Você está na página 1de 8

J. Am. Ceram. Soc.

, 89 [7] 21882195 (2006)


DOI: 10.1111/j.1551-2916.2006.00920.x
r 2006 The American Ceramic Society

Journal

A Model for the Nanodomains in Polymer-Derived SiCO


Atanu Saha and Rishi Rajw
Department of Mechanical Engineering, University of Colorado, Boulder, Colorado 80309-0427

Don L. Williamson
Department of Physics, Colorado School of Mines, Golden, Colorado 80401

inform about the spatial distribution of the various tetrahedra.


The carbon-to-carbon bonding has been investigated by C13
NMR, which shows that it is of predominantly sp2 character.1
Limited amount of Raman spectroscopy suggests the existence
of graphene sheets of carbon with some degree of long-range
order.5,6 Most significantly, the Raman peaks appear to narrow
as the SiCO is annealed at higher temperatures. However, quantitative interpretation of these spectra requires as yet unavailable
models of Raman emission from carbon structures in complex
multicomponent molecular structure of the PDCs. (Published
models of Raman spectra for long-range structure of carbon
based upon the positions of the G and D bandsapply
only to pure, one component, amorphous carbon7; they may
not be automatically applied to the complex constitution of the
PDCs.)
Studies of bonding in SiCO by other spectroscopic techniques, including X-ray photoelectron spectroscopy (XPS), infrared spectroscopy (IR), and the characterization of the radial
distribution functions by X-ray and neutron scattering, are consistent with the results from NMR as described above. The XPS
spectra3,8,9 and IR spectroscopy10 reveal the presence of silicon
oxygen bonds similar to those in silicates, as well as silicon
centered tetrahedra with mixed bonds to carbon and oxygen.
Carbon is found to be bonded to either silicon or to itself, but
not to oxygen.3 Radial distribution functions obtained from Xray and neutron diffraction11,12 gives two distinct peaks, one at
162 pm and the other in the range 188189 pm, which are attributed to the nearest neighbor bonds between SiO and SiC,
respectively. A peak at 306 pm is assigned to the second nearest
neighbor distance between two Si atoms. The average SiOSi
bond angle was found to be smaller than in amorphous silica
and decreases with higher carbon content, which is in agreement
with the ab initio calculations of SiCO structure.13
The microstructure of SiCO, of various compositions, has
been studied by transmission electron microscopy (TEM).3,5,1416
The most common result is that SiCO pyrolyzed at 10001C is
generally featureless. However, high-resolution TEM studies by
Kleebe et al.15 and Turquat et al.16 suggest the presence of
graphitic carbon in nanoscale units. They also nd that in some
regions the local arrangement of silicon and carbon atoms is
close to that of SiC; but the extent of such units is far short of
the volume fraction that would be predicted from the composition with the assumption that all SiC is segregated into near
crystalline structures. The wide angle X-ray diffraction of the
SiCO, annealed below the temperatures where they would decompose, also fails to show significant crystallization of silicon
carbide.17
A second, and a very unusual, feature of the PDCs is the
discovery of nanodomains by small-angle X-ray scattering, of
size ranging from 1 to 5 nm, that persist to very high temperatures. The presence of these domains was rst reported in the
SiCN system by the Max Planck group,18 and later by us.19 As
far as we know the present work is the rst to report the measurement of the domains in SiCO ceramics by SAXS. The detailed chemical structure of these domains remains obscure. In

The polymer-based synthesis of ceramics such as SiCO (and


SiCN) leads to the incorporation of significant amounts of carbon into their molecular structure. A key feature of the nanostructure of these polymer-derived ceramics is the revelation of
persistent, 15 nm size domains by small-angle X-ray scattering.
Here we present a model for these nanodomains, which is consistent with the nuclear magnetic resonance (NMR) data and
with the phenomenological properties of SiCO (high resistance
to creep and viscoelastic behavior). The model consists of clusters of silica tetrahedra encased within an interdomain wall constituted from mixed bonds of SiCO, and from a network of sp2
carbon. The model predicts the domain size as a function of the
carbon content. These predictions are in reasonable agreement
with the measurements of the nanodomains in SiCO synthesized
with varying carbon contents (the domain size decreases with
higher carbon). Simple maps are developed for easy reading of
the domain size and the width of the interdomain boundary in the
composition diagrams.

I. Introduction

OLYMER derived ceramics (PDCs) of two families, siliconoxycarbide and silicon-carbonitride share a common trait:
they incorporate large quantities of carbon, up to about onethird molar into their molecular structure. The carbon may be
incorporated only via the organic route where a highly crosslinked polymer (or a gel) is pyrolyzed into the ceramic. The carbon bonded to silicon in the silicon-based organic oligomers
(silsesquioxanes and polysilazanes) is retained because the siliconcarbon bonds survive the pyrolysis better in comparison
with the CH bonds. Thus hydrogen is released, in the 7001
9001C temperature range,1 leaving behind excess carbon in the
ceramic. In this way carbon, which is insoluble in silica (very
little of it dissolves if it is mixed into molten silica glass),2 can be
incorporated into it by the polymeric process.3 Nuclear magnetic resonance (NMR) studies of Si29 have shown that the carbon in SiCO exists in mixed bond, silicon centered, sp3
tetrahedra, of all possible SiCO congurations1,4; even further,
quantitative analysis of the NMR data suggests that, in certain
compositions, the number of mixed tetrahedra of different types
(i.e. SiO1C3, SiO2C2, etc.) are in proportion to the random
number fractions that can be predicted from the composition
of the SiCO.4 It must be noted that the NMR data tells us only
that the number fractions are randomly distributed; it does not

G. Sorarucontributing editor

Manuscript No. 21050. Received October 4, 2005; approved December 12, 2005.
This work was supported by the Air Force Ofce of Scientific Research under the
direction of Dr. Joan Fuller under Grant no: FA9550-04-1-0154, and by a grant from
the National Science Foundation from the Ceramics Program of the Division of Materials
Research.
w
Author to whom correspondence should be addressed. e-mail rishi.raj@colorado.
edu

2188

July 2006

A Model for the Nanodomains in Polymer-Derived SiCO

Fig. 1. Two concepts for the nanostructure of polymer-derived SiCO.


In (A) the graphitic carbon forms the nanodomains, which are dispersed
in a silica matrix. In (B) the domains are formed by clusters of silica
tetrahedra that are encased in a network of graphene.

this paper we propose and analyze a model for these nanodomains. The underlying background of the work is the assumption that the unusual ultrahigh temperature properties of the
PDCs, the incorporation of carbon by the polymer-based process, and the nding of mixed bonds by NMR despite very low
intrinsic solubility of carbon in silica, are intertwined with the
existence of the nanodomains. In a companion paper20 it has
been shown that silica domains can be leached away by etching
in hydrouoric acid, leaving behind pores. The size of these
pores, measured by BET, is comparable with the domain size as
measured by SAXS. Therefore, it can be inferred that the
domains seen in SAXS, as reported here, are clusters of silica
molecules.
The proposed model for the nanodomains meets three critical
boundary conditions: (a) it is consistent with the NMR results of
mixed bonds and the sp2 nature of the carboncarbon bonds, as
described above, (b) it explains the unusual resistance to crystallization and creep4,5,21 and the nding of viscoelasticity in
SiCO,22 and (c) it predicts how the domain size, as measured by
SAXS, varies with the composition of the SiCO, especially with
different carbon contents. The paper centers on the last point,
that is, the experimental verication of the predicted changes in
the domain size with carbon content in SiCO ceramics.
To begin we consider two possible models for the nanodomains, which are sketched in Fig. 1. Both types are consistent
with the NMR data. In type A, shown on the left, the domains
consist of a discontinuous dispersion of sp2 carbon embedded in
a continuous phase of silica. The mixed SiCO tetrahedra are
formed at the interface between the carbon nanoparticles and
the silica matrix. In this instance the size of the nanodomains
will be determined by the free carbon, that is, the carbon in
excess of the stoichiometric mixtures of SiC and SiO2, and by the
relative ratio of SiC/SiO2. In the model on the right, type B, the
domains are presumed to consist of clusters of silica tetrahedra
that are encased in mixed bonds of SiCO; these units are then
embedded in a cage-like network of sp2 graphene carbon that
forms an interconnected network.
It is intuitively self-evident that only the type B model is consistent with the quite remarkable creep and viscoelastic behavior
of SiCO. These ceramics do not exhibit long-term steady-state
creep,21 even at very high temperatures; instead they are viscoelastic22 at temperatures where pure silica ows readily and viscously. The model in A should behave like pure silica, albeit
with a somewhat higher viscosity. On the other hand, the graphene network in B would prevent long-term creep, as the ow in
the silica domains will gradually shed load on to the graphene
network, which can deform only by elastic deformation. Thus
long-term, steady-state creep in the type B structure is not admissible. At the same time, this structure will clearly show viscoelastic behavior as, upon unloading, the elastic strain stored in
the graphene network will give back the applied strain as the
silica domains are restored to their original shape. In Scarmi
et al. 22 the time constant for the viscoelastic relaxation has been
shown to be consistent with the viscosity of pure silica.
The SiCO ceramics have another peculiar characteristic. They
contain large amount of silica but the silica does not crystallize

2189

into cristobalite, as does pure silicaeven at temperatures as


high as 14001C. Indeed, crystallization of silica in SiCO is yet to
be reported under any conditions. It has however been shown
that the silica passivation layer formed by oxidation of SiCO
does crystallize, implying that the presence of carbon in the
SiCO is the reason that the silica resists crystallization. The
model in Fig. 1(A) would not resist crystallization as silica forms
a continuous phase as in pure silica. In model B, however, the
nite domain size of silica may inhibit crystallization if the domain size is comparable with the critical size for the nucleation
of the crystallization phase. Embedded in this idea is the importance of the energy of the interface formed between the layer
of mixed bonds surrounding the cluster of silica tetrahedra. It
would appear that this interface has a particularly low interfacial
energy such that it inhibits the heterogeneous nucleation of
cristobalite.
It is important to recognize that Type A and Type B models
are inverse of one another. For example the inuence of an
increase in carbon content on the domain size will be opposite
for these two cases. In A the domain size will increase
(assuming that the number density of the domains remains the
same), while in B it will decrease with higher carbon. Experiments presented below are consistent with the prediction from
case B.
Thus, we move forward with the assumption that the model
in Fig. 1(B) is consistent with the NMR data and with the phenomenological properties as discussed just above. The remainder
of this article is devoted to experiments that measure the domain
size as a function of the carbon content, and the development of
a model, which successfully predicts the domain size on the basis
of the composition of the SiCO ceramic. The experiments are
described rst.

II. Experimental Procedure


Four compositions of SiCO were prepared by the polymer pyrolysis route. Polyhydro methyl siloxane (PHMS) and
1,3,5,7-tetramethyl-1,3,5,7-tetravinylcyclotertrasiloxane (Gelest)
were mixed in various proportions, followed by cross-linking
at room temperature by adding Pt-catalyst (1 wt% of
polymer). The cross-linked polymer was aged for 24 h in Ar
and the elastomeric mass was heat treated at 4001C for 5 h in
Ar. Heat-treated mass was ball-milled in a plastic jar using zirconia balls as the grinding media. The ball-milled powder was
pyrolyzed at 10001C for 5 h. and subsequently heat treated at
12001C for 1 h in Ar. SiCO of four different compositions were
prepared by adding 5, 15, 50, and 95 wt% of vinyl bond containing precursor to PHMS as denoted by A, B, C, and D,
respectively.
Elemental analysis of SiCO powder was done for C and O
content while the silicon content in the samples was calculated as
the difference of the sum of the measured carbon and oxygen
content to 100%. The carbon and oxygen content in SiCO
was measured by combustion (Model C-200, LECO Corp., St.
Joseph, MI) and fusion (Model TC-600, LECO Corp.) method,
respectively.
Phase and crystalinity were identied by X-ray diffraction
technique using X-ray diffractometer (Scintag Inc., Cupertino,
CA) with CuKa radiation. For SAXS analysis, powder samples
were sandwiched between 99.999% pure, 10 mm thick, aluminum foil and measured with Kratky Compact SAXS System
attached to a Cu rotating anode X-ray generator.23 The sample
data were corrected for background scattering from the Al- foil.
The size distribution of the short-range feature present in the
SiCO was determined by deconvoluting the SAXS data.23

III. A Model for the Nanodomains


The model for the nanodomains is constructed from three constituents: clusters of silica tetrahedra that form the heart of the
domain, the surrounding monolayer of mixed bonds of the type

2190

Journal of the American Ceramic SocietySaha et al.

Vol. 89, No. 7

Fig. 3. The domain size is determined by the surface to volume ratio of


the domains, and therefore depends on the composition of the point S.
The domain size becomes smaller as S moves closer to SiC. The free
carbon content, given by c, determines the carbon chemistry of the
interdomain boundary. The overall composition of the specimen is
given by P.
Fig. 2. A concept of the molecular make-up of the nanodomains. Note
that the interdomain boundary consists of graphene layers with mixed
SiCO bonds forming the interface with the silica domains.

SiCnO4n where n gives the fourfold coordination of silicon to


carbon and oxygen, and the graphene cage-like network that
encases the domains. The carbon bonded to silicon is called sp3
or carbidic carbon and the carbon in graphene is sp2 or graphitic
carbon. A conceptual molecular make up of such a structure is
drawn in Fig. 2.
The rst step in quantitative analysis of the model is to separate the overall composition of the silicon-oxycarbide into the
stoichiometric part and the graphitic part, as illustrated in
Fig. 3. The line drawn from the C node through the overall
composition, P, meets the tie-line joining SiCSiO2 at the point
S. It is immediately evident that the composition of the silica
cluster plus the surrounding mixed bonds is given by S, as all
silicon atoms in the cluster and the interface are tetrahedrally
bonded. It is further self-evident that the domain size in this
picture will be determined by the position of S on the tie-line: the
closer S is to SiO2 the larger will be the domain size simply from
the surface-to-volume ratio argument. Indeed, knowledge of the
volume of the silica tetrahedra and the surface area occupied by
the average mixed bond will explicitly yield a value for the domain size. Thus every point on the SiC/SiO2 tie-line corresponds
to a specific value of the domain size. Once the domain size is
known then the distance c in Fig. 3, which gives the mole
fraction of the graphitic carbon can be translated into a thickness for the domain wall. In this way the domain size and the
effective width of the interdomain boundary can be determined
from the average composition of the silicon-oxycarbide ceramic.
The quantitative analysis described below is based upon these
concepts.
Following the procedure outlined above, the rst step is to
separate the total composition of the SiCO into the stoichiometric part and the graphitic part, the latter being carbon bonded mostly to itself. Thus the overall composition, SiOXCY
partitions as follows:
SiOX CY SiOX C10:5X CY10:5X

(1)

where the rst term on the right refers to the predominantly


tetrahedrally bonded stoichiometric composition and the second
term to the graphitic carbon.
The domain size is related only to the stoichiometric term
in Eq. (1), and is calculated as follows. We assume the shape of
the domains to be space-lling cuboids. The size of the domains

is dened by the volume of the silica cluster within this cuboid.


We call it d 0 . It follows that the number of silica tetrahedra
within this domain volume, nSiO2 , is given by

nSiO2

d 03
VSiO2


(2)

where VSiO2 is the effective volume per silica tetrahedra.


The mixed-bond interface layer that surrounds the silica domain also consists of Si-centered tetrahedra but with various coordinations of O and C bonds. The population of different kinds
of co-ordinations will determine the average composition of the
interface layer. First we consider the general case such that this
composition is dened by p, as in SiO2(1p)Cp; note that the
stoichiometry is conserved. The number of such tetrahedra in
the interface layer, which is called nSiOC, is then related to the
surface area of the domain, and the average area occupied by the
interfacial tetrahedra, ASiOC, as follows:

nSiOC

6d 02
ASiOC


(3)

By equating the composition of the domains and the interface


layer to the number of silica and SiCO mixed-bond units we
obtain
nSiO2 SiO2 nSiOC SiO21p Cp SiOX C10:5X

(4)

Substituting from Eqs. (2) and (3) into Eq. (4) we get



 02 
d 03
6d
SiO2
SiO21p Cp SiOX C10:5X
VSiO2
ASiOC

By enforcing mass balance for oxygen on both sides of the


above equation we obtain
X

2d 03 =VSiO2 121  pd 02 =ASiOC


d 03 =VSiO2 6d 02 =ASiOC

Rearranging terms in the above equation gives the nal


results for the domain size:
d0 6



VSiO2
2p
1
ASiOC 2  X

(5)

July 2006

The above result is for an arbitrary interface composition,


specied by p. Its value depends on the assumption of the distribution of the mixed-bond tetrahedra. If the mixed bonds in
the interface are equally populated by tetrahedra consisting of
SiO1C3, SiO2C2, and SiO3C1, so that the average coordination is
given by SiO2C2, then the average composition will be SiO1C0.5,
that is, p 5 0.5. (Please note that SiC4 tetrahedra are not included in the mixed bonds that are assumed to exist in the domain
wall; this is because the interfacial bonds, by definition, must
connect the graphene wall to the SiO2 tetrahedra within the domains. Therefore, they must contain all three constituents Si, O,
and C. Stoichiometric silicon carbide bonds are likely to precipitate out as crystallites, which is indeed observed when the
PDCs are heated to very high temperatures17; when this occurs
then the domain size can be expected to increase as explained
schematically in Appendix A.) Substituting this value for p into
Eq. (5), the domain size for uniformly distributed tetrahedra in
0
the interface, which we call d0:5
, is given by the following result:


VSiO2 X  1
0
(6)
6
d0:5
ASiOC 2  X
This result is valid only when 1oXo2.
The effective width of the interdomain wall (see Fig. 2) can be
similarly derived from geometric arguments and compositional
constraints. The number of carbon atoms in this layer, nC, is
given by the volume occupied per carbon atom in the interdomain wall, which we call VW, divided by the total volume of the
interdomain boundary layer on a per-domain basis:
nC

2191

A Model for the Nanodomains in Polymer-Derived SiCO

 02

6d dw =2
VW

(7)

Here dW is the effective width of the interface (the factor of


one half accounts for two domains abutted against the same
graphene layer). Note that the interdomain boundary width includes the thickness of the graphene carbon as well as the monolayers of mixed CSiO bonds sandwiched between adjacent
silica domains.
Setting the result in Eq. (7) to the carbon composition as
given by the second term in the right-hand side in Eq. (1) gives


d 0 VW
VW
dw
(8)
Y 6Y  p: 0
3 VSiO2
d ASiOC
Substituting for d 0 from Eq. (5) in the above equation we
obtain the nal result for dW:


2VW p 2Y
dW
1
(9)
ASiOC 2  X
The results presented below are based on estimates for the
domain size as given by Eq. (5), and the estimate for the effective
width of interdomain wall as given by Eq. (9). The application
for these equations requires values for VSiO2 , VW, and ASiOC.
The value for VSiO2 is calculated from the density of amorphous
silica; assuming a value of 2.2 g/cm3 gives VSiO2 0:0453 nm3 .
VW is volume of the domain wall for one unit of carbon atom; it

is therefore equal to the area occupied by one carbon


atom multiplied by the width of the wall. The area occupied
per carbon atom is estimated from the volume per carbon
atom in graphite, which is 0.0091 nm3; the area per carbon
atom is then equal to the volume per atom to the power of twothirds, which gives 0.044 nm2. The volume of the interdomain
wall, per carbon atom, is then equal to the area per carbon atom
times the width of the boundary. For the latter we choose a
number equal to one unit of interplanar spacing between
graphene layers in graphite plus one unit of the dimensions of
the silica tetrahedra. The rst quantity is 0.335 nm and the second we assume to be the cube root of the volume of one
SiO2 molecule, which was just quoted above. In this way the
width of the domain wall is approximated to be equal to 700 pm,
and VW 5 0.044  0.7 5 0.03 nm3. These values can be rened
with time as we gain greater quantitative understanding of the
interdomain wall via ab initio molecular dynamics simulations.
It now remains to estimate a value for ASiOC. We calculate it
using the following relationship:

ASiOC

MSiOC
AV rSiOC

2=3
(10)

where MSiOC and rSiOC are molecular weight and density of an


average mixed-bond unit and AV is the Avogadros number. We
assume MSiOC 5 47.5, which is obtained by assuming p, to be
equal to the average of 0.5 and 0.75. The density is calculated
from taking the average of the values for amorphous silica, 2.2
g/cm3 24 and for amorphous silicon carbide, 2.75 g/cm3.25 Substituting these values into Eq. (10) we obtain ASiOC 5 0.1 nm2.
We are now ready to estimate the domain size d 0 and the width
of the interdomain boundary, dW, by substituting the above
values into Eqs. (5) and (9). These results are presented in the
next section. We begin by describing the experimental results.

IV. Results
(1) Experimental
The main experimental results are summarized in Table I. They
give the compositions of the four samples, and the partitioning
of the composition into stoichiometric compositions of SiO2/SiC
and free carbon. The measurements of the average domain
size are given in the middle column. The SAXS data and the
distribution of the domain sizes obtained from deconvolution of
these data are included in Fig. 4. Later in this article we show
that in some instances, the width of the domain walls can be
expected to be comparable with the size of the domains themselves. In these instances the SAXS data may include the scattering from the domain walls as well as from the clusters that
form the volume of the domains. However, the proximity in the
size of the domain walls and the domain size does not negate the
basic conclusions in this article. The question arises whether or
not more detailed information about the domain walls and the
domains themselves can be elicited from the SAXS data. The
answer to this question is yes, but more controlled specimens
need to be prepared and estimates of densities of the walls and
the volumes of the domains must be inserted into the models for
deconvoluting the SAXS data. These experiments and modeling
are currently underway in our laboratory.

Table I. Experimental and Predicted Values of the Domain Size


Predicted domain size (nm)
SiCO sample

A
B
C
D

SiOXCY

SiCSiO2 and free carbon

Average domain size


by SAXS (nm)

SiO1.49C0.9
SiO1.38C1.31
SiO1C1.6
SiO0.92C1.94

SiO1.49C0.25510.645 C
SiO1.38C0.3111.0 C
SiO1C0.511.1 C
SiO0.92C0.5411.4 C

2.3
1.7
1.1
1.4

Predicted interdomain boundary width (pm)

p 5 0.5

p 5 0.75

p 5 0.5

p 5 0.75

2.6
1.7
N/A
N/A

5.3
3.9
1.4
1.1

760
968
N/A
N/A

1140
1450
1000
1170

2192

Journal of the American Ceramic SocietySaha et al.

Vol. 89, No. 7

Fig. 4. Small-angle-X-ray data, and the distribution of the nanodomains obtained from these data.

The X-ray diffractograms for the as-prepared four samples


are given in Fig. 5. As expected, the structure is predominantly
amorphous. The very broad peak near B221, in samples with
the lower carbon content, is attributed to amorphous silica.5 The
peak is so broad that it is barely discernible in the samples with
the higher carbon content. The broadening of this silica peak
may be due to a combination of two factors: a wide dispersion in
size of the silica nanodomains, and the elastic distortion in these
clusters because of the constraint imposed by the surrounding
mixed bonds and the interdomain layer. Molecular ab initio
calculations of the structure predict large strains in such structures.13

(2) The Predictions from the Model


The model presented in the preceding section gives two results:
the size of the silica domains, and the width of the interdomain
boundaries. These results are given by Eqs. (5) and (9). The
equations include a parameter, p, which gives the composition of
the mixed-bond monolayer surrounding the silica domains. The
range of the allowed values for p are constrained by the composition. The average composition of the materials is written as
SiOXCY, while the composition of the mixed bond monolayer is
SiO2(1p)Cp. As the overall composition includes the SiO2 tet-

Fig. 5. X-ray diffraction spectra showing a predominantly amorphous


structure.

rahedra that form the domains, it is required that:


X  21  p

(11)

which leads to a condition for the minimum allowed value of p


in comparing theory with experiment:
p1

X
2

(12)

Note that if X  1 then p may assume a value of r0.5 (as p is


not allowed to go negative). However, when Xo1 then p must
be greater than 0.5. This rule is important when comparing theory with experiment.
As d 0 and dW depend only on the composition it is possible to
plot loci for given values of these parameters in the composition
diagrams. As d 0 is determined by the stoichiometric part of the
overall composition (i.e., by the point S in Fig. 3), plots for
constant values of d 0 are obtained by drawing straight lines from
the C node to a point on the lie lines that joins SiC with SiO2.
The plots dW are also linear except very close to the SiO2 corner.
The values for d 0 and dW are dependent on the choice of p. Thus,
the maps constructed from Eqs. (5) and (9) for three values of p
are given in Fig. 6. They can be read immediately for estimating
d 0 and dW from the overall composition of the silicon oxycarbide.

(3) Comparison of Experiment with the Model


The measurements of the domain size are compared with the
experiment in Table I, and, graphically, in Fig. 7. The low carbon compositions, A and B, are shown on the left in Fig. 7; for
these maps for p 5 0.5 is assumed as X  1. The points lie on the
same contour for dW 5 700 pm. However, they lie on different
contours for the domain size. Sample A, with the lowest carbon
content, ts d 0  2 nm, while B is closer to d 0  1 nm.
The samples C and D with the higher carbon content are
compared with the model by assuming p 5 0.75, as Xo1. The
positions of C and D in the composition maps are shown on the
right in Fig. 7. The domain size varies between 1 and 2 nm, while
both specimens fall on the same contour for the width of the
interdomain boundary (dW 5 1400 pm).
The exact predictions of the width of the interdomain wall
and the domain size are quoted in Table I. Given the approx-

July 2006

A Model for the Nanodomains in Polymer-Derived SiCO

2193

imate nature of the model, the agreements between the predicted


and measured domain size are remarkable. The samples A and
B, the agreement is obtained with p 5 0.5, and for the higher
carbon samples p 5 0.75 gives a good comparison. The calculated width of the domain wall is varies from 700 to 1400 pm.
These values are consistent with the picture presented in Fig. 2.
The higher value of dW is predicted for the samples with the
higher carbon content, suggesting an increase in the wall thickness with carbon content. It is particularly satisfying that samples A and B are predicted to have the same value for dW. The
same is true for samples C and D, although they lie on a larger
value of dW. These results suggest that within certain range of
carbon content the structure of the domain walls remains invariant.

V. Discussion

Fig. 6. The domain size maps for p 5 0.25, 0.5, and 0.75.

The results presented in this article show that even a simple geometrical model based upon surface-to-volume ratio of the domains is moderately successful in predicting the domain size in
the silicon-oxycarbides. The principal assumptions in the model
lie in the framework of the molecular structure, which is justied
on the basis of the NMR data and the creep and viscoelastic
behavior of these materials at high temperatures. The model has
only one adjustable parameter, given by p, which denes the
composition of the layer of the mixed bonds that are assumed to
surround the silica domains. Reasonable values of this parameter give good agreement with experiment.
The ideas sketched in Fig. 1(B), and in Fig. 2, form the foundation for the model. We discuss three aspects of the model: Is
there additional evidence (beyond the arguments based upon
creep, viscoelasticity, and the resistance to crystallization) that
supports the proposed view of the molecular structure of the
nanodomains? What are the approximations in the model? and
what may be the scientific and technological consequences of the
model?
On the rst point, there is indeed additional evidence that
supports the schematic in Fig. 2. Experiments in progess at the
University of Trento and in our laboratory continue to support
the suggested structure of the nanodomains. Results from
Trento submitted separately for publication20 show that SiCO
amorphous ceramics can be etched with hydrouoric acid leaving behind a porous structure with a median pore diameter of
approximately 3 nm. The change in the composition after etching shows that only silica (not carbon) is removed by etching;
implying that the silica was occupying these pores before etching. Further experiments are underway to combine the studies at
Colorado and Trento for more detailed investigation with different compositions, and characterization of the domain and the
pore sizes by small angle X-ray scattering.
The second point concerns the approximations in the model.
The values for the domain size and the interdomain wall thick-

Fig. 7. The positions of the four specimens on the domain-size maps.

2194

Vol. 89, No. 7

Journal of the American Ceramic SocietySaha et al.

ness, both experimental and theoretical, are of the order of the


size of silica tetrahedra; therefore, their shape and size must be
viewed in terms of molecular networks, rather than as if in a
continuum. Furthermore, the domain structure is highly stochastic as is clear from the SAXS data (Fig. 4). Given these limitations, the convergence of experimental and theoretical values
quoted in Table I may be coincidental. In general, though, the
scale of the composition maps in Fig. 6, is appropriate for estimating a round number for the domain size and the width of
the interdomain walls. For example mapping the compositions
of six silicon-oxycarbides, extracted from the literature,5,26,27 in
Fig. 8, gives domain sizes that range from 1 to 6 nm and the
interdomain boundary widths from 3001000 pm. Composition
#6 is unusual because it lies on the SiCSiO2 tie-line, and therefore does not contain any free carbon. In such instances, the
domain size will still be predicted by the maps in Fig. 6; however, the interdomain wall thickness will now be given simply by
the thickness of the layer of mixed SiCO bonds, and will be independent of composition.
The third point of discussion concerns the further scientific
and technological consequences of the model. One issue concerns the nature and the properties of the domains when the
compositions lie on the SiCSiO2 tie-line. These materials may
be less robust in their resistance to creep and crystallization than
the materials that contain free carbon. Another issue concerns
the range of domain sizes and the range of compositions where
silicon-oxycarbide materials with similar properties can be prepared. The maps in Fig. 6 suggest that, realistically speaking,
domain sizes greater than about 6 nm will yield materials that
are likely to be more akin to silica than to polymer-derived ceramics in their creep and crystallization behavior, because these
compositions lie very close to the SiO2 corner. At the other extreme experiments suggest that as the carbon content is increased the domain size appears to reach a lower limit of
approximately 1 nm. These results may be explained by assuming that higher carbon content translates into a carbon-rich
constitution of domain boundarybut such an argument also
has its compositional limit. The domain structure of materials
with ultrahigh carbon contents fabricated by Blum et al.28 is
expected to be different from what envisaged in the present
model. A detailed study of these materials can provide new insights into the structure of the PDCs.
Finally we look at the technological potential of the nanoscale structure of the PDCs. One of the directions will be to
functionalize the domains by doping the silica tetrahedra for
obtaining novel electrical, optical, and sensor properties. Another possibility is to consider novel applications of the highsurface area, nanopore materials which can be produced by
etching of silicon-oxycarbide materials.20

Fig. 8. Data from six specimens from the literature5,26,27 placed on the
domain size map.

The model leads to maps that are built on the simple concept
of the surface-to-volume ratio of the domains. The maps can be
quickly read to estimate the domain size and the interdomain wall
thickness. The examples of data included in Figs. 7 and 8 provide
a perspective of the range of compositions where the maps should
apply. Surprisingly, the model not only predicts the trend in the
change of the domain size with carbon content, but also gives
reasonable quantitative agreement with experimental data.

Appendix A
In this appendix we consider the inuence of phase separation of
stoichiometric SiC from the overall composition on the domain
size. Stoichiometric SiC can be precipitated as crystallites at high
temperatures,17 which will change the composition of the remaining material as explained in Fig. A1. If the initial composition of the SiCO is given by P, then precipitation of SiC
will change the composition of the remaining non-crystalline
material such that it moves away from the SiC point along the
line connecting the SiC point to the initial composition. This
new composition of the non-crystalline material is shown as P0
in Fig. A1. The relative mole fractions of the composition P0 and
the mole fraction of stoichiometric SiC will then be given by the
lever rule as explained in Fig. A1.
The domain size (as well as the width of the domain wall) in
composition P0 can now be calculated using the same procedure
as described in this paper. The domain size is derived by nding
the intersection of the line drawn from the C apex through the

VI. Conclusions
There is incontrovertible evidence that SiCO made from the
polymer route contain nanodomains that persist up to very high
temperatures. It is conceivable that these domains evolve from
the organic state for two reasons. First, the pyrolysis of the organic leaves behind significant amounts of carbon in the ceramic, and second, the carbon is essentially insoluble in silica. It is
therefore, reasonable to postulate that as the silica tetrahedra
coarsen during pyrolysis, they reject carbon to their outer surfaces. When these surfaces form a continuous structure of mixed
bonds and graphene, the coarsening stops, and stable nanodomains are created. Whether such a nanodomain structure is
kinetically stable, or thermodynamically stable (as a result of
low interfacial energies) remains an open question.
The results presented here are based upon a molecular structure of the domains sketched in Fig. 2. This structure is consistent with the nearest-neighbor chemistry of silicon, oxygen, and
carbon bonds revealed by NMR, and with the highly unusual
phenomenological properties of SiCO that include astonishing
resistance to creep and to the crystallization of silica, and its
viscoelastic behavior.22

Fig. A1. Procedure for estimating the increase in the domain size as a
result of precipitation of stoichiometric SiC. The initial composition of
the SiCO is given by P. The composition of the noncrystalline material
remaining after precipitation of SiC is shown as P0 .

July 2006

A Model for the Nanodomains in Polymer-Derived SiCO

composition, with the SiCSiO2 tie-line. As shown in the composition map, the precipitation of SiC will lead to an increase in
the domain size. If the mole fraction of the precipitated SiC is
known (indeed it can be measured by X-ray diffraction), then
the new non-crystalline composition can be determined and the
change in the domain size can be predicted. Experiments along
these lines are underway in our laboratory.

Appendix B
The method of incorporating different amounts of carbon in
SiCO in recent years has followed the developments for making
black glass at Allied Signal29 and SRI30 dating back to the
mid-1980s. In one case29 the reaction between cyclohydridomethylsiloxane and vinyl compound was used to achieve practical preceramic polymer with high carbon content. In the
second instance31,32 a hydrosilylation reaction between PHMS
and water and/or tetravinyltertamethylcyclosiloxane, in the
presence of a transition metal catalyst (e.g., Pt), was used as a
practical way to cure and achieve high ceramic yields. This
method has been useful in the fabrication of composites and
coatings made from silicon oxycarbide and silica. In the present
experiments the second reaction, with multivinyl crosslinking
compounds (e.g., tetravinyltetramethylcyclosiloxane), was employed.33,34

Acknowledgments
Stimulating discussions with Prof. G.D. Soraru and Dr. R. Pena-Alonso
throughout the course of this work are gratefully acknowledged.

References
1

L. Bois, J. Maquet, F. Babonneau, H. Mutin, and D. Hahloul, Structural


Characterization of SolGel Derived Oxycarbide Glasses. 1. Study of Pyrolysis
Process, Chem. Mater., 6, 796802 (1994).
2
T. H. Elmer and H. E. Meissner, Increase of Annealing Point of 96 Percent
SiO-2 Glass on Incorporation of Carbon, J. Am. Ceram. Soc., 59 [56] 2069
(1976).
3
G. M. Renlund, S. Prochazka, and R. H. Doremus, Silicon Oxycarbide
Glasses. 2. Structure and Properties, J. Mater. Res., 6 [12] 272334 (1991).
4
L. Bois, J. Maquet, F. Babonneau, H. Mutin, and D. Hahloul, Structural
Characterization of Sol-Gel Derived Oxycarbide Glasses. 2. Study of the
Thermal Stability of the Silicon Oxycarbide Phase, Chem. Mater., 7, 97581
(1995).
5
G. D. Soraru, G. DAndrea, R. Campostrini, F. Babonneau, and G. Mariotto,
Structural Characterization and High-Temperature Behavior of Silicon Oxycarbide Glasses Prepared from SolGel Precursors Containing SiH Bonds, J. Am.
Ceram. Soc., 78 [2] 37987 (1995).
6
C. G. Pantano, A. K. Singh, and H. Zhang, Silicon Oxycarbide Glasses,
J. SolGel Sci. Technol., 14, 725 (1999).
7
D. S. Knight and W. B. White, Characterization of Diamond Films by
Raman Spectroscopy, J. Mater. Res., 4 [2] 38593 (1989).
8
G. D. Soraru, G. DAndrea, and A. Glisenti, XPS Characterization of Gel
Derieved Silicon Oxycarbide Glasses, Mater. Lett., 27, 15 (1996).
9
R. J. P. Corriu, D. Leclecq, P. H. Mutin, and A. Vioux, Preparation and
Structure of Silicon Oxycarbide Glasses Derived from Polysiloxane Precursors,
J. SolGel Sci. Technol., 8, 32730 (1997).
10
G. Das, G. Mariotto, and A. Quaranta, Vibrational Spectroscopy Characterization of Low-Dielectric Constant SiOC:H Films Prepared by PECVD Technique, Mater. Sci. Semicond. Process., 7, 295300 (2004).

2195

11
H. Brequel, G. D. Soraru, L. Schifni, and S. Enzo, Radial Distribution
Function of Amorphous Silicon Oxycarbide Compound, Mater. Sci. Forum,
343346, 67782 (2000).
12
H. Brequel, J. Parmentier, S. Walter, R. Badheka, G. Trimmel, S. Masse, J.
Latournerie, P. Dempsey, C. Turquat, A. Desmartin-Chomel, L. Le NeindrePrum, U. A. Jayasooriya, D. Hourlier, H.-J. Kleebe, G. D. Soraru, S. Enzo, and F.
Babonneau, Systematic Structural Characterization of the High-Temperature
Behavior of Nearly Stoichiometric Silicon Oxycarbide Glasses, Chem. Mater., 16
[13] 258598 (2004).
13
P. Kroll, Modeling and Simulation of Amorphous Silicon Oxycarbide,
J. Mater. Chem., 13, 165768 2003.
14
E. Breval, M. Hammond, and C. G. Pantano, Nanostructural Characterization of Silicon Oxycarbide Glasses and Glass Ceramics, J. Am. Ceram. Soc.,
77 [11] 30128 (1994).
15
H.-J. Kleebe, C. Turquat, and G. D. Soraru, Phase Separation in an SiCO
Glass Studied by Transmission Electron Microscopy and Electron Energy-Loss
Spectroscopy, J. Am. Ceram. Soc., 84 [5] 107380 (2001).
16
C. Turquat, H.-J. Kleebe, G. Gregori, S. Walter, and G. D. Soraru, Transmission Electron Microscopy and Electron Energy Loss Spectroscopy Study of
Non-Stoichiometric SiliconCarbonOxygen Glasses, J. Am. Ceram. Soc., 84 [10]
218996 (2001).
17
A. Saha and R. Raj, Crystallization Maps for SiCO Amorphous-Ceramics,
J. Am. Ceram Soc. to be published.
18
J. Bill, T. W. Kamphowe, A. Muller, T. Wichmann, A. Zern, A. Jalowieki, J.
Mayer, M. Weinmann, J. Schuhmacher, K. Muller, J. Q. Peng, H. J. Seifert, and
F. Aldinger, Precursorderived Si(B)CN Ceramics: Thermolysis, Amorphous
State and Crystallization, App. Organometallic Chem., 15 [10] 77793 (2001).
19
A. Saha, R. Raj, D. L. Williamson, and H.-J. Kleebe, Characterization of
Nanodomains in Polymer-Derived SiCN Ceramics Employing Multiple Techniques, J. Am. Ceram. Soc., 88 [1] 2324 (2005).
20
R. Pena-Alonso, R. Raj, and G. D. Soraru, Preparation of Ultrathin Walled
Carbon Based Structures by Etching Pseudo-Amorphous Silicon-Oxycarbide
Ceramics, J. Am. Ceram. Soc., (2005) submitted.
21
T. Rouxel, G. D. Soraru, and J. Vicens, Creep Viscosity and Stress Relaxation of Gel-Derived Silicon Oxycarbide Glasses, J. Am. Ceram. Soc., 84 [5]
10528 (2001).
22
A. Scarmi, G. D. Soraru, and R. Raj, The Role of Carbon in Unexpected
Visco(an)elastic Behavior of Amorphous Silicon Oxycarbide above 1273K, J.
Noncryst. Sol., 351 [2729] 223843 (2005).
23
D. L. Williamson, Microstructure of Amorphous and Microcrystalline Si
and SiGe Alloys Using X-Rays and Neutrons, Solar Energy Mater. Solar Cells,
78, 4184 (2003).
24
K. Vollmayr, W. Kob, and K. Binder, Cooling Rate Effects in Amorphous
Silica: A Computer-Simulated Study, Phys. Rev. B, 55 [22] 1580827 (1996).
25
V. Heera, F. Prokert, N. Schell, H. Seiforth, W. Fukarek, M. Voelskow, and
W. Skorupa, Density and Structural Changes in SiC After Amorphization and
Annealing, Appl. Phys. Lett, 70 [26] 35313 (1997).
26
M. A. Schiavon, C. Gervais, F. Babonneau, and G. D. Soraru, Crystallization Behavior of Novel Silicon Boron Oxycarbide Glasses, J. Am. Ceram. Soc., 87
[2] 2038 (2004).
27
G. D. Soraru and D. Suttor, High Temperature Stability of SolGel-Derived
SiOC Glasses, J. SolGel Sci. Tech., 14 [1] 6974 (1999).
28
Y. D. Blum, D. B. Macqueen, and H. J. Kleebe, Synthesis and Characterization of Carbon-Enriched Silicon oxycarbide, J. Eur. Ceram. Soc., 25 [23]
1439 (2005).
29
R. Y. Leung, S. T. Gonczy, and M. S. Shum. Carbon Containing Black Glass
Monoliths; U.S. Patent 5,242,866, 1993.
30
Y. D. Blum and R. M. Laine. Polysilazanes and Related Compositions,
Processes, and Uses; US Patent 5,008,422, 1991.
31
Y. D. Blum Hydroxysiloxane Precursors for Ceramic Manufacture; U.S.
Patent 5,128,494, 1992.
32
S. M. Johnson, Y. D. Blum, C. Kanazawa, and H.-J. Wu, Low-Cost Matrix
Development for an OxideOxide Composite, Metals Mater., 4 [6] 111925
(1998).
33
Y. D. Blum, D. B. MacQueen, and H.-J. Kleebe, Synthesis and Characterization of Carbon Enriched Silicon Oxycarbides, J. Eur. Ceram. Soc., 25, 1439
(2005).
34
S. Modena, G. D. Soraru, Y. D. Blum, and R. Raj, Passive Oxidation of an
Efuent System: The Case of Polymer-Derived SiCO, J. Am. Ceram. Soc., 88,
33945 (2005).
&

Você também pode gostar