an-
nealing for 50 h at 1500 ꢀC in nitrogen

290
J. Bill et al.
maps reveal that the microstructure consists of crystalline SiC and of carbon segregations.
Compared to the product obtained at 1350 ꢀC the nitrogen content is considerably reduced.
In contrast to the PHMS-derived ceramic nitrogen is removed almost completely dur-
ing the crystallization process. Only small amount of nitrogen can be detected by ESI
investigations.
4. Creep Behaviour
Typical creep curves of PVS- and PHMS-derived ceramics are shown in Figs. 24 and 25,
respectively.
Both materials show similar creep behaviour characterized by a high initial deforma-
tion (up to 2%) and extremely low deformation rates at a later stage of creep (e.g.
1.7 Â 10±±8 s±±1 after 100 h at 1550 ꢀC). However, while the strain rates after 100 h creep
deformation are almost the same, e.g. 1.5 Â 10±±8 and 2 Â 10±±8 s±±1 at 1550 ꢀC for the
PVS- and for the PHMS-derived material, respectively, and reproducible, the difference
in total strain between the two materials is higher and the primary creep deformation at
the beginning of the experiments varies from specimen to specimen. This leads to the
result, that tests at a lower temperature may reveal higher deformation due to an arbi-
trary higher deformation at the beginning of the creep test as e.g. shown in Fig. 24.
The total strain after 4Â105 s creep deformation at 1550 ꢀC and 100 MPa compression
stress is 2.5% for the PHMS-derived material while the deformation of the PVS-derived
material is only 1.4%.
4.1 Primary creep deformation
It is assumed that a part of the extraordinary initial deformation during compression
loading is a stress-independent shrinkage due to the densification of the residual porosity
Fig. 24. Creep curves of the PVS-derived ceramic at 100 MPa compression stress for different tem-
peratures

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
291
Fig. 25. Creep curves of the PHMS-derived ceramic at 100 MPa compression stress for different
temperatures
of the material also observed in stress-free dilatometer measurements [15]. Since the
maximum processing temperature of the precursor-derived materials attained during
pyrolysis is only 1050 ꢀC, it is possible that a sinter shrinkage with a reduction of
porosity takes place at the much higher creep testing temperature. Another explana-
tion for the high primary creep deformation could be a shrinkage due to the densifi-
cation of the amorphous state, also observed during heat treatment of quenched
glasses [27].
The higher deformation of the PHMS-derived materials is maybe due to its higher
porosity (14%) compared to that of the PVS-derived material (4%). The porosity was
determined by optical microscopy and a digital image analysis system.
4.2 Stress dependence of the creep deformation
Assuming power law creep:
ꢀ
ꢁ
Q
n
tꢁc
:
ꢃ5†
_
e ˆ Cs exp
ꢁ
RT
The results reveal that the stress exponent n increases with increasing creep stress.
E.g. the stress exponent of the PHMS-derived ceramic measured after 100 h creep
deformation at 1400 ꢀC increases from n ˆ 0.24 at stresses between 30 and 100 MPa
to n ˆ 1.35 between 100 and 200 MPa [28]. A possible explanation for this beha-
viour is the superposition of a stress dependent and a stress independent deformation
mechanism:
_
_
_
e ˆ eꢃs; T; t† ‡ eꢃT; t† :
ꢃ6†
However, further experiments are necessary to explain this phenomenon.

292
J. Bill et al.
4.3 Influence of crystallization on the temperature dependence of the
creep rates
4.3.1 PHMS-derived material
The creep rate of the PHMS-derived ceramic after 100 h decreases from 3.2 Â 10±±8 s±±1
at 1500 ꢀC to 1.7 Â 10±±8 s±±1 at 1550 ꢀC. While the materials are still X-ray amorphous
Fig. 26. TEM bright field image and elemental distri-
bution images of the material obtained from PHMS
after a creep test at 1550 ꢀC with 100 MPa compression
stress for 4 Â 105 s in air

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
293
after the creep tests at 1550 ꢀC, TEM investigations revealed the formation of nanosized
crystalline Si3N4 and carbon segregations. Parts of the matrix still remained amorphous
Si±C±N phase (Fig. 26).
In some regions of the matrix silicon and oxygen as silicon oxide can be detected.
Elemental distribution images for silicon, carbon, nitrogen and oxygen of this material
show the results very clearly. As can be seen from these images nanosized silicon nitride
crystals grow from the amorphous matrix into the silicon oxide rich parts in this sample.
4.3.2 PVS-derived material
The temperature dependence of the strain rates during creep measured after 100 hours
creep deformation is unusually low. From the creep curves at 1400 and 1500 ꢀC an acti-
vation energy Q of about 250 kJ/mol was calculated for both materials assuming a ther-
Fig. 27. TEM bright field image and elemental distribution images of the material obtained from
PVS after a creep test at 1550 ꢀC with 100 MPa compression stress for 4 Â 105 s in air

294
J. Bill et al.
mally activated process. At 1550 ꢀC the creep rate of the PVS-derived material is only
7% higher than at 1500 ꢀC, which would give an even lower activation energy of only
27 kJ/mol. The reason for this result may be due to different material states. Whereas
after creep tests at 1550 ꢀC the matrix of the PVS-derived material consists mainly of
nanosized Si3N4 grains and segregations of carbon at the grain boundaries beside some
residual amorphous Si±C±N phase and some silicon oxide (Fig. 27) the samples tested
at lower temperatures are much less crystalline or even all-amorphous.
Obviously, crystallization of the material at 1550 ꢀC reduces the creep rate at this
temperature as compared to that of amorphous material tested at lower temperature.
Therefore, comparing the creep curves in the temperature range between 1400 and
1550 ꢀC means dealing with different materials.
4.4 Influence of atmosphere
Owing to the oxygen-containing atmosphere during the creep tests the materials are
sealed by a passivating coating made of silicon oxide [29]. As a consequence, the phase
formation during these investigations leads to microstructures which still contain an
amorphous phase as well as crystalline silicon nitride and carbon segregations. No de-
composition reactions by loss of nitrogen according to Eq. (1) as described in the pre-
vious chapter is observed because this self-sealing effect provides an effective diffusion
barrier that suppresses the loss of nitrogen. Additionally, due to this passivating layer a
remarkable oxidation resistance of the materials is observed.
5. Conclusion
The ceramization of the polysilazanes PHMS and PVS at temperatures up to 1050 ꢀC
leads to the formation of amorphous Si±C±N ceramics. The composition of both cera-
mics is strongly determined by the molecular structure and the composition of the pre-
cursor polymer. In the case of the PVS the vinyl groups polymerize in an early stage of
the pyrolysis (around 350 ꢀC) yielding in a separation of carbon and in a ceramic which
shows finally a higher carbon content compared to PHMS where methyl groups with a
reduced reactivity are split off during the condensation process.
With the disappearance of the Si±C bonds combined with the vinyl group polymeriza-
tion a formation of Si±N bonds takes place resulting finally in tetrahedral configurations of
silicon nitride and leaving behind only a small portion of Si tetrahedra ªmixedº in C and N
if at all. The condensation reaction of PHMS polymer chains on the other side is domi-
nated by the reaction of methyl groups (connected exclusively with Si) among each other
or with H (bonded to N or Si or to C resulting from the methyl±methyl reaction) leaving
behind finally tetrahedral configurations of silicon carbide, silicon nitride and silicon units
with mixed C/N environment as well as trigonal sp2-hybridized carbon units of graphite.
Since the types of short-range ordering observed are in accordance with the structural
units of the thermodynamically stable (crystalline) phases the system obviously mini-
mizes its free energy during pyrolysis equivalent to the stable phase equilibria of the
ternary system Si±C±N indicating that the free energy of the amorphous Si±C±N phase
also has minima located in the same composition ranges as that of the crystalline
phases. In consequence, during ceramization of PVS the composition approaches the tie
line Si3N4 ‡ C and that of PHMS the three-phase field of Si3N4 ‡ SiC ‡ C.

Structure Analysis and Properties of Si±C±N Ceramics from Polysilazanes
295
This result gives a simple explanation for the phase separation observed experimen-
tally in as-pyrolyzed PHMS-derived material and in material similar in composition to
the PVS-derived material reported elsewhere [30]. Following this explanation conse-
quently one can predict that small angle scattering experiments with PVS-derived mate-
rial in the as-pyrolyzed stage should also reveal a heterogeneous material consisting of
amorphous Si3N4 and amorphous C.
As with crystallization of the heterogeneous amorphous materials the stable phases
are formed, the ceramization of PHMS and PVS provides a means to build up silicon
nitride/carbide-based composite ceramics from the corresponding structural units on an
atomic scale. Moreover, certain amounts of carbon can be incorporated into the materi-
als depending on the type of polymer and the composition of the ceramics.
Owing to the different architecture of these ternary Si±C±N ceramics their properties
differ. In this connection the electrical conductivity of the PHMS- and PVS-derived cera-
mics can be seen in Fig. 28.
These measurements provide a sensitive probe for the different structural arrange-
ments and compositions of the two materials. The studies reveal that the PVS-based
material which contains a higher amount of sp2-hybridized carbon compared to the
PHMS-based material also exhibits higher values of the electrical conductivity [13, 16].
It is interesting to note that both ceramics exhibit semiconducting properties.
The results from the preliminary creep studies reveal an interesting potential of creep
resistance at temperatures at least as high as 1550 ꢀC since the creep rates after an
initial stage of substantial deformation are quite reproducible and become very small
(e.g. 1.5 Â 10±±8 s±±1). Even after 100 h of creep the deformation rate was still decreasing
indicating that creep still has not approached secondary creep. Furthermore, the creep
rates in this stage were quite reproducible although the materials investigated were still
poorly developed and e.g. differed substantially in porosity.
Fig. 28. Electrical conductivity of the amorphous silicon carbonitride ceramics obtained by pyroly-
sis of the polysilazanes PVS and PHMS at 1050 ꢀC
20 physica (a) 166/1

296
J. Bill et al.: Structure Analysis and Properties of Si±C±N Ceramics
The reason for the substantial creep resistance of precursor-derived ceramics is prob-
ably due to the lack of grain boundaries and a rather high viscosity, i.e. a very low
diffusion coefficient in the predominantly covalent bonded amorphous material.
In the light of this potential of creep resistance it seems to be worthwhile to investi-
gate the initial creep deformation in more detail and to find out whether the extraordi-
nary initial deformation can be decreased by a decrease of porosity.
Acknowledgements The authors thank the Keramikverbund Karlsruhe Stuttgart
(KKS), the KSB foundation, Stuttgart, the Japan Science and Technology Corporation
(JST), the New Energy and Industrial Technology Development Organization (Japan),
and the Deutsche Forschungsgemeinschaft for financial support.
References
[1] F. F. Lange, J. Amer. Ceram. Soc. 56, 445 (1973).
[2] P. Greil, G. Petzow, and H. Tanaka, Ceram. Int. 13, 19 (1987).
[3] F. F. Lange, J. Amer. Ceram. Soc. 57, 84 (1974).
[4] R. W. Rice, Amer. Ceram. Soc. Bull. 62, 916 (1983).
[5] D. Seyferth and G. H. Wiseman, J. Amer. Ceram. Soc. 67, C-132 (1984).

[6] M. Peuckert, T. Vaahs, and M. Bruck, Adv. Mater. 2, 398 (1990).
[7] J. Bill and F. Aldinger, Adv. Mater. 7, 775 (1995).
[8] H.-P. Baldus, O. Wagner, and M. Jansen, Mater. Res. Soc. Symp. Proc. 271, 821 (1992).
[9] J. Seitz and J. Bill, J. Mater. Sci. Lett. 15, 391 (1996).
[10] A. Jalowiecki, J. Bill, M. Friess, J. Mayer, F. Aldinger, and R. Riedel, Nanostruct.
Mater. 6, 279 (1995).
[11] J. Seitz, J. Bill, N. Egger, and F. Aldinger, J. Europ. Ceram. Soc. 16, 885 (1996).

[12] S. Schempp, J. Durr, P. Lamparter, J. Bill, and F. Aldinger, submitted to Z. Naturf.
[13] J. Seitz, PhD Thesis, Stuttgart 1996.
[14] M. Friess, J. Bill, F. Aldinger, D.V. Szabo, and R. Riedel, Key Engng. Mater. 89/91,
95 (1994).
[15] T. Nishimura, R. Haug, J. Bill, G. Thurn, and F. Aldinger, submitted to J. Mater. Sci.
[16] U. Eins, J. Bill, and F. Aldinger, Tagungsbericht der Werkstoffwoche 96, Symp. 7, Stutt-
gart 1996.
[17] G. Thurn and F. Aldinger, Proc. Symp. Grain Boundary Dynamics of Precursor-Derived
Covalent Ceramics, Ringberg, 1996, to be published.
[18] Product information VT 50, Hoechst AG, Frankfurt.

[19] M. Ruhle, J. Mayer, J. C. H. Spence, J. Bihr, W. Probst, and E. Weimer, Proc. 49th
EMSA-Meeting, 1991 (p. 706).
[20] L. Reimer, I. Fromm, and R. Rennekamp, Ultramicroscopy 24, 339 (1988).
[21] J. Mayer, Microscopy and Analysis 25, 21 (1993).
[22] A. Berger, J. Mayer, and H. Kohl, Ultramicroscopy 55, 101 (1994).
[23] J. Weiss, H. L. Lukas, J. Lorenz, G. Petzow, and H. Krieg, Calphad 5, 125 (1981).
[24] G. Socrates, Infrared Characteristic Group Frequencies, 2nd ed., Wiley, New York 1994.
[25] E. Breitmaier and W. Voelter, Carbon-13 NMR Spectroscopy, 3rd ed., VCH, Weinheim/
New York 1987.
[26] M. R. J. Gibbs, J. E. Evetts, and J. A. Leake, J. Mater. Sci. 18, 278 (1983).
[27] H. Salmang, Die physikalischen und chemischen Grundlagen der Glasfabrikation, Springer-

Verlag, Berlin/Gottingen/Heidelberg 1957.

[28] B. Z. Janos, Diplomarbeit, Universitat Stuttgart, 1997.
[29] J. Bill and F. Aldinger, Z. Metallk. 87, 827 (1996).

[30] J. Durr, P. Lamparter, J. Bill, S. Steeb, and F. Aldinger, submitted to J. Non-Crystall.
Solids.