JOURNAL OF MATERIALS SCIENCE 25 II99O) 3886-3893

Structural evolutionsfrom polycarbosilaneto

SiG ceramic
G . D . S O R A R U -F, L O R E N CBEA B O N N E A UJl ., D . M A C K E N Z I E
Departmentof MaterialsScienceand Engineering,Universityof California,Los Angeles,
California90024. USA
into a microcrystalline
The pyrolysisprocessof a polycarbosilane
silicon carbideceramichas
2sSi
beenfollowedup to 1700"C mainlyby meansof solidstate
and 13Cnuclearmagnetic
resonance,transmissionelectronmicroscopyand X-ray diffractionanalysis.A structuralmodel
has been proposedfor the amorphoussilicon carbidephasethat is formedduring the pyrolysis
process.
is formedby a mixtureof p-SiC and a-SiC;
The ceramicobtainedat high temperature
however,some difficultiesin the identificationof the crystallinephaseshave been pointed out.

was usedin this study. All the firing treatmentswere

performed in flowing argon with a heating rate of
2'C min I up to the completeremovalof the organic
componentsat around8400C. The resultinginorganic
solid, that is amorphous according to XRD and
TEM/SAED experiments,can be considered as
the ACC phase precursorfor SiC microcrystalline
ceramics.
In orderto studythe subsequent
densification
and crystallization process, the ACC phase was
fired at different temperaturesup to 1700'C al
l0'Cmin I. The amountsof silicon,carbon and
hydrogenwereanalysedfor selectedsamples.Oxygen
contentwasnot analysed.Thermogravimetric
analysis
wasperformedin flowing argon usinga Perkin Elmer
equipment.2esi,'rC and 'H liquid NMR spectrawere
recordedon a AM 360Brukerspectrometer
al.'71.5,
90.5 and 360MHz, respectively.
The polymer was
dissolvedin CDC!. For "Si NMR experiments,
a
pulsewidthof l0 p secwas applied with a relaxation
delayof 6 sec. H and rrCNMR spectrawererecorded
w i t hp u l s e w i d t hosf 2 4 s e cl o r r H a n d 5 g s e cf o r | C .
and delaysbetweenpulsesof I secfor I H and 2 secfor
'rC. Solid-state'esi and r3C NMR spectra were
obtainedon a MSL 300 Bruker spectrometerat 59.6
and 75.5MHz. A pulsewidthof 2.5p secand a delay
betweenpulsesof 60sec were used for the zosi
MAS-NMR spectra.A contacttime of 2msec was
r'C CP
appliedfor thecross-polarization
experimenrs.
MAS-NMR spectrawererecordedwith a contacttime
of 3 m sec.Tetramethylsilane(TMS) was used as a
referencefor all the NMR data. ESR expedments
were carried out on a Varian E09 spectrometer.A
Bruker gaussmeter
wasusedto measurethe magnetic
field with diphenylpicrylhydrazyl(DPPH) as a standard. The numberof spinswas estimatedby comparison with copper sulphate as reference.XRD
pattems were recordedon a Philips diffractometer
2. Experimental procedure
usinga Cu,(c radiationwith a nickel filter. For TEM
Commercially available polycarbosilane(PC, Dow
CorningX9-6348)with a molecularweightof 1400, observationsthe sampleswere ground to very fine

1. lntroduction
Recently,it has beenshown that non-oxideceramics
suchascarbidesand nitridescan be obtainedby firing
suitablemetal-organicpolymer precursorsin controlled atmospheresn]. A common feature of the
polymerroute to carbidesor nitridesis the formation
of intermediateswhich are amorphoussolids.These
are formed after the removal of the organic components and before crystallization [2]. We have
alreadypointedout the importanceof regardingthese
amorphouscovalentceramics(ACC) as an entirely
new family of disorderedsolidswhosestructureand
propertiesare worthy of independentstudy [3, 4].
Among the various systemsalready synthesized,
SiC obtained from polycarbosilaneis certainly the
most widely studied [5-7]. Commercially available
Nicalon SiC fibresare obtainedfrom polycarbosilane
following the processfirst developedby Yajrma et al.
[8]. The feasibilityof this method in producingfibres
or coatings is not restrictedto this material. It is
mainly due to the polymeric nature of the ceramic
precursors,and it accountsfor the many research
effortsthat are currentlybeingconducted.Despitethe
increasingnumberofstudieson SiC fibresobtainedby
the Yajima process,the structuralconversionfrom the
starting polycarbosilaneto the resulting ceramic
matedal is not well understood.In this paper the
pyrolysismechanismof polycarbosilane,
the structure
ofthe intermediateACC phaseand its conversioninto
the microcrystallinesilicon carbideceramics,will be
discussedmainly basedon 2eSiand r3Cmagic-angle
(MAS-NMR),
spinningnuclearmagneticresonance
electron spin resonance(ESR), X-ray diffraction
(XRD) and transmissionelectronmicroscopy/selected
areaelectrondiffraction (TEM/SAED) experiments.

* Permanentaddress.Dipartimentodi Ingegneria,Univrsitdi Trento, 38050Mesiano,Trento, Italia.

I Permanentaddress:Chimiede la MatireCondense.
Univelsil Paris6. Tour 54. 4 olaceJussieu.75005Paris.France.

3886

nd Hall Ltd.
0022-2461/q0
$01.00+ .12 O 1990Chapman

ffi

50

-50

Chemical shift {g.o.m.}

Frgrre 1 (a) ?'SiMAS-NMR and O),,Si Cp MAS-NMR sr,ctraof
polycarbosilane.

powders which were then dispersedon to a TEM

copper grid using an eye dropper. TEM/SAED
investigationswere performed using a Jeol STEM
100 CX equipment. Density measurementswere
performed on fine powders by immersion in CClo
following the Archimedesmethod. BET measurements \ryeredone with a Flow Sorb II 2300 Micromeriticsequipment.

8.0

4.0

0.0

3. Results
3.1. Characterization
of the polymer
orecursor
The chernicalanalysisof PC, reported in Table I,
s h o w sa S i :C : H r a t i o o f l : 2 . 2 : 5 . b s i M A S - N M R
and 2esiCP MAS-NMR spectrahave beenrecorded
on this starting material (Fig. l). The two spectra
revealthe samefeatures,with an enhancement
of the
resolution for the cross-polarizedspectrum. The
spectra show two different silicon units, already
reportedin the literature[7]. The peakat - 0.8p.p.m.
is due to silicon atoms bondedto four carbon atoms
(SiCa) as type I units and the second peak at
- 17.6p.p.m.is due to type II units in which the
silicon atoms are surroundedby three carbon atoms
and one hydrogenatom (SiCrH). No distinct peaks
appeararound - 35p.p.m: Si-Si bonds,if they exist,
are thus not abundant in this starting PC [7]. These
two typesof units are shownbelow.
CH'

I
CH"_Si_CH,
- l
CH,

(r)

cH.,

CH"-si-cH,
I
H

OD

TABLE I Atomic ratio Si:C:H in the preculsorand in some

fired samples
Sample

PC
PC8,l0
PCl200
PCt500

2.2
1.6
t.44
1.43

5
0.65
0.10
0.07

ffi
Chemicalshift ( p.p.m.l
F8rle 2"Si,rH andr3CNMR spectraof polycarbosilane
dissolved
in CDClr.

PC wasdissolved
in CDClr. 2esi,I H and ,,C NMR
spectrawererecordedin solution(Fig.2). The,,Si
NMR spectrumis quite similarto the CP MAS-NMR
spectrumon th powder. The ratio betweenthe two
kinds of units (I and II) appearsto be I : 0.8. The 'H
NMR spectrumof PC shows two regions for the
resonance,
from 4 to 5p.p.m. due to Si-H bonds and
around0 p.p.m.due to C-H bonds.The integrationof
the peaksgivesa valueof 11 for the C-H/Si-H ratio.
In the Si-H region,severalpeaksarepresentat 4.1,4.3
and 4.6p.p.m. due to differentSi-H sites.The C-H
regionshowsa mainpeakat 0.17p.p.m.,due to CHj
groups,and two shoulderswith lower chemicalshift
values,at 0 and 0.5p.p.m. assigned,respectively,to
CH, and CH groups.The '3C-{rH} spectrumof the
solution shows a broad peak centred on 3 p.p.m.
correspondingto the aliphatic carbon atoms present
in PC in CH3, CH, and CH units. Somesharppeaks
are superimposed,
certainlydue to somequite mobile
units insidethe polymer.
All the NMR spectraindeedhavebroad peaks,and

3887

Fgurc3 TEM bright-field micrograph with SAED

pattemsof polycarbosilane
pyrolysedat 8,10"C.

no resolutionwasobtainedespecially
for rH and r3C
NMR experiments.
Thisseems
to indicatethata large
distributionof unitsis presentin thispolycarbosilane.
The structureof polycarbosilane
that emerges
from
theseresultsis morecomplicated
than a simpleIinear
chainof type I and II unitswith a I : I ratio suchas
CH.
t

CH
t

-cH, si cH, si-cH,-

CH..

In this casethe chemicalanalysisshouldbe Si : C : H :

I :2.5:7. The low carbon and hydrogen contents in
the studiedPC and the presenceof CH units in the ' H
NMR spectrum suggestthat some cross-linknghas
already occurred between the chains. This has already
been suggestedby Okamura et a/. [9].

3.2. Characterizalion
of he pyrolysisprocess
prccursor
3.2.1. From the
polymer to the
ACC phase
The low molecularweightcomponentsof the polycarbosilanehave been previouslyremovedby melting the polymer in flowing nitrogengas. Thermogravmetricanalysis(TGA) performedon this material
(PCD) showedthat the weight lossesend at around
800'C.At thistemperature
polymerhas
theprecursor
beenconvertedinto an inorganicsolid that appea$
to be amorphousby XRD and TEM/SAED investigations(Fig. 3).
According to TGA experiments,the pyrolysis
processleading to the formation of the amorphous
slcon carbide phase,consistsof two stages:from
300 to 500'C with a weightlossof l37o and from
500 to 800'C with a further weight loss of 1270.
The PCD was heatedat 2'Cmin ' in argon up
to 500"c (PC500), 700'c (PC700) and 840.C
(PC840).The chemicalanalysisof PC840appearedto
b e S i : C : H : l : 1 . 6 : 0 . 6 5( T a b l eI ) . E x c e scsa r b o n

was thus presentin this amorphoussiliconcarbide

phase.
The "Si MAS-NMR and r3C CP MAS-NMR
spectrarecordedon thesesamplesare shownrn Fig. 4
aswell asthe specraofthe precursorasreference.
The
"Si MAS-NMR spectrumof PC showsthetwo peaks
assigned
to SiCoand SiCrH units.[n the samplefred
at 500'C, the peak at
l6p.p.m. assignedto the
SiC,H units has disappeared.
The major peak at
0 p.p.m.dueto SiCaunitsappearsalmostunchanged
with onlya slightshiftofthe positionofrhe maximum
and a small increaseof the linewidth (Table II). By
increasingthe firing temperatureover 500oC and up
to 840'C thistrendis maintained:
the peakrelatedto
the SiCaunits is continuouslymovingtoward lower
values of chemical shift typical of the crystalline
siliconcarbidephasewhile ts Iine',vidthis increasing
up to 700'C. The evolutionoi rhe ' esi MAS-NMR
spectrahas alreadybeen explainedin detail in a
previousstudy [3]. In the first stageof the pyrolysis
process!up to 500"C, a possible reaction is the
consumptionof Si-H groupsand the formationof
bridging Si-C bonds betweenthe polycarbosilane
chains.Above this temperaturethe shift of the peak
dueto SiCounitsreflectsan increase
ofthe connectivity
29silaes-rulrR

13ccp uls-t'tl,,ta

TABLE II "Si MAS-NMR data for the precursorand rhe

fircd samples
Sample

Chemicalshift (p.p.m.)

Linewidth (p.p.m.)

PC
PC500
PCTOO
PC84O

0
- 4.5
-6.3
-8.0

l0
12.5

3888

16

*ito---*-t*

-Zo-:n-*-

Chemicalshift (p.p,m
)
FErre 4 Evolutionwith the firing temperatureof DSi MAS-NMR
and ''C CP MAS-NMR spectraof polycarbosilane.

'.

r'-'.1

u- l.,.tn\,
r..,2"',.-lr
i,
St
,L.
I
,').,-'./"'t.-i.
I .r'"

,"/

-c' |
sl-"\ ,
I
|
-sr-c\
/
si-l
I
)sr..
r,-cr
/

-cH-t.

i'

cfl7.r

i,-

./
si-uc/
/ \

_sf\

Fr,g/fe Proposed
structure
phase
oftheamorphous
silconcarbide
obtained
from polycarbosilane-

FurJ ESRspectrum
recorded
at roomtemperature
on PC840of the network by the formation of Si C-Si bonds.
At the same time the evolution of the linewidth is
correlated to an increase of the disorder of the local
environment around the silicon atoms. During the
pyrolysis process, condensation reactions lead to
the consumption of CH., and CH, groups and to the
formation of CH or C units with the evolution of H,
or CH" [7]. These reactons obviously increasethe
number of different SiC, units that can exist in the
material and may account for the observed increaseof
the peak linewidth already mentioned. "Si MAS-NMR
study of the fired samples shows no other peaks than
thoseassignedto SiCaunits. No Si-Si nor Si O bonds
seemto be ormed during the pyrolysis process.
Information about the evolution of the local
environment of the carbon sites during the pyrolysis
processcan be obtained from the rrC CP MAS-NMR
spectrain Fig. 4. The spectrumof PC showsone peak
at 4.2 p.p.m. and a small absorption near 55 p.p.m.
that has been identified as a spinning side band. By
increasingthe firing temperaturethe main peak shifts
toward higher values of chemical shift with a corresponding increase of its line\,/idth. The resonance
presentat 4.2p.p.m. in the spectrumof PC is due to
t h e c o n t r i b u t i o no f a l l t h e c a r b o n g r o u p sp r e s e n ri n
the starting polymer, namely CH,, CH, and CH. It is
wel known that the chemical shifts of CH,, units
increasewhen the r value is decreased.
Then, according to thesedata, the shift of the peak with the firing
temperature reflects a consumption of the more
hydrogenated species due to the occurrence of the
condensationreactionsduring the pyrolysis process.
The ncreasein the linewidth has to be relatedwith the
same evolution in the resi MAS-NMR spectra. It
revealsan increasein the disorderofthe Iocaenvironment ofthe aliphatic carbon atoms during the pyrolysis
process.Moreover, an interestingfeature appearsat
above 700'C: new peaks are present in the 100 to
200p.p.m. range.Thesepeaksbecomemore intenseat
840'C. The broad peak centred around 135p.p.m.,
can be assignedto the presenceof aromatic carbon
atoms. and could be relatedto the formation of C:C
bonds. It has already beensuggestedthat such bonds
could be presentin the intermediateamorphousphase
but no experimentalevidencewas given.They should
be precursorbonds for graphiticcarbon that is formed
around 1200'C accordingto Raman data [9].
ESR was perormed on the sample fired at 840" C

(Fig. 5). The spectrumexhibitsa singleisotropic signal

centredat g : 2.0030,with a linewidth of 3.9 G. This
valuecorrespondsto carbon danglingbonds [0]. The
integration of the signal gave the number of defect:
to be 2.5 x lOrecm I, considering a densit,v o
2.2gcm-l for this sample.This value is in agreemenL
with the number of defectsfound in amorphous SiC, .
samplesprepared from chemical vapour deposition

ll0l.
3.2.2. Characterizationof the ACC phase
The proposed structure of the amorphous silicon
carbide phase obtained at 840'C is illustrated in
Fig. 6. It is basedmainly on rhe 2esiMAS-NMR and
rC CP MAS-NMR resultsalthough
the broadnessof
the NMR peaks in thesesamplesmade it diflculr to
obtain precise structural data. These amorphous
phasesshow a wide distribution of siliconand carbon
atoms types. However, some relevant features can be
pointed out. (a) The silicon carbide phase is not
stoichiometric,but an excessof carbon is present
(C/Si : 1.6). (b) All the silicon atoms seem to be
bonded to four carbon atoms and Si Si or Si-H
bonds,if they are presentshould be minimal. (c) Some
C:C bonds are presentasclearlyshown by the rrC CP
MAS-NMR experiments.(d) The residual hydrogen
content(H/Si : 0.65)should be mainly presentin the
structure as CH groups as suggestedby the chemical
shift of the main peak in the rrc Cp MAS-NMR
spectra. (e) The condensation reactions occurdng
during the pyrolysisprocesscan lead to the formation
of six-memberrings like those presentin crystalline
SiC, but the formation ofdistorted five- or seven-aton
rings cannot be ruled out and should lead to the
presence of C{ bonds. (f) The presence of paramagnetic defectshas been shown by ESR experiments
and these defects were assigned to carbon dangling
bonds.
A density of 2.21gcm r has been measuredon a
sample of fine powders of this phase. This value is
lower than the density of p-SiC (3.21gcm r). A
theoretical density (2.70gcm r) can be estimated
from the chemicalanalysisofPC840 (Table I) by using
the rule of mixtures and assumingthat all the silicon
atoms are engagedin forming amorphous SiC (density
3.0gcm r) [tl]while the remainingcarbon is present
as graphite(density2.2gcm 3;. The low value of the
densityof the amorphousphasecannot be ascribedto
the presenceof porosity in the material. SEM investigations (Cambridge Sterescan)have shown only few

3889

10000c
^

2.4
9500c

9000c

- 2.3

10

12

Time (h)
Figwe 7 Etoltion of the density,p, of the ACC phasewith the
lring time.

isolated pores for the ACC phase while TEM studies

(Fig. 3) have shown no porosity at a sub-microscopic
level- This evidence is also supported by BET analysis
that gave a value of surfacearea lower than I m' g- |
lor the same sample.Therefore, a quite open structure,
with a large amount of free volume must be invoked
to account for the low value of density of the amorphous ceramics. With the aim of following the
evolution of the density of PC840 as a function of
the firing time, isohermal treatments in an argon
atmosphereat 900, 950 and 1000'C were performed
and the resultsreportedin Fig. 7. An activationenergy
of 82 kcal mol I was obtained from the initial slopeof
the densilcationcurves(Fig. 8). Assuminga negligible
value ol porosity in PC840, the observedincreaseof
density during the firing treatmentmust be primarily
ascribedto a reduction of its free volume. Two main
processesmay account for this effect. (i) The progress
of the condensationreactions between residual CH
groups in the structurewith the eliminationofH, and
CHo and the formation of new Si-C Si bridges with a
consequentincreaseofthe crosslinkingofthe network.
Actually, chemicalanalysisof the ACC phasefired at
950 and 1000"C for the longest times, showed a
decreaseof the hydrogen down to H/Si : 0.2 compared to the initial value of H/Si : 0.65; (ii) a rearrangement of the open amorphous covalent structure
toward more compact configurations with no change
in chemical composition. For both of these mechanisrns,a key step should be the cleavageof chemical
bonds. either C-H (99 kcal mol r) in the first caseor
Si-C (T6kcalmol ') and C C (82kcalmol r) [2] in
-1

4."
f= 82 kcolmot-r

4.Q

4.1

the latter one.The obtainedvalueofactivation energy

suggests
that, at leastin the initial stages,the densification processoccurs via the cleavageof chemical
bondspresentin the material.On the ground of 'esi
MAS-NMR experiments,
it hasalreadybeenreported
in a previous study [3] that, during the isothermal
treatments,the rarrangementof the structureleads
to an ordering of the local environmentof the
siliconatoms.Moreover,at the highesttemperature
(1000'C), XRD and TEM/SAED investigations
showeda concomitantreorganizationof the network
also in the medium rangewith the forrnation of SiC
microcrystals
[3].
3.2.3. From the ACC Dhase to
microcrystalline SiC
The amorphoussiliconcarbidephasecan be converted
into a microcrystallineceramic by firing it at high
temperatures.
In order to follow suchtransformation,
thePC840washeated2!t10'C min-r in argonflow at
1000'c(PCl000),1200"
c (PCl200),1500.c(pcls00)
(PCl700).
and 1700'C
XRD and "Si MAS-NMR
spectra recorded on these samplesare shown in
Figs 9a and b, respectively,togetherwith the spectra
obtainedon a commericalB-SiCas reference.
In the XRD patterns(Fig.9a), broad peaks,correspondingto crystallineSiC phase,start to appearat
1000'C and sharpenby increasingthe firing temperature. The correspondngcrystallite sizes,evaluated
from the diffraction resultsby usinga peak broadening procedure,are reportedin Table III. The microstructureof the sampleheatedat 15000
C, as revealed
by TEM investigations,
is shownin Fig. 10.Although
a detailedanalysisof the crystalsizesfrom TEM
micrographswasnot performed,the meancrystalsize
seems
to be slightlyhighercomparedto XRD results.
This could be due to the fact that the crystalliteshave
a rangeof sizedistributionin the sampleand the
largest particles are more readily observed. Fine
porosity,as revealedby the white spotsin the micrograph, seemsto be present in this sample. This
observationis in agreementwith a recentstudy [ 3]
that showedthe formation of porosityin Nicalon
fibresafterannealingat 14000
C in argonatmospheres.
BET analysisresultedin a valueoi surfacearealower
than I m' g I indicatingthat closedporosityis present
in thesesamples.
The position of the diffraction lines in the fired
phaseis
samplesindicatesthat the microcrystalline
mainly B-SiC.In the XRD pattern of the sample
heatedup to 1700"Ca small shoulderis seenat
around20 : 34".It hasbeenassigned
to a-SiC,suggesting that, at the highest temperaturesthe
phaseconsists
crystalline
of a mixtureof cubicp-SiC
with tracesof the hexagonalform
TABLE III Sizesof the SiC microcrystals
n thc fired polycarbosilanecalculatedfrom the broadnessof the main peak in the
diffractionpattern
Firing temperature("C)

4.2

1000

1041Rr
Fre 8 Arrheniusplot for the inital densificalionrate.

3890

Crystalsize(nm)

t 500

1700
16.0

Figule 9 Evolution of (a) XRD patterns and (b) "Si

MAS-NMR spectra of ACC phase during the firing
process(P-SiCsamplprovidedby Superiorcraphite).

,[

\--./q*JL_L

Pcl7oo

PC1500

\"*-/"..***--".-

PC1200

PC1000
\.-l\.*....-...,-

\-_-.r,r.*.,*.,.*"40

Pc84o

"t*"*,/

200 100

20 (des
)

Chemicalshift (p.p.m.)

In the X-ray patterns of the fired samples a small

peak around 20 : 26'is clearly visible.Its intensity
reaches the maximum value in the sample heated at
1200'C and decreaseswith further heating. lt could be
assignedeitherto the (l 0 l) reflectionof(-quartz or to
the (002) line of carbon. In the literature, the disappearanceofthis peak after a treatment with HF has
been reported [7]. This result has been assumed to be
proof for the existenceof crystalline silica in the fired
ceramics. However, in the present study, the samples
were heatcd in an inert atmosphere to avoid major
oxygen contamination; moreover SiOo units should
give rise ro a peak in the "Si MAS-NMR spectra
around - I l0 p.p.m. Sucha peak is completelyabsent
in the spectra. Therefore it seemsthat the X-ray diffraction peak af 20 : 26" should be assigned to the
presenceof small clusters of graphite rather than
crystallitesof d-quartz.
The evolution with the firing temperatureofthe' eSi
MAS-NMR spectra is shown in Fig. 9b. The main
peak correspondingto the SiC4units is shifting downfield approaching the value corresponding to the
crystalline form (Table IV). At the same time, irs
linewidth is decreasing suggesting an ordering of the
ocal environment of the silicon atoms in the SiCr
units. At 1500"C some structuresstart to appear in
this peak that becomemore evidentat 1700"C. At this
temperature,the MAS-NMR spectrum revealsthree

d i s t i n cpt e a k sa t - 1 6 . 2 , - 2 0 a n d - 2 5 p . p . m . S u c h a
spectrumhas alreadybeenpublishedin the literature
the main peakat
[4, l5]. In a first approximation,
16.2p.p.m.can be assigned
to B-SiCand the two
minor peakscould be due to somea phase.However,
this assignmentwill be discussedlater in the discussion.
4. Discussion and conclusion
In the pyrolysisof PC, the removal of organic components occur via condensationreactions beween
CH.,and CH, groupsof the strting polymer.When
thisprocessis completethepolymerhasbeenconverted
into an amorphouscovalentceramic(ACC) phase.
The temperatureat which the condensationreactions
end and the ACC is formed can be obtainedfrom a
TGA experiment:it can be definedas the temperature
at which the weightlossesare completeand the curve
approaches
a constantweightvalue.However,this is

TABLE IV 2'SiMAS-NMRdataof thefiredoolvcarbosilane

PC840 PC1000 PCr200 PCl500 PCt700
Chemical
shift (p.p.m.)

l6

- t6.2
20

- 16.2
-20
FBrre 10 TEM bright-fildmicrographof PC frredat 1500'C.

3891

not an absolutethresholdvalue:it is reasonable

to
think that the temperature
of formationof the ACC
phase is dependentupon the heating rate and the
heatingatmosphere.For example,by processingthe
PC in vacuum or by using a lower heating rate it
shouldbe possibleto completethe pyrolysisprocessat
lowertemperatureand thusit shouldbe possibleto get
an amorphousinorganic solid at lower temperature.
Moreover,the structureitself and the compositionof
the forming inorganicdisorderedphasecan be affected
parameters
by the sameprocessing
i.e. temperature,
heatingrate and heatingatmosphere.
In the presentstudy, TGA experimentsobtained
usinga heatingrate of 2"Cmin I in flowingargon,
showedthat the weight lossesend at around 800oC.
Thus according to these results the PC has been
convertedinto the ACC phase for the subsequent
structuraland crystallizationstudiesby heatingit with
the same processingparameters at temperatures
slightlyhigherthen 800'C, namely840'C.
A wide tange of differenttypesof defectseemsto
existin thisamorphous
siliconcarbidephaseasshown
in Fig. 6. NMR and ESR experimentsshowedthe
presenceof C:C bonds and carbon dangling bonds
respectively.The presenceof distortedfive- or sevenatom rings can be reasonablyassumed[3]; excess
carbonand residualhydrogen(C/Si : 1.6;H/Si :
0.65)havebeenevincedby chemicalanalysis.Chemical
and structural modificationsoccur in this phaseby
increasingthe temperatureover 840'C. Residual
hydrogencontentand carbon excessare considerably
(TableI) at 1200'C(H/Si : 0.1;C/Si : 1.44)
reduced
and are still decreasingat l500oc (H/Si:0.07;
C/Si : 1.43)due mainly to the completionof condensation reactions.The other major modification
occurringduringthe firingprocess
of the ACC phase
is its structural rearrangementthat leads to the
formation of a microcrystallineceramic.This transformatonstarts,accordingto XRD andTEM/SAED
experiments,
at around1000'C.It is known[16]that
the crystallizationmechanismof disorderedcovalent
four-coordinatedmaterialslike amorphoussilicon or
germaniuminvolvesthe rupture ofthe Si-Si or Ge-Ge
bonds,respectively.In thesecasesthe experimentally
observedactivationenergiesare closeto the covalent
bond energiesfor both silicon !7] and germanium
[8]. In the presentcase,the crystallizationmechanism
ofthe amorphoussiliconcarbidephaseshouldbe more
complicated becauseit occurs together with the
mentionedchemicalmodificationof the system.The
crystallizationof the ACC phaseresultsin an increase
of its density. Kinetic studies of the densification
processin the early stagesof crystallization,below
1000'C,gavean activationenergycloseto the energies
of the Si-C and C-C bonds.However,it is not possible
to regard this value as the activation energy lor
crystallizationbecausethe observeddensificationis
due not only to crystallizationbut also to the conprocess.
comitantcompletionof the condensation
At 1200'C crystallitesof SiC with dimensionof
2.5nm are presentin the material(Table III); brighr
fieldTEM observationswerestill featureless
like those
obtained on PC840. However, the diffraction rings
3892

80

40

-40

-B0

Chemicalshift (p.p.m.)
Fgrle / 1 Coparison betweenthe zesiMAS-NMR spectraof the
samplelired at 1700'Cand a commericall-SiC (Supriorcraphite).

were quite narrow, confirming the microcrystalline

nature of this sample.As suggestedby X-ray diffractionpattems,clustersof graphiteare presentin
the material.As alreadyreported[9] they should
be presentat the edgeof SiC microcrystalsand could
play an important role preventingor slowingthe rate
of crystalgrowth.
A t 1 5 0 0a n d 1 7 0 0 " Ct h ec r l s t a ls i z ei n c r e a suep L o
(TableIll). As reportedin
8 and 16nm,respectively
the literature[7] this processshouldbe connectedwith
an evolution of CO from the system.In the present
case,however,
ifthis reactionakesplace,it, shouldbe
minimal concerningonly the oxygen presentin the
systemas impurities:indeed,in the firing processof
PC care has beentaken to avoid major oxygencontaminaion.
Thedensityofthe samplefiredat 1500o
C approaches
L h e v a l u e o f 2 . 7 g c m - ' . T h e t h c o r e t i c adl e n s i t l .
calculatedas previouslydescribedfrom the chemical
analysisand usingin this casethe densityof B-SiC
( 3 . 2 g c m ' ) . i s 2 . 9 g c m ' . T h ed i f f e r e n cbee t w e etnh e
two valuescan be ascribedto a possiblecloseporosity
presentin the sampleasrevealedby TEM observations.
A valueol 77o of porosityshouldaccountfor the
observeddifferencein the densityvalues.
Accordingto XRD analysisB-SiCseemsto be the
principal crystallinephasein the samplesfired at the
highesttemperaturestogetherwith small amountsof
the hexagonalform. The presence
ofa-SiC in Nicalon
fibresheatedat temperaureshigherthan 1400'C has
alreadybeenreportedin the literature[13].However,
due to the broadnessof the diffraction peaks, a
definitiveassignmentseemsdificult. Many different
polytypesofthe hexagonalphaseare known, differing
from eachother only in the stackingsequenceof the
siliconand carbonlayers[20].
"Si MAS-NMR has beensuccesslully
appliedto
distinguish between the different SiC polytypes
-"Si
4 . 2 1 . 1I .n F i g . l l a c o m p a r i s o bn e r w e e n
lhe
MAS-NMR spectrumof the samplefired at 1700'C
and that recordedon commericalP-SiC is reported.
The NMR spectrum of SiC from polycarbosilane
shows three distinct peaks at -16.2, -20 and

TABLE

V Chemicalshifts of siliconcarbidepolytypes

Sample

Chemicalshift
(p.p.m.,

Reference

,-sic,3c

- 18.3
- 18.9

I2r]

a-SiC,6H

- 13.9)

-"-|
d-SiC,l5R

, r4.e
- 20.81

siliconsiteswith slight differencesin the Si-C bondlengths.This is only an assumption,

and the identification of the crystalline phase formed during the
pyrolysisprocess
ofPC seems
worthy offurther studies.

Present
[2t

12u

24.4)

- 25p.p.m. whereas the commerical one displays

just one peakat l9 p.p.m.A detailedNMR study
of silicon carbide polytypes has been published
by Hartman ef al. l2ll. Among the large number
of known polytypes,it seemsthat only four typesof
siliconenvironmentexist,designated
A, B, C and D
by theseauthors.The cubicsiliconcarbide(B phaseor
3C polytype) has only type A silicon sites,while the
H or 15R polytypesexhibit three silicon sitesA, B
and C in relativeintensitiesI : I : I and I : 2 : 2, respectively. The chemical shifts for these polytypes are
reportedin Table V. The fype D siteis more unusual:
it is the only site presentin the 2H polytype, but
no NMR data seem available.The chemical shift
has beenpredictedto be -31p.p.m. [21].The two
resonances
at -20 and -25p.p.m. in the PC sample
pyrolysedat 1700'C, can be assignedto the presence
of type B and C silicon units of somea phases.The
third componentdue to type A units could lie under
the major peak at - 16.2p.p.m.The assignment
of
this peak is more complicated,becausethe chemical
shift doesnot correspondto the usuallyreportedvalue
for l-SiC. Thisvalue,around- l6 p.p.m.hasalready
beenreported for powderedsamplesthat were supposedto be P-SiC[4, l5], and alsofor p-SiCsingle
crystals[22. Inkrott e/ a/. [ 5] found suchpeak in a
plasmasynthesizedSiC sample.After annealingthis
material above 1600'C under inert atmosphere,the
expectedp-SiC peak at l8.3p.p.rn.appeared.
B-SiC
has only type A silicon units. However,the chemical
shift is differentfrom that oftype A unitsin 6H or l5R
polytypes(-13.9 and - 14.9p.p.m.,respectively).
Hatman et al. [21]assignedthis differenceto different
Si-C bondlengths:in p-SiC, the siliconsite has a full
tetrahedrasymmetry,while one long and threeshort
Si-C distancesare presentin the 6H polytype. The
peakaround- l6 p.p.m.,cancertainlybe
anomalous
assignedto typeA units.The shift comparedto the 3C
polytype could be due to a lower symmetryof the

Ackn ow ledg eme nts

Mike Jeckleand Richard Lysseare thankedfor their
respective
contributionin the MAS-NMR and TEM/
SAED investigations.
NSF is alsoacknowledged
for
financialsupportofthis study,Contractno. DMR 87
063'79.
References
LL K. J. WYNNE and R. W. P.ICE, Ann. Re\,.Mater. Sci.
14 (1984\291.
2. B. A. BENDER, R. W. RICE and J. R. SPANN, "/.
Amet. Ceram.Soc.70 (1987)C58.
3. G. D. SORARU, F. BABONNEAU and
J, D.
MACKENZIE. J. Non-Cr))st.
SolrZr106(1988)256.
4. G. D. SORARU,F. BABONNEAUANdJ. D. MAC.
"Procedings of the 7th International
KENZIE,
Symposiumon Ceramics", 14to 16December1988,Bologna,
Italy, in prcss.
5. S. YAJIMA et al.,J. Matet.Scl. 13 (1978)2569.
6. Y. HASEGAWAM
, . I I M U R A a n dS . Y A J I M A , , i d 1 5
(1980)720.
7 - Y . H A S E G A W A and K. OKAMURA, rrid l8 (1983)
3633.
8 . S , Y A J IM A , J , H A Y A S H a n d M . O M O R t , C h c m l. , ? ! ! .
(1975)931.
9. K. OKAMURA et d1., "UltrastructureProcessingof
Advaced Ceramics", edited by J. D. Mackenzie and
D. R. Ulrich (Wiley, New York, 1988)p. 501.
i0. S. LIEDIKE et al., J. Non-Crts. Solrzs97198(1987)1083.
ll. J. LIPOWITZ et al., Adv.Ceram.Matet.2 (1987)l2l.
12. T. L. COTTRELL, "The Strengthof ChemicalBonds"
(Butterworths,London, 1958).
13. R. CHAIM, A. HEUER and R. T. CHEN,
Ceran. Soc.7l (1988)960.
14. c. L. MARSHALL et al., Proc. Symp. Sc. Cerum. l4
(t987) 341.
1 5 . K. E. INKROTT, S. M. WHARRY
and
D. J.
O DONNEL. Mater. Res.Soc.Symp.Proc.73(1986)155.
6 . D. TURNBULL, J. Non-Cr)st.So/rzs.
75 (1985)l9?.
1 7 . L. CZEPREGI et al., J. Appl. Phrs. 49 (1978)3906.
8 . L- CZEPRECI et al., SolidState Commun.21(19'7'l)1019.
1 9 . J. AYACHE et al.,J. Mater.Srt. a?r. 7 (1988)885.
20. v. HAASE er a/., "Gmeln Handbook of Inorganic
Chemistry",(Springer,Berlin, Heidelberg,New York, 1984)
Supplement
Volume82, Si-Silicon,"Propertesof Crystalline
SiliconCarbide".
21. J. S. HARTMAN et al., J. Amer. Chen. Soc. 109 (198'7)
6059.
22. J. R. GUTH and W. T. PETRUSKEY. ]. Ph},s,Chem.
9t (1987)5361.

Received 4 May
and accepted 29 September 1989

3893