Available online at www.sciencedirect.com

www.elsevier.com/locate/actamat

ScienceDirect

Acta Materialia 72 (2014) 22–31

Quantitative study on structural evolutions and associated energetics inpolysilazane-derived amorphous silicon carbonitride ceramics

Yaohan Chen a, Xueping Yang a, Yejie Cao a, Zhehong Gan b, Linan An a,⇑

a Department of Materials Science and Engineering, Advanced Materials Processing and Analysis Center, University of Central Florida,

Orlando, FL 32816, USAb Center of Interdisciplinary Magnetic Resonance, National High Magnetic Field Laboratory, Tallahassee, FL 32310, USA

Received 23 October 2013; received in revised form 20 March 2014; accepted 23 March 2014Available online 18 April 2014

Abstract

Several important structural changes and their energetics during high-temperature annealing of polysilazane-derived amorphous sil-icon carbonitride ceramics were quantitatively studied. A 29Si solid-state NMR study indicated that the structural transition in the Si-containing area can be described by an equilibrium reaction, 4SiCN3 = SiC4 + 3SiN4. The enthalpy and entropy for the reaction werecalculated to be positive. Raman and electron paramagnetic resonance (EPR) studies revealed that the structural evolution within thefree carbon area includes the graphitization of amorphous carbon and the lateral growth of nanographite, accompanied by a decreasein the point defect concentration. EPR results also suggested that the materials contain two kinds of point defects: carbon-danglingbonds at the edge and in the interior of the nanographite. It was found that the lateral growth of the nanographite followed a 2-D graingrowth process, and that the decrease in the defect concentration was mainly due to the growth of the nanographite. The energetics of thestructural changes was rationalized according to a simple structural model, and the effects of these changes on the stability of thematerials were discussed.� 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Polymer-derived ceramics; Structure evolution; Thermodynamics; Energetics; Nanodomain

1. Introduction

Amorphous Si-based ceramics, such as SiCN, SiCO,SiBCN and SiAlCN, synthesized by thermal decomposi-tion of polymeric precursors (referred to as polymer-derived ceramics (PDCs)) have received extensive attentionin recent decades [1]. Compared to polycrystalline ceramicsprepared by sintering the corresponding powders, this newclass of materials exhibits a set of unique and superiorproperties, such as excellent creep resistance [2–4], remark-able thermal stability [5], outstanding oxidation/corrosionresistance [6–9], high-temperature semiconducting

http://dx.doi.org/10.1016/j.actamat.2014.03.049

1359-6454/� 2014 Acta Materialia Inc. Published by Elsevier Ltd. All rights r

⇑ Corresponding author. Tel.: +1 0407 288 23436; fax: +1 04078230208.E-mail addresses: lan@mail.ucf.edu, linan74@gmail.com (L. An).

behavior [10,11] and anomalously high piezoresistivity[12]. These properties, together with the direct chemical-to-ceramic processing, make the materials very promisingfor a wide range of applications, including high-tempera-ture ceramic fibers, ceramic composites, microelectrome-chanical systems/microsensors for harsh environments,gas separation/absorption, and thermal protection coatings[13–18].

The unique properties, and thereby the applications, ofPDCs result directly from their unusual structures. A vari-ety of spectroscopic and microscopic techniques has beenemployed to investigate the structure of polymer-derivedSiCN [19–25]. These efforts have revealed a large amountof useful information, which paint a fairly clear pictureabout the PDC-SiCN structure. While it strongly depends

eserved.

Y. Chen et al. / Acta Materialia 72 (2014) 22–31 23

on the chemistry of the precursors and processing condi-tions [1,23,26], the structure of carbon-rich PDC-SiCNpossesses several common features. First, all SiCN ceram-ics contain a silicon-containing area and a carbon area(referred to as free carbon). The silicon-containing area isamorphous, and comprises of SiN4 and SiC4 tetrahedracores embedded in mix-bonded SiCxN4�x (x = 1, 2, 3)units. The mix-bonded SiCxN4�x tetrahedra have beenfound in polysilazane-derived SiCN ceramics, but not inpolysilylcarbodiimide-derived SiCN ceramics. On the otherhand, the free carbon area can be viewed as an assembly ofmany relatively ordered carbon nanoclusters with a fairlylarge number of defects. These phases are arranged into aunique heterogeneous amorphous nanodomain structure[27].

Upon annealing at higher temperatures, SiCN under-goes a series of structural evolutions, towards orderingand eventual crystallization. The structural evolutions canbe roughly divided into two stages: structural rearrange-ment at relatively lower temperatures (<1440 �C) andlarge-scale crystallization and carbothermal reaction athigher temperatures. The low-temperature structural evo-lution is rather interesting since many of the aforemen-tioned functional properties are related to this uniquenanodomain structure. It was revealed that the bondingtype and bonding ratio in SiCN remained unchanged dur-ing annealing between 1000 and 1400 �C [28], suggestingthat there is no net interaction between the Si-containingphase(s) and the free carbon phase within this temperaterange. Therefore, the low-temperature structural evolutioncan be analyzed by considering the changes in the Si-con-taining area and in the free-carbon phase separately. Ofcourse, the changes in the interface between the Si-contain-ing area and free carbon area are also expected to play akey role in determining many of the properties of SiCN.It was revealed that the major structural evolutions includethe following changes: the demixing of the SiCxN4�x mixedbonds into SiN4 and SiC4, resulting in the shrinkage of theSiCxN4�x area and the coarsening of the SiN4 and SiC4

areas [23,26]; and the graphitization of the free carbon,accompanied by an increase in the order of the carbonnanoclusters and a decrease in the concentration of thedefects [22,29,30].

Further important progress in the understanding ofPDCs has come from the measurement of their energeticsusing high-temperature oxide melt solution calorimetricmethods [31]. It was revealed that most of the reportedSiOC, SiCN, SiCNO and SiBCN ceramics exhibited anegative heat of formation with respect to a mixture oftheir crystalline counterparts of the same composition[23,24,26]. Upon high-temperature annealing, the SiCNceramics derived from polysilazanes showed a decrease inenthalpy [23].

In spite of these extensive studies, two aspects of thestructural evolution of SiCN have received little attention.First, even though the variation of different phases withannealing temperature has been well documented, these

previous studies primarily focused on showing the qualita-tive trend; no quantitative study on the structural evolutionhas been reported. Second, the previous calorimetric mea-surements can only provide energetic information on theentire material as whole; the energetics involved in eachprocess of the structural evolution has not been reported.In this paper, we report a quantitative study on thestructural evolution of a SiCN derived from polysilazane.Several important structural changes in both the Si-con-taining area and the free carbon area are quantified individ-ually. The energetics involved in each of these changes arederived and discussed in terms of their role in determiningthe stability and evolution of the SiCN.

2. Experimental

The SiCN studied here was prepared by thermal decom-position of a commercially available polysilazane(HTT1800, Kion Corp., Columbus, OH, USA) [32]. Theprecursor was first solidified without any catalyst in a Tef-lon tube at 200 �C for 16 h under ultrahigh-purity argon(UHP Ar) protection. The obtained solid was then pyro-lyzed under a steady flow of UHP Ar in a quartz tube fur-nace (GSL-1100X, MTI Corporation, Richmond, CA)using the following schedule: (i) heating to 900 �C at aheating rate of 3 �C min�1, (ii) further heating to 1000 �Cat a heating rate of 1 �C min�1; (iii) keep at 1000 �C for4 h; and (iv) cool to room temperature at a cooling rateof 5 �C min�1. The obtained ceramic was further annealedat different temperatures up to 1350 �C for 4 h under UHPAr protection in an alumina tube furnace (GSL-1600X,MTI Corporation, Richmond, CA). The resultant sampleswere analyzed using X-ray diffraction (XRD; Rigaku,Tokyo, Japan), which reveals that regardless of annealingtemperature, all samples were amorphous. The polymer-to-ceramic conversion of the polysilazane was studied bythermogravimetric analysis (TGA; SDT Q600, TA Instru-ments, New Castle, DE) in UHP Ar atmosphere up to1400 �C, using a heating rate of 5 �C min�1 and an argonflow rate of 150 ml min�1.

The resultant ceramics were characterized using high-resolution solid-state magic angle spinning (MAS) nuclearmagnetic resonance (NMR) on a Bruker DRX spectrome-ter at a static magnetic field of 19.6 T (1H frequency:830 MHz). 29Si solid-state NMR spectra were recorded atLarmor frequencies of 165.5 MHz. All spectra wereacquired using a standard single-pulse sequence with a tipangle of �45� and a recycle delay time of 5 s and basedon a total of 2048 scans. The 29Si chemical shifts were ref-erenced to tetramethylsilane (TMS).

The materials were also characterized by Raman spec-troscopy using a Renishaw inVia spectrometer (RenishawInc., Gloucestershire, UK), equipped with a 532 nm Sisolid laser excitation source and a sensitive Peltier-cooledcouple-charged device (CCD) detector. The laser beamwas focused on the sample through a �50 objective lens.The laser spot size was �10 lm in diameter and the laser

24 Y. Chen et al. / Acta Materialia 72 (2014) 22–31

power on the sample was kept below 2.5 mW to avoiddecomposition of surface carbon. At least 10 Raman spec-tra were obtained for each sample by mapping the imagesin order to minimize the measurement error.

The ceramics were further analyzed using electron para-magnetic resonance (EPR). The X-band (9.5 GHz) EPRspectra were recorded on a Bruker ECS-106 spectrometer(Bruker Instruments, Billerica, MA), using a TE102 cavitywith a microwave power of 2 mV. The modulation fre-quency was 100 kHz. The G-band (406 GHz) EPR spectrawere recorded on a home-built spectrometer at theNational High Magnetic Field Laboratory (Tallahassee,FL). The instrument was a transmission-type device inwhich microwaves are propagated along a 10 mm cylindri-cal waveguide through the sample compartment and fur-ther towards the bolometer detector. The microwave of406.4000(4) GHz (g � 2.00 around 14.52 T) was generatedby a phase-locked Virginia Diodes source. A superconduc-ting magnet (Oxford Instruments) capable of reaching afield of 17 T was employed. Thin Teflon vessels with anouter diameter of 9 mm and a depth of up to 10 mm ormore are used as sample containers.

3. Results

3.1. General

Fig. 1 shows the TGA/differential scanning calorimetry(DSC) curves for the polysilazane pre-solidified at 200 �C.The precursor exhibits a four-step weight loss [30,32–34]:(1) an initial weight loss of 2% between 200 and 380 �C,due to the release of low-molecular-weight oligomers; (2)5% weight loss between 380 and 570 �C, resulting fromthe evolution of ammonia (NH3) by transamination reac-tions and the consequently formed oligomeric units; (3) amajor weight decrease of �8% between 570 and 750 �C,due to the release of hydrogen (H2) and methane (CH4);and (4) 1.8% weight loss between 750 and 1000 �C because

Fig. 1. TGA/DSC curves of the starting precursor pre-solidified at 200 �C.

of H2 evolution. Note that there is no detectable weightchange between 1000 and 1400 �C, suggesting the materialis stable in terms of thermal decomposition in this temper-ature range. This is consistent with the previous observa-tion that the carbothermal reaction usually occurs attemperatures above 1440 �C.

3.2. Structural evolution of the Si-containing area

The structural evolution of the Si-containing area wascharacterized using 29Si MAS NMR. Fig. 2a comparesthe 29Si NMR spectra of the SiCN ceramics pyrolyzed atdifferent temperatures. The spectra exhibit a broad reso-nance peak centered at about �34 ppm, which is character-ized the amorphous SiCxN4�x (0 6 x 6 4) and caused bythe heterogeneity in the local silicon coordination environ-ments. To obtain more information from the spectra, thepeaks were fitted using Lorentz curves. The results showthat the broad curve can be split into three curves centeredat chemical shifts of �19, �34 and �50 ppm (Fig. 2b).These peaks correspond to SiC4, SiCN3 and SiN4 units,respectively [22,30,33,35,36]. The possibility of the exis-tence of a small amount of other mixed bonding environ-ments, such as SiC3N and SiC2N2, cannot be ruled out.

Fig. 2. (a) 29Si solid-state NMR spectra of the SiCN pyrolyzed at differenttemperatures; and (b) curve fitting of 29Si NMR for the SiCN ceramicpyrolyzed at 1300 �C.

Y. Chen et al. / Acta Materialia 72 (2014) 22–31 25

It can also be seen from Fig. 2a that with increasing anneal-ing temperature, the intensity of the SiCN3 peak decreases,but that of the SiN4 and SiC4 peaks increases. This suggeststhat the SiCN3 mix-bonds undergo a demixing process withincreasing annealing temperature, consistent with previousreports on similar materials [23].

In order to acquire quantitative information on thestructural evolution, the concentrations (mol.%) of thethree kinds of Si coordination environments were calcu-lated from the NMR fitting curves using the followingequations [37–39]:

½SiCN3� ¼ASiCN3

ASiCN3 þ ASiN4 þ ASiC4

ð1aÞ

½SiN4� ¼ASiN4

ASiCN3 þ ASiN4 þ ASiC4

ð1bÞ

½SiC4� ¼ASiC4

ASiCN3 þ ASiN4 þ ASiC4

ð1cÞ

where ASiN4, ASiC4 and ASiCN3 are the integration areasunderneath their corresponding peaks. The results are plot-ted in Fig. 3. It can be seen (Fig. 3a) that the concentrationsof the three units vary linearly with increasing annealing

Fig. 3. The mole percentages of different silicon bonding environments (a)and the mole ratio of Si–C to Si–N bonds (b) of the SiCN ceramicspyrolyzed at different temperatures.

temperature, with that of the SiCN3 decreasing and thatof the SiC4 and SiN4 increasing. However, the ratio ofSi–C bonds to Si–N bonds remains constant and was notaffected by the annealing (Fig. 3b), suggesting there is nonet exchange between the Si-containing phase and the freecarbon phase. This is also consistent with previous reportsthat the carbothermal reaction only occurs at temperatureshigher than 1440 �C and the TGA result that showed noobvious weight change between 1000 and 1400 �C(Fig. 1). Comparing the change in the concentration ofeach unit reveals that the relative changes in the concentra-tions can be described by the following stoichiometricreaction:

4SiCN3 $ 3SiN4 þ SiC4 ð2ÞAssuming the above reaction is an equilibrium reaction,

its equilibrium constant (K) can be expressed as:

K ¼ ½SiN4�3½SiC4�½SiCN3�4

ð3Þ

According to the Van’t Hoff equation, the temperaturedependence of the equilibrium constant can be related tothe Gibbs free energy by:

ln K ¼ �DG�

RT¼ �DH �

RTþ DS�

Rð4Þ

where DH� is the standard enthalpy of the reaction, DS� isthe standard entropy, R is the universal gas constant, and T

is the annealing temperature. Combining Eqs. (3) and (4)leads to:

ln½SiN�4

3½SiC4�½SiCN�4

¼ �DH �

RTþ DS�

Rð5Þ

Fig. 4 plots the experimental results in terms ofln([SiN4]3[SiC4]/[SiCN3]4) vs. 1000/T. The linear relation-ship suggests that (i) the assumption of the reaction beingin equilibrium is reasonable, and (ii) the enthalpy andentropy of the reaction are independent of temperature.

Fig. 4. A plot of ln ([SiN4]3 � [SiC4]/[SiCN3]4) vs. 1000/T.

26 Y. Chen et al. / Acta Materialia 72 (2014) 22–31

The enthalpy and entropy were then calculated from thecurve in Fig. 4:

DH � ¼ �slope � R ¼ 365:9 ðkJ mol�1Þ ð6aÞDS� ¼ intercept � R ¼ 206:6 ðJ K�1 mol�1Þ ð6bÞ

The positive enthalpy suggests that the demixing processis an endothermic reaction; that is to say, increasing tem-perature will increase the reaction constant, thus promot-ing the formation of SiC4 and SiN4 units. The positiveentropy suggests that the reaction leads to the increase ofdisorder, also leading to an increase in the reaction con-stant and the formation of SiC4 and SiN4 units withincreasing annealing temperature. The change in the stan-dard Gibbs free energy (DG�) of the reaction can then becalculated using the enthalpy and entropy data:

DG� ¼ DH � � DS�T ¼ 365900� 206:6T ð7ÞIt can be seen that while the change in Gibbs free energy

for the reaction is positive at any temperature in the testingrange of 1000–1350 �C, it decreases with increasing anneal-ing temperature. This suggests that the temperature-induced demixing reaction caused the decrease in Gibbsfree energy, and thus is energetically favorable.

The changes in the enthalpy and entropy can be betterunderstood by considering the detailed changes in thestructures of the material. According to previous studies[23,26] on the similar materials and our NMR resultsshown above, a possible structure can proposed for theSiCN. As shown in Fig. 5, the material contains a Si-con-taining area and a free carbon area. Within the Si-contain-ing area, there are SiN4 domains, SiC4 domains andinterfacial SiCN3 domains. The free carbon phase containsa large number of nanosized graphite crystals. The demix-ing process of the SiCN3 will lead to the following changeswithin the structure of the Si-containing area: (i) anincrease in the amount of SiC4 and SiN4 (which can occurvia coarsening the existing domains, or nucleating newdomains, or both); and (ii) an increase in the total interfa-cial area between the SiC4/SiN4 and SiCN3 domain. Mean-while, we can assume that the structure of the interfacesbetween different Si-containing domains remainsunchanged. Since there is no net chemical bonding changefor the demixing reaction, the measured enthalpy change

Fig. 5. Schematic showing the structure of the polysilazane-derived SiCNceramics.

must come from the change in the structure. The mixedSiCN3 should have higher strain energy than SiC4 andSiN4 due to the structural distortion in the former, andhence the transition from SiCN3 to SiC4 and SiN4 will leadto a decrease in enthalpy. On the other hand, the increasein the interfacial area can result in an increase in enthalpy(because the rather disordered interface should have higherenthalpy). It is likely that in our case the enthalpy changedue to the increase of the interfacial area is much higherthan that due to the release of the distortion energy, andthus the net change in the enthalpy is positive. As forentropy change, the transition from SiCN3 to SiC4 andSiN4 can lead a decrease in entropy due to the increase inorder. On the other hand, the increase in interface will leadto an increase in entropy. In our case, the interface-inducedentropy increase played a dominant role over the transi-tion-induced entropy decrease, leading to a net increasein entropy.

3.3. Structural evolution in free carbon area

The structural evolution of the free carbon phase wasfirst examined using Raman spectroscopy (Fig. 6). Thespectra obtained from the samples annealed at differenttemperatures exhibit similar shapes, containing a D peakat �1350 cm�1 resulting from the breathing modes of sp2

carbon atoms in rings, and a G peak at �1600 cm�1 result-ing from in-plane bond stretching of sp2 carbon [40–49].Additionally, several minor peaks are also observed forsome spectra, such as a D” peak at 1500 cm�1 indicatingthe existence of amorphous carbon in samples, and a D’peak at 1620 cm�1 from the disordered graphitic lattice[42,43].

The Raman spectra were analyzed by curve-fitting themusing the Lorentzian function for the D peak and the Bre-it–Wigner–Fano function (BWF) for the G peak [40]. Thepeak positions, the full width at half maximum (FWHM),and the intensity ratio ID/IG were obtained from the fittingand are listed in Table 1. It can be seen that with increasingannealing temperature, both the G and D peaks up-shifted.

Fig. 6. Raman spectra of the SiCN pyrolyzed at different temperatures.

Table 1Curve-fitting parameters for the SiCN ceramics pyrolyzed at differenttemperatures.

Pyrolysis temperature (�C) 1000 1100 1200 1300G peak position (cm�1) 1550 1611 1612 1608D peak position (cm�1) 1352 1339 1354 1360G peak FWHM 138 69 61 79D peak FWHM 193 130 71 65ID/IG 1.48 1.26 1.77 2.04La (nm) 1.54 1.42 1.69 1.81

Fig. 7. Lateral size of the free carbon nanoclusters as a function ofannealing temperature.

Y. Chen et al. / Acta Materialia 72 (2014) 22–31 27

The up-shift of the G peak, leading to the conjunction of Gand D’ peaks, was due to the transition from disorderedcarbon to nanocrystalline carbon [44]; while the up-shiftof the D peak was caused by the increase in the numberof ordered threefold aromatic rings. These, together withthe decrease in the FWHM of both the D and G peaksand the disappearance of D” peak, suggest that the freecarbon phase is graphitizing from amorphous carbon tonanocrystalline graphite with increasing annealing temper-ature [42,44].

The D-to-G peak intensity ratio was used to calculatethe size of the free carbon nanoclusters. For highly disor-dered graphite, the lateral size (La) can be calculated usingthe Ferrari-Robertson equation [42,45]:

ID=IG ¼ C0ðkÞ L2a ð8Þ

where C0(k) = C0 + kC1 (C0 = �12.6 nm and C1 = 0.033)and k is the wavelength of the excitation source. The resultsare listed in Table 1. It can be seen that the size decreasedwhen annealing temperature increased from 1000 to1100 �C, and then continually increased with a further in-crease in annealing temperature. The results are consistentwith the Ferrari model [42,44], i.e. for the transformationof amorphous carbon to nanocrystalline graphite, the sp3-to-sp2 transition and rearrangement of distorted aromaticrings to six-membered rings occurs between 1000 and1100 �C, resulting in the shrinkage of lateral size; a furtherincrease in temperature results in the in-plane growth ofnanopolycrystalline graphite.

The in-plane growth of the nanocrystalline graphite canbe treated as a 2-D grain growth process. Accordingly, thegrowth rate of the graphite nanocrystals can be describedby following equation [50]:

dLa

dt¼ A

Lað9Þ

where A is a temperature-dependent constant and t is theannealing time. A can be related to annealing temperatureby [50]:

A ¼ Ao exp � G

RT

� �; ð10Þ

where Ao is a temperature-independent constant and G* isthe activation energy. Combining Eqs. (9) and (10),

L2a / exp � G

RT

� �ð11Þ

Fig. 7 is a plot of ln(La2) vs. 1000/T. The linear relation-

ship suggests that the in-plane growth of the graphite nano-crystals indeed follows a 2-D grain growth process. Theactivation energy for the growth of the nanocrystallinegraphite is then calculated from the slope of the curve tobe 43.6 kJ mol�1.

The structural evolution of the free carbon phase wasfurther examined using EPR. Previous studies revealed thata fairly large amount of unpaired electrons (danglingbonds) were formed during pyrolysis due to the release ofhydrogen and methane [25,30,46,47]. For SiCN ceramics,three kinds of unpaired electrons have been observed: sili-con-dangling bond, carbon-dangling bond in free carbonand carbon-dangling bond in Si-containing phases[47,48]. The EPR signals for these three dangling bondsoverlap to a considerable extent and thus cannot be distin-guished in low-frequency EPR. Hence, EPR spectra of ourSiCN ceramics were first obtained using G-band EPR at406.4 GHz with an external standard (Fig. 8) [49]. It isfound that regardless of annealing temperature, all SiCNceramics show only one peak, with g = 2.00258 ±0.00002, which is the signal for carbon-dangling bonds inthe free carbon phase. There are no silicon-dangling bonds(g = 2.005) and carbon-dangling bonds in the Si-containingphase (g = 2.0032) in the SiCN studied here.

The SiCN ceramics were then analyzed using X-bandEPR at 9.5 GHz to determine the concentration and loca-tion of the carbon-dangling bond (Fig. 9). It is interestingto see that the EPR spectra obtained from the SiCNannealed at lower temperatures show two peaks: a widepeak with a width of �8.5 gauss and a narrow peak witha width of �2 gauss (Fig. 9a). This “split” characters ofthe EPR signal has been observed in coal due to electrontransfer between small and macromolecular scales [51];and in the conversion from diamond to graphite due todangling bond defects and conduction p electrons[52–54]. Most recently, Sehlleier et al. [46] reported a simi-lar phenomenon for 13C- and 15N-labeled polymer-derivedSiBNC ceramics pyrolyzed at 1400 �C. They observed that

Fig. 8. G-band EPR spectrum of the SiCN ceramics pyrolyzed at 1100 �C.The asterisks () are the signals of the standard.

Fig. 9. (a) X-band EPR spectra of the SiCN ceramics pyrolyzed atdifferent temperatures; and (b) the relative density of the carbon-danglingbond as a function of annealing temperature.

28 Y. Chen et al. / Acta Materialia 72 (2014) 22–31

the samples had a small sharp peak on top of a broad onewhen X-band EPR was recorded at 2 K, but the spectra ofthe same samples measured at higher temperature containonly a single peak. They assigned the sharp peak to elec-tron spin coupling to the 14N nucleus and attributed the

broad peak to a doublet due to unpaired electrons near13C. For our system, the G-band EPR clearly showed thatthe paramagnetic centers are only related to the unpairedelectrons of carbon in the free carbon phase, not in theSi-containing phase; and there is no dangling bond fromother atoms. Therefore, the two-peak phenomenonobserved is likely similar to that for the pure carbon mate-rials and closely related to graphitization of the free car-bon. The wide peak could be attributed to thenonbonding p electrons at the zigzag edge of the carbonclusters [52–54]; while the narrow peak could be assignedto the localized dangling bonds associated within the car-bon clusters. With increasing pyrolysis temperature, thenarrow peak gradually decreased and disappeared, eventu-ally merging with the broad one; meanwhile the intensity ofthe signal decreased. To quantify the change in the dan-gling bond, the relative concentration of unpaired electronswas obtained by a second-integration of the X-band EPR(Fig. 9b), which shows that the dangling bond concentra-tion decreased rapidly between 1000 and 1100 �C, and thenslowly increased with a further increase in annealing tem-perature. The results are keeping with the Ferrari model[44,46]. At lower temperatures, the carbon is highly disor-dered, and thus contains a large amount of both types ofdangling bonds. The transition from amorphous carbonto nanocrystalline graphite that occurs between 1000 and1100 �C caused a significant increase in order, and thus arapid decrease in dangling bond density. In particular,the graphitization process reduced the defects within thecarbon clusters (corresponding to the narrow peak) morethan those at the edge (corresponding to the broad peak),leading to the disappearance of the narrow peak. With afurther increase in the annealing temperature, the majorchange in the carbon structure is the in-plane growth ofgraphite nanocrystals, which cause a decrease in theboundary area (edge area), and thus a further decrease indefect concentration (mostly due to the decrease in thedefects at the edge).

According to the above discussion, the decrease in thedefect density in the temperature range 1100–1350 �C canbe related to the increase in the lateral size of the graphitenanocrystals. Assuming the graphite nanocrystals have asquare shape, the number (N) of the nanocrystals in a unitvolume of the free carbon phase is:

N ¼ 1

L2ah

ð12Þ

where h is the average thickness of the nanocrystals. Thetotal area of the edge (S) can then be calculated by:

S ¼ 4NLah ¼ 4

Lað13Þ

Assuming the number of defects per unit edge area is no,the total defect per unit volume (nd, defect density) will be:

nd ¼ noS ¼ 4no

Lað14Þ

Table 2Enthalpy and entropy changes in different areas with increasing annealingtemperature.

Si-containing area Free-carbon area Interface area

Enthalpy change Positive Negative NegativeEntropy change Positive Negative Negative

Y. Chen et al. / Acta Materialia 72 (2014) 22–31 29

This equation suggests that the defect density is inver-sely proportional to the lateral size of the graphite nano-crystals. Inserting Eq. (11) into Eq. (14), we obtain:

nd / expG

2RT

� �ð15Þ

Fig. 10 plots the logarithm of the defect density vs. thereciprocal of the annealing temperature. The “activationenergy” (=slope�R) calculated using the data between1100 and 1350 �C is �20.5 kJ mol�1. This value is veryclose to half of the activation energy (G*) for the lateralgrowth of the graphite nanocrystals, as predicted by Eq.(15), suggesting that the decrease in the density of the car-bon-dangling bond within the temperature range is indeedmainly due to the decrease in the interface (edge) area.

3.4. Final remarks

A previous study using a high-temperature oxide meltsolution calorimetric method revealed that the enthalpyof polysilazane-derived SiCN decreases with increasingannealing temperature [23]. This seems to contradict ourresult in Eq. (6a), which shows that the enthalpy of demix-ing SiCN is positive. This contradiction can be rationalizedby considering the overall structural changes with increas-ing annealing temperature. According to previous studiesand the results of current paper, structural changes in theSiCN can be classified into three types: changes in Si-con-taining area, changes in the free carbon area, and changesin the interface area between the Si-containing area and thefree carbon area. As shown in Eq. (6a), the structuralchanges in the Si-containing area caused an increase inenthalpy. On the other hand, the structural changes inthe free-carbon area, including the graphitization and thedecrease in the defect density, should cause a decrease inenthalpy. While changes in the interface area (betweenthe Si-containing area and the free-carbon area) cannot

Fig. 10. A plot of the dangling-bond density as a function of annealingtemperature.

be detected in the current paper, such changes must involveordering and defect reduction, leading to a decrease inenthalpy. The above analysis on the changes in enthalpyis summarized in Table 2. The overall decrease in theenthalpy [23] indicated that the overall enthalpy change islikely dominated by the changes in the free-carbon areaand the interface area.

We can perform a similar analysis on the change inentropy (Table 2). While the entropy change in the Si-con-taining area was measured to be positive, it must be nega-tive in both the free-carbon area and the interface areasince the structural changes in these two areas causedincreases in order and decreases in defects. Therefore, theoverall change in entropy will be determined by the relativeentropy changes in the different areas.

Since the overall Gibbs free energy change is related tothe changes in enthalpy and entropy by DG� = DH� �DS �T, the above analysis can lead to the following interest-ing deductions. First, the free-carbon concentration canhave a significant effect on the stability of polymer-derivedSiCN. An increase in the amount of free carbon will makethe change in overall entropy less positive (or more nega-tive); hence, the change in Gibbs free energy will be lessnegative (or more positive). Therefore, the material withhigher free carbon concentration will be more stable vs.temperature as compared to those with lower free carbonconcentration. This conclusion is consistent with the exper-imental observation that PDCs with more free carbonexhibited higher stability [55]. Second, the relative amountof different Si-containing phases can have a significanteffect on the stability of polymer-derived SiCN. As dis-cussed above, the positive changes in the enthalpy andentropy of the Si-containing area are mainly due to theincrease in the interface areas between different Si-contain-ing phases. If the amount of SiN4 and/or SiC4 increasesand these become the dominant phases (which can occurby annealing at higher temperatures or by changing thechemistry of precursors), demixing the SiCxN4�x could leadto a decrease in the interface area, thus resulting in a neg-ative change in the enthalpy and entropy. Consequently,the Gibbs free energy will increase with increasing anneal-ing temperature, leading to the materials becoming morestable.

4. Conclusions

High-temperature annealing induced structural evolu-tions in polysilazane-derived amorphous SiCN ceramicswhich were quantitatively studied using 29Si solid-state

30 Y. Chen et al. / Acta Materialia 72 (2014) 22–31

NMR, Raman spectroscopy and EPR. In particular, thestructural changes in the Si-containing area and the freecarbon area were studied separately. According to theNMR study, the major structural change in the Si-contain-ing area was a transition from SiCN3 to SiC4 and SiN4. Thedemixing process can be described by an equilibrium reac-tion, 4SiCN3 = SiC4 + 3SiN4. The reaction showed posi-tive changes in enthalpy and entropy, which is likely dueto the increase in the interface area between differentphases. Raman spectra suggested that the structural evolu-tion in the free carbon area involved graphitization of theamorphous carbon and lateral growth of the nanographite.The lateral growth of the graphite nanocrystals followed a2-D grain growth mechanism. EPR results revealed thatthe materials contained only carbon-dangling bonds aspoint defects. Two types of defects were identified: car-bon-dangling bonds at the edge and in the interior of thenanographite. The defect concentration decreased withincreasing annealing temperature. By comparing their acti-vation energies, it was found that the decrease in the defectconcentration was mainly due to the lateral growth of thenanographite. The energetics associated with the structuralchanges in different phases were discussed in terms theireffects on the stability of the materials.

Acknowledgment

This work is supported by the U.S. Department ofEnergy (DE-FE0007004) and the US National ScienceFoundation (CMMI-1301099 and CMMI-0927441). A por-tion of this work was performed at the National HighMagnetic Field Laboratory (NHMFL), which is supportedby National Science Foundation Cooperative AgreementNo. DMR-0654118, the State of Florida, and the USDepartment of Energy. The authors thank Dr. A. Ozarow-ski for performing the high-field EPR measurements, andDr. P.G. Fajer for the X-band EPR characterization.

References

[1] Colombo P, Mera G, Riedel R, Soraru GD. J Am Ceram Soc2010;93(7):1805.

[2] Riedel R, Ruwisch L, An L, Raj R. J Am Ceram Soc1998;81(12):3341.

[3] An L, Riedel R, Konetschny C, Kleebe H-J, Raj R. J Am Ceram Soc1998;81(5):1349.

[4] Zimmermann A, Bauer A, Christ M, Cai Y, Aldinger F. Acta Mater2002;50(5):1187.

[5] Riedel R, Kienzle A, Dressler W, Ruwisch L, Bill J, Aldinger F.Nature 1996;382(6594):796.

[6] An L, Wang Y, Bharadwaj L, Zhang L, Fan Y, Jiang D, et al. AdvEng Mater 2004;6(5):337.

[7] Wang Y, Fei W, Fan Y, Zhang L, Zhang W, An L. J Mater Res2006;21(07):1625.

[8] Wang Y, Fei W, An L. J Am Ceram Soc 2006;89(3):1079.[9] Wang Y, Fan Y, Zhang L, Zhang W, An L. Scripta Mater

2006;55(4):295.[10] Wang Y, Jiang T, Zhang L, An L. J Am Ceram Soc 2009;92(7):1603.[11] Ryu H, Wang Q, Raj R. J Am Ceram Soc 2010;93(6):1668.

[12] Zhang L, Wang Y, Wei Y, Xu W, Fang D, Zhai L, et al. J Am CeramSoc 2008;91(4):1346.

[13] Verbeek W. Production of shaped articles of homogeneous mixturesof silicon carbide and nitride. US Patent 3853567; 1974.

[14] Yajima S, Hayashi J, Omori M. Chem Lett 1975;4(9):931.[15] Colombo P, Paulson TE, Pantano CG. J Am Ceram Soc

1997;80(9):2333.[16] Liew L, Zhang W, An L, Shah S, Luo R, Liu Y, et al. Am Ceram Soc

Bull 2001;80(5):25.[17] Liu Y, Liew L, Luo R, An L, Dunn M, Bright V, et al. Sensors Actuat

A: Phys 2002;95(2–3):143.[18] An L, Xu W, Rajagopalan S, Wang C, Wang H, Fan Y, et al. Adv

Mater 2004;16(22):2036.[19] Iwamoto Y, Volger W, Kroke E, Riedel R, Saitou T, Matsunaga K. J

Am Ceram Soc 2001;84(10):2170.[20] Ziegler G, Kleebe H-J, Motz G, Muller H, Traßl S, Weibelzahl W.

Mater Chem Phys 1999;61(1):55.[21] Gregori G, Kleebe H-J, Brequel H, Enzo S, Ziegler G. Non-Cryst

Solids 2005;351(16–17):1393.[22] Seitz J, Bill J, Egger N, Aldinger FJ. Eur Ceram Soc 1996;16(8):885.[23] Gao Y, Widgeon SJ, Tran TB, Tavakoli AH, Mera G, Sen S, et al.

Scripta Mater 2013;69(5):347.[24] Morcos RM, Mera G, Navrotsky A, Varga T, Riedel R, Poli F, et al.

J Am Ceram Soc 2008;91(10):3349.[25] Traßl S, Motz G, Rossler E, Ziegler G. J Non-Cryst Solids 2001;293–

295:261.[26] Widgeon S, Mera G, Gao Y, Stoyanov E, Sen S, Navrotsky A, et al.

Chem Mater 2012;24(6):1181.[27] Saha A, Raj R, Williamson DL. J Am Ceram Soc 2006;89(7):2188.[28] Durr J, Lamparter P, Bill J, Steeb S, Aldinger F. J Non-Cryst Solids

1998;232–234:155.[29] Kleebe H-J, Stormer H, Traßl S, Ziegler G. Appl Organomet Chem

2001;15(10):858.[30] Traßl S, Suttor D, Motz G, Rossler E, Ziegler G. J Eur Ceram Soc

2000;20(2):215.[31] Varga T, Navrotsky A, Moats JL, Morcos RM, Poli F, Muller K,

et al. J Am Ceram Soc 2007;90(10):3213.[32] Dhamne A, Xu W, Fookes BG, Fan Y, Zhang L, Burton S, et al. J

Am Ceram Soc 2005;88(9):2415.[33] Li Y-L, Kroke E, Riedel R, Fasel C, Gervais C, Babonneau F. Appl

Organomet Chem 2001;15(10):820.[34] Brodie N, Majoral JP, Disson JP. Inorg Chem 1993;32:4646.[35] MacKenzie KJD, Smith ME. Multinuclear solid-state NMR of

inorganic materials. New York: Pergamon Press; 2002.[36] Berger F, Muller A, Aldinger F, Muller K. Zeitschrift fur Anorgan-

ische und allgemeine Chemie 2005;631:355.[37] Hartman JS, Richardson MF, Sherriff BL, Winsborrow BG. J Am

Chem Soc 1987;109(20):6059.[38] Carduner KR, Carter RO, Milberg ME, Crosbie GM. Anal Chem

1987;59:2794.[39] Sarkar S, Gan Z, An L, Zhai L. J Phys Chem C 2011;115:24993.[40] Ferreira EHM, Moutinho MVO, Stavale F, Lucchese MM, Capaz

RB, Achete CA, et al. Phys Rev B 2010;82:125429.[41] Tuinstra F, Koenig JL. J Phys Chem 1970;53:1126.[42] Ferrari AC, Robertson J. Phys Rev B 2000;61:14095.[43] Sadezky A, Muckenhuber H, Grothe H, Niessner R, Poschl U.

Carbon 2005;43:1731.[44] Ferrari AC, Robertson J. Phys Rev B 2001;64:75414.[45] Zickler GA, Smarsly B, Gierlinger N, Peterlik H, Paris O. Carbon

2006;44:3239.[46] Sehlleier YH, Akdogan Y, Verhoeven A, Roduner E, Jansen M.

Chem Mater 2008;20:7563.[47] Andronenko SI, Stiharu I, Misra SK. J Appl Phys 2006;99:113905.[48] Ishii N, Kumeda M, Shimizu T. Jpn J Appl Phys 1981;20:L673.[49] Stoll S, Ozarowski A, Britt RD, Angerhofer A. J Magn Reson

2010;207:158.[50] Thompson CV. Annu Rev Mater Sci 1990;20:245.

Y. Chen et al. / Acta Materialia 72 (2014) 22–31 31

[51] Keneko T, Sasaki M, Yokono T, Sanada Y. Novel characterizationof argonne premium coals with electron donors or acceptors byelectron paramagnetic resonance spectroscopy. Magnetic resonanceof carbonaceous solids. Washington, DC: American Chemical Soci-ety; 1992.

[52] Prasad BLV, Sato H, Enoki T, Hishiyama Y, Kaburagi Y, Rao AM,et al. Phys Rev B 2000;62:11209.

[53] Andersson OE, Prasad BLV, Sato H, Enoki T, Hishiyama Y,Kaburagi Y, et al. Phys Rev B 1998;58:16387.

[54] Tomita S, Sakurai T, Ohta H, Fujii M, Hayashi S. J Chem Phys2001;114:7477.

[55] Mera G, Riedel R, Poli F, Muller K. J Eur Ceram Soc 2009;29:2873.