Embed Size (px)
Citation preview
Developing a Macro-scale SiC-cladding Behavior Model Based on Localized
Mechanical and Thermal Property Evaluation on Pre- and Post-Irradiation
SiC-SiC Composites
Fuel Cycle Research and Development Peter Hosemann
University of California, Berkeley
CollaboratorsOregon State University
University of Illinois, Urbana Champaign
Frank Goldner, Federal POCYutai Katoh, Technical POC
Project No. 15-8439
Project Title: Developing a macro-scale SiC-cladding behavior model based on localized
mechanical and thermal property evaluation on pre-and post-irradiation SiC-SiC composites
Reporting Frequency: Final Report December 2018
Recipient: UC-Berkeley
Award number: CFA-15-8439
Awarding Agency: DOE
Working Partners: University of California, Berkeley
Oregon State University
University of Illinois Urbana Champaign
General Atomics
University of Oxford
Principal Investigator:
P. Hosemann, Dep. of Nuclear Engineering; UC-Berkeley
Title: Associate Professor
Phone: 510-717-5752
Email: [email protected]
Collaborators:
J. Tucker, B. Bay Oregon State Univ.;
D. Cahill, Univ. of Illinois Urbana Champaign;
C. Deck, General Atomics;
S.G. Roberts and D. Armstrong Univ. of Oxford, UK
Table of contents
Proposal overview ........................................................................................................................................... 3
Proposal abstract ................................................................................................................................................. 3
Milestone deliverables outlined by DOE Work Package ...................................................................................... 3
Collaboration roles and responsibilities ............................................................................................................... 3
Project status .................................................................................................................................................. 4
Executive summary of achievements ................................................................................................................... 4
Budget status ....................................................................................................................................................... 4
Communication and reporting status .................................................................................................................. 4 Journal publications ...................................................................................................................................... 5 Conference publications & presentations .................................................................................................... 5
Technical review .................................................................................................................................................. 6 SiC/SiC samples ............................................................................................................................................. 6
2.4.1.1 SiC/SiC manufacturing ........................................................................................................................... 6 2.4.1.2 SiC/SiC sample preparation ................................................................................................................... 8 2.4.1.3 SiC/SiC Microstructure characterization ................................................................................................ 9
Small scale mechanical testing (SSMT) ....................................................................................................... 14 2.4.2.1 Nano-indentation ................................................................................................................................ 15 2.4.2.2 Micro-cantilever testing ....................................................................................................................... 17 2.4.2.3 Fiber Pushout testing ........................................................................................................................... 23 2.4.2.4 Micro-pillar compression ..................................................................................................................... 27
Spatially resolved thermal conductivity ...................................................................................................... 35 Irradiation Effects ....................................................................................................................................... 40
2.4.4.1 Irradiated samples ............................................................................................................................... 40 2.4.4.2 Neutron irradiation: ............................................................................................................................ 43 2.4.4.3 Ion Irradiation ...................................................................................................................................... 45
Macroscopic composite characterization ................................................................................................... 50 2.4.5.1 X-ray Tomography ............................................................................................................................... 50 2.4.5.2 Hysteresis testing of unidirectional mini-composites .......................................................................... 53 2.4.5.3 Mechanical testing of woven composites ........................................................................................... 57
FEA model development............................................................................................................................. 60 Microscale Models ...................................................................................................................................... 60
2.4.7.1 Micropillar Compression Model .......................................................................................................... 60 2.4.7.2 Fiber Pushout Model ........................................................................................................................... 61
Mini-composite model ................................................................................................................................ 62 2.4.8.1 Modelling Approaches ......................................................................................................................... 63
Summary and Conclusions .............................................................................................................................. 66
References ..................................................................................................................................................... 68
List of Figures ..................................................................................................................................................... 71
Proposal overview
Proposal abstract
Silicon carbide is being investigated for accident tolerant fuel cladding applications due to its good neutronic
performance, high temperature strength, exceptional stability under irradiation, and reduced oxidation
compared to Zircaloy under accident conditions. The development and investigation of these materials is
particularly important in the light of the Fukushima event and subsequent emphasis in DOE on accident
tolerant fuel (ATF) concepts. In this work, we propose to develop and improve upon existing small-scale
mechanical and thermal characterization methods, including micro-cantilever bend tests and fiber push-out
tests, and time-domain thermo-reflectance measurements. These techniques will be applied to evaluate
micro-scale properties of SiC-SiC composite constituents (matrix, fibers, interphase). The results will be
coupled with meso-scale fiber and void structural information and be used as input to develop a
comprehensive, finite element-based model based on constituent properties. The goal of this multi-scale
model is to be used to predict anisotropic mechanical and thermal cladding behaviors. The model will be
benchmarked against measurements of bulk properties and validated for radiation-damaged materials by
utilizing irradiated SiC-SiC composites available from General Atomics. The characterization approach
developed under this program will have particular application in the property prediction of reactor irradiated
materials, as only small volumes need to be tested, minimizing associated hazards and costs. The micro-scale
test methods developed here will be used along with meso-scale structural data to inform a finite element
model of the full SiC-based cladding system. This model will capture fabrication and irradiation effects and
importantly include for the first-time separate material data for each system component. The incorporation
of these micro-scale effects will lead to a more accurate model of SiC-SiC composite behavior, enabling
further advancements in the development and design of SiC-based materials for improved fuel cladding
performance.
The aim of this work is to develop advanced localized material characterization techniques to directly
measure mechanical and thermal properties of the individual constituents of SiC-based claddings at the
relevant micro-scale. SiC-SiC composites will be evaluated before and after irradiation and these results will
be coupled with macro-scale properties and microstructural information in order to provide the input
parameters for a comprehensive finite element model, which will be developed in this program.
Milestone deliverables outlined by DOE Work Package
1) Developing micro bend bar testing, fiber push out testing, and other small-scale mechanical testing
techniques: micro-pillar compression and nano-indentation.
2) Finish the manufacturing of all SiC samples
3) Receiving neutron irradiated SiC/SiC from GA
4) Spatially resolved thermal conductivity measurements on the as produced materials.
5) Determination of thermal conductance of interfacial layers of as produced materials.
6) Tomography measurements of the SiC materials
7) Macroscopic property measurements
8) Simulation of macro-scale component testing for comparison with experimental data.
9) Delivering a FEA model that incorporates all data generated
10) Localized mechanical property measurements on the irradiated material
11) Final Report
Collaboration roles and responsibilities
PI university (UC Berkeley) was to conduct small-scale testing on the composite constituents with specific
focus on debond and friction parameters as a function of interface characteristics. In situ SEM/FIB micro-
mechanical testing and TEM microstructural investigation were performed. The in situ synchrotron X-ray
based tomography work was also led by this institution.
Industrial Partner I (General Atomics) was responsible for sample and component manufacturing and
fabricated SiC-SiC composites with a range of composite parameters. Industry partner I supported the bulk
mechanical property characterization, fracture analysis, and X-ray tomography work to investigate porosity
size and distribution effects.
University Partner I (Oregon State University) was to develop a FEA model for the SiC-SiC composite.
Mechanical, thermal, microstructural properties were compiled from the literature, and by conducting first
principles calculations when needed. The FEA model was pursued to predict macro-scale component
behavior. University Partner I also developed sub-models to describe interface failure mechanisms as a
function of irradiation and fabrication.
University Partner II (University of Illinois UC) was responsible for high spatial resolution measurements
of the thermal conductivity of fibers, matrix, and interfacial phases by time-domain thermoreflectance
(TDTR) between room temperature and 300°C. Ion beam irradiation experiments will also be carried out to
complement the studies of materials damaged in reactor environment.
Foreign Institution I (University of Oxford) was responsible for development of high temperature
micromechanical testing techniques. Focused ion beam machining was used to manufacture micro-cantilever
samples in SiC- SiC composite materials supplied by Industrial Partner I. These cantilevers sampled the key
microstructural features of the materials, including, fiber, matrix and fiber matrix interphase. Failure
mechanisms were studied using high resolution SEM and correlated with observed local microstructure.
Project status
Executive summary of achievements
• All milestones and deadlines have been met.
• 1D, 2D, and layered composites were successfully fabricated with variable parameters including fiber
type and PyC interface thickness. This showed versatility in manufacture capability and enabled
investigation of relationship between constituent properties and bulk behavior. Additional unique micro-
composites and specialized laminates were fabricated for enhanced fundamental studies.
• Microstructure characterization was successfully applied to evaluate macroscopic composite density and
constituent grain structure. This was achieved through XCT, HRTEM, EBSD and Transmission
Kickuchi Diffraction (TKD).
• Constituent thermo-mechanical properties were successfully evaluated using micro-cantilever testing,
fiber pushout, micro-pillar compression, nano-indentation, and thermoreflectance measurements. This
was done for pristine and irradiated samples.
• Macroscopic four-point bend, tube expansion, c-ring compression, and unidirectional mini-composite
tests were successfully carried out to understand bulk failure behavior.
• Constituent level FEA models of micro-pillar compression and fiber pushout were developed to capture
interface failure mechanisms.
• Macroscopic FEA models were developed that attempt to integrate the interface failure characteristics
and follow observe geometry and porosity of unidirectional mini-composites.
Budget status
All budgeting items were kept valid.
Communication and reporting status
Communication of the research that evolved within this project was a critical component for advancing SiC-
SiC ATF technologies. It allowed for development and understanding of experiment techniques, and opened
discussion for the impact and direction of these materials. The primary methods for dissemination under this
project included journal publications and conference proceedings.
Journal publications
E. K. Pek, J. Brethauer, D. G. Cahill: Hish Spatial Resolution Thermal Condictivity Mapping on SiC-SiC
Composites: planned submission to Acta Materialia (2019)
Y. Zayachuk, P. Karamched, C. Deck, P. Hosemann, D. E. J. Armstrong: Linking microstructure and local
mechanical properties in SiC-SiC fiber composite using micromechanical testing, submitted for publication
in Acta Materialia, (2019)
J. Kabel, Y. Zayachuk, D. E. J. Armstrong, T. Koyanagi, K.A. Terrani, Y. Katoh, and P. Hosemann. Ceramic
composites: A review of toughening mechanisms and demonstration of micropillar compression for interface
property extraction. J. Mater. Res. 33, (2018).
J. Kabel, Y. Yang, M. Balooch, C. Howard, T. Koyanagi, K.A. Terrani, Y. Katoh, and P. Hosemann: Micro-mechanical
evaluation of SiC–SiC composite interphase properties and debond mechanisms. Composites, Part B Eng. 131, 1–18
(2017).
Conference publications & presentations
ANS ICAPP 2016: J. Kabel, M. Balooch, D. Frazer, C. Deck, T. Koyanagi, K. Terrani, P. Hosemann.
Characterization of SiC-SiC Composites for Application in Current and Advanced Reactors. ANS
Transactions (2016).
ANS Winter Meeting 2016: J. Kabel, M. Balooch, Y. Yang, T. Koyanagi, K. Terrani, P. Hosemann. SiC-
SiC Composite Interphase Evaluation via Small Scale Mechanical Testing. ANS Transactions (2016).
ANS Annual Meeting 2018: J. Kabel, P. Hosemann, T. Koyanagi, Y. Katoh. Micro-Cantilever Testing of
Environmental Barrier Coatings on CVD SiC. ANS transactions.
ACS ICACC 2017: J. Kabel, M. Balooch, Y. Yang, T. Koyanagi, K. Terrani, P. Hosemann. Influence of the
PyC Interphase on Mechanical Properties and Failure Mechanisms of SiC-SiC Composites. ICACC
presentation (2017).
ACS ICACC 2018: J. Kabel, C. Deck, T. Koyanagi, Y. Katoh, I. Love, P. Hosemann. Micro-Mechanical
Characterization of the PyC interphase in SiC/SiC composites. ICACC presentation (2018).
ACS ICACC 2019: J. Kabel, P. Hosemann, T. Koyanagi, Y. Katoh. Small scale mechanical testing of dual-
purpose barrier coatings on CVD SiC. ICACC presentation.
TMS 2017: J. Kabel, M. Balooch, Y. Yang, T. Koyanagi, K. Terrani, P. Hosemann. Micro-Mechanical
Interphase Property Evaluation for SiC-SiC Composites. TMS presentation.
TMS 2017: Y. Yang, J. Kabel, M. Balooch, T. Koyanagi, K. Terrani, P. Hosemann. A TEM study of
microstructure of Hi-Nicalon Type S SiC composite beyond ultimate shear strength. TMS presentation.
TMS 2018: J. Kabel, C. Deck, T. Koyanagi, Y. Katoh, P. Hosemann. Experimental characterization of micro-scale
failure mechanisms and governing properties in SiC/SiC composites. TMS presentation.
TMS 2019: Joey Kabel, Darren Parkison, Christian Deck, Yutai Katoh, Peter Hosemann. Application of
small-scale mechanical testing to link interface properties to macroscopic hysteresis behavior of SiC/SiC
composites. TMS poster.
ANS Env. Deg. 2017: Ian Love, Peter Hosemann, Ph.D., Joey Kabel, Brian K. Bay, Ph.D., Julie D. Tucker,
Ph.D. “Image Analysis of SiC-SiC Composites for Quantification of Mechanical Properties under Tensile
Loads.” Env. Deg. Published paper.
ANS Env. Deg 2019: Joey Kabel, Takaaki Koyanagi, Djamel Kaoumi, Yutai Katoh, Peter Hosemann.
Interface characterization of candidate dual-purpose barrier coatings for SiC/SiC accident tolerate fuel
cladding. Env. Deg. Published paper.
NuMat 2016: Y. Zayachuk, D. E. J. Armstrong, S. G. Roberts, C. Deck, P. Hosemann. Microstructural and
micromechanical characterization of SiC-SiC fiber composites for fuel cladding applications. NuMat
presentation
NuMat 2018: Y. Zayachuk, D. E. J. Armstrong, C. Deck, P. Hosemann. Micromechanical characterization
of the radiation and temperature effects on SiC-SiC fiber composites for accident-tolerant fuel applications.
NuMat presentation.
MRS Fall Meeting 2016: Y. Zayachuk, D. E. J. Armstrong, S. G. Roberts, C. Deck, P. Hosemann.
Microstructural and micromechanical characterization of SiC-SiC fiber composites for fuel cladding
applications. MRS Fall poster.
MRS Fall Meeting 2018: Y. Zayachuk, A. Hussey, D. E. J. Armstrong, C. Deck, P. Hosemann. Radiation
and temperature effects on localized properties of sic-sic fiber composites – a micromechanical study. MRS
presentation.
Materials for Extreme Environments workshop 2017: Y. Zayachuk, D. E. J. Armstrong, S. G. Roberts,
C. Deck, P. Hosemann. Micromechanical characterization of SiC-SiC fiber composite for accident tolerant
fuel cladding applications. MEE presentation.
Nanomechanical Testing in Materials Research and Development VI, ECI conference series 2017: Y.
Zayachuk, D. E. J. Armstrong, S. G. Roberts, C. Deck, P. Hosemann. Micromechanical testing of SiC-SiC
fiber composites for nuclear fuel cladding applications. ECI presentation.
ICFRM 2017: Y. Zayachuk, D. E. J. Armstrong, S. G. Roberts, C. Deck, P. Hosemann. Development of
microstructural and micromechanical tools for characterization of composite materials with application to
SiC-SiC fiber composites. ICFRM presentation.
Radiation damage workshop 2018: Y. Zayachuk, D. E. J. Armstrong, S. G. Roberts, C. Deck, P. Hosemann.
Microstructural and micromechanical characterization of SiC-SiC fiber composites for accident-tolerant fuel
applications. Poster.
Thin Film and Small Scale Mechanical Behavior, Gordon Research Seminar 2018: Y. Zayachuk, A.
Hussey, D. E. J. Armstrong, C. Deck, P. Hosemann. Micromechanical characterization of the radiation and
temperature effects on SiC-SiC fiber composites for accident-tolerant fuel applications, GRS presentation.
Thin Film and Small Scale Mechanical Behavior, Gordon Research Conference 2018: Y. Zayachuk, D.
E. J. Armstrong, C. Deck, P. Hosemann. Microstructural and micromechanical characterization of SiC-SiC
fiber composites for accident-tolerant fuel applications, GRC poster.
Technical review
This section of the final report is written in pieces that correspond to the milestones outlined in the DOE
work package, shown in 1.2 . Justification and description of experimental techniques and corresponding
results are presented. Discussion and interpretation of findings are embedded within each section. Section 3
presents a summary and conclusion to open discussion for the impact of our findings and next steps to
continue advancing SiC/SiC as accident tolerant fuel.
SiC/SiC samples
2.4.1.1 SiC/SiC manufacturing General Atomics supplied several monolithic and composites samples fabricated by chemical vapor
infiltration (CVI) techniques. The macro-scale SiC/SiC composite samples provided were composed of
woven SiC fiber with a pyrolytic carbon (PyC) interphase layers and SiC matrix. In some of these samples,
the PyC first layer thickness was varied (10nm-500nm). In addition, several single tow samples were
provided to the program. The single tow test specimens (mini-composites) consist of a single fiber tow
(composed of a bundle of several hundred fibers) which has been infiltrated via CVI similar to the woven
composites. In this case, the fiber is held straight during processing, and is not woven or wound into a planar
or tubular geometry. Figure 1 shows the mini-composite samples while Figure 2 shows the flat woven panel
samples. The single tow specimen are especially interesting since the uncertainty on the mechanical
properties was reduced by not considering the weave structure. These samples were utilized in the in-situ
tomography tensile testing and constituent level fiber-matrix interaction tests allowing accurate modeling
composite behavior. There are two premier nuclear grade fiber types; Tyranno SA3 (Ube Industries, Ltd.,
Ube, Japan, ‘‘SA3’’ hereafter) fiber 1 and Hi-Nicalon Type S (Nippon Carbon Co., Tokyo, Japan, ‘‘HNLS’’
hereafter) fiber 2,3. Both are nuclear-grade generation III SiC fibers that are characterized by near-
stoichiometric chemical composition with low oxygen but free carbon concentrations, as well as high
crystallinity. The primary difference is that SA3 fibers contain larger grains (d~50-400nm) and consequential
increased root mean squared (RMS) surface roughness (RMS~8.04nm) as compared to HNLS fiber with
grain size (d~10-50nm) and RMS ~2.33nm 4–6. Y. Katoh 7 tabulates the detailed attributes of these composites
on the bulk scale in table 1 and table 7 respectively. Early mini-composites produced by GA also contained
ZMI fibers, which are slightly less stoichiometric SiC but similar roughness to SA3 fibers. Most composites
from GA contained SA3, however some studies were carried out using HNLS mini-composites fabricated by
Hypertherm and passed to the PI institution through supporting NSUF RTE’s. Although not explicitly part
of this NEUP program, the comparisons between fiber types are enlightening and worth highlighting
throughout this document.
Figure 1 - Single tow mini-composites with ZMI SiC fibers (left) and SA3 fibers (right), fabricated by GA that were
provided for characterization.
Figure 2 - Six planar SiC-SiC panels fabricated for this work (top row are baseline panels with a single PyC layer at
the fiber matrix interface; bottom row are panels with multi-layer interphase).
Additional samples including mini-composites of different fiber and interface type, as well as monolithic
SiC/PyC layers samples were provided to allow for diverse and comprehensive investigation of constituent
impact on composite behavior. One unique and novel sample configuration fabricated and provided in this
work was a single fiber composite, referred to as a micro-composite. This sample type is composed of a
single, individual fiber (extracted from a fiber tow), which was then coated with pyrolytic carbon interphase,
followed by CVD SiC coating. An initial fabrication attempt at these single fiber samples produces a “fuzzy”
SiC coating appearance, but this was resolved and a subsequent attempt produced individual SiC fibers with
uniform, continuous SiC coatings. Examples of these are found in Figure 3.
Figure 3 – Examples of single-fiber mico-composites. Initial fabrication produced a non-continuous SiC coating (left)
while subsequent fabrication produced individual fibers coated with a uniform CVD SiC later (right).
2.4.1.2 SiC/SiC sample preparation Samples for microstructure characterization and micro-mechanical testing require high quality surface
preparation, allowing for improved analyses. Quality surface finishes exhibit very little scoring, limited
chipping, and minimal edge-rounding. Most samples were prepared using slow speed saws and cut into ~1
mm thick slices for subsequent experiments. Samples were mechanical ground and polished with SiC grit
papers, then polished to 1 µm with diamond suspensions. General structure of the composite samples was
characterized using Scanning Electron Microscopy (SEM), and advanced crystal structure and compositional
data was acquired with tools including electron backscatter detection (EBSD), Transmission Electron
Microscopy (TEM), and Energy Dispersive Spectroscopy (EDS). Figure 4 below presents a general view of
the polished surface of the sample.
Figure 4 - SEM images of the polished surface of the composite: (a) general appearance, (b) close-up of the vicinity of
fiber bundle, (c) close-up with individual fibers, denoted by arrows.
The fiber bundles are clearly visible, both the ones perpendicular to the surface and those parallel to it. The
presence of large pores between the bundles is evident. Figure 4b presents a more detailed view of the vicinity
of one of the fiber bundles. It emphasizes the lower-scale intra-bundle porosity with pores located between
the individual fibers within the bundle. It is important to note that at larger magnification individual fibers
can be easily resolved, which enables the precise aiming at the regions of interest (individual fibers,
interphases or inter-fiber regions) for micromechanical testing and lifting out the TEM samples. Each fiber
is very noticeable thanks to the presence of the dark thin “halos” surrounding each fiber, which are the PyC
interlayers. The fibers are typically coated with a PyC deposition layer to allow for micro-crack deflection
and fiber sliding, while the carbon concentration observed at the fiber center (Figure 4c) is an artifact of fiber
production. The distribution of porosity within the bundle is non-uniform. For the most part pores are
concentrated close to the center of a bundle, where the areal concentration of the fibers is highest. Conversely,
at the periphery of a bundle individual fibers are further apart, and each is surrounded by a monolithic block
of matrix material with relatively few or no large pores, as shown.
Micro-mechanical testing, which is performed in situ the SEM sometimes requires that two surfaces of the
bulk sample be polished to create a sharp edge so that testing can be viewed directly by the electron beam. a
Below in Figure 5 is an SEM image showing this type of edge polish. The chipping at the edge of the fiber
is an artifact of polishing, and is generally milled away during focused ion beam (FIB) fabrication of micro-
pillars etc. In order to verify the surface quality of the polished samples, an AFM (purchased with NEUP
infrastructure funds) was used to evaluate roughness. Figure 5 shows an AFM height image collected with
the new tool. It can be seen that the RMS roughness of the image is 26.3nm including the regions of
preferential graphite removal, which was deemed negligible roughness for most experiments.
Figure 5 - (Left) SEM image of an edge polish required for pillar fabrication. (Right) Height image of a SiC fiber in the
SiC matrix. It can be seen that the graphite containing regions were removed preferentially during polishing.
2.4.1.3 SiC/SiC Microstructure characterization The microstructure was investigated in greater detail using Scanning Transmission Electron Microscopy
(STEM) and EBSD. Figure 6 below presents a general view of a typical area with several relatively close-by
fibers (with separation of the order of ~10 µm, typical for the periphery of the fibers) and surrounding matrix
region.
Figure 6 - STEM image of a typical region containing fibers and surrounding matrix, arrow indicates submicron-sized
pores.
An inherent weakness of TEM, which only allows analysis of a small localized area (corresponding to a
single lamella) at a time, making the investigation of the large areas difficult and time-consuming. At the
same time, such an investigation is essential since microstructure is very non-uniform in composite SiC. An
alternative way of large-scale microstructure characterization is provided by image quality (IQ) mapping
using EBSD, where each location is assigned a value reflecting the ability of an indexing software to detect
Kikuchi bands, producing a grayscale map. Figure 7 (Left) below presents the IQ map of a typical area at the
periphery of a fiber bundle, where fibers are relatively far apart and large regions of bulk matrix material are
present.
Pores Interphase
M
a5 µm
b
Figure 7 – (Left) Image quality map of a large area, including several fibers and a surrounding region of the matrix,
artifact CVI rings are visible. (Right) STEM image of the matrix material.
It is evident that the microstructure of the matrix is complex, being organized in several hierarchical levels.
First of all, it consists of the grains elongated in the direction away from the fibers. Such directional growth
is typical for SiC grown by CVI. The close-up of these grains is shown in the right image of Figure 7. Their
width is ~ 100 – 200 nm, and their length is in the order of ~ 2 – 5 µm. The matrix grains contain fine structure
within them, consisting of series of parallel dark fringes, running approximately normal to the grain growth
direction. Comparison with similar structures presented in the literature indicates that these are likely stacking
faults. In the regions of matrix in between fibers, the growth of grains is constrained by the presence of other
grains, originating at other fibers and growing in the opposite direction. Porosity is found at these interfaces
running along the boundary between the domains, parallel to the direction of the fibers, as shown in Figure
6. These grains are organized into ring-like structures. Notably, each fiber is surrounded by several nested
rings. These are thought to originate from the discontinuities during the CVI process, with each ring
corresponding to a consecutive stage of the matrix growth process. Finally, these rings form domains of
matrix associated with each fiber. Multiple levels of microstructural organization give rise to multiple types
of boundaries within the matrix. Elongated radial grains are separated by the regular grain boundaries. Nested
concentric rings are separated by the inter-ring boundaries, and finally matrix domains are separated by the
domain boundaries. And indeed, these domain boundaries are the regions associated with the submicron-
sized porosity observed in STEM images. The presence of these boundaries, together with the submicron-
sized porosity, within the matrix indicates that the matrix itself is non-uniform, giving rise to the possibility
that mechanical properties, and in particular fracture properties, are locally non-uniform as well. This has
been investigated by micromechanical testing, as described later on.
Similarly, the structures of the pyrolytic carbon interphase and fiber were investigated. First the PyC layer,
ranging from 10 to 1000nm depending on deposition parameters often exhibits two different phases. Most
commonly found to be a turbostatic nanocrystalline graphite-like structure, and sometimes found to be
amorphous in nature. As will be discussed later, irradiation increase the amorphous character and can lead to
porosity. TEM foils were fabricated to characterize the initial state of the PyC, Figure 8 shows SA3 (A) and
HNLS foils (C), high resolution TEM (HRTEM) images of the PyC first layer (B&D), and diffraction pattern
at that location (E).
Rings
10 µm 500 nm
Figure 8 - A) SA3 TEM foil. B) SA3 first layer PyC/fiber interface with oriented graphitic structure. C SEM-STEM
image of HNLS foil. D) HNLS first layer PyC/fiber interface with oriented graphitic structure. E) Diffraction ring
pattern at HNLS first layer PyC with characteristic graphite rings (002).8
The SiC and carbon phases were clearly distinguishable. ImageJ was used to measure the interplanar spacing
of the carbon phase and found values on the order 3.57Å ±0.01 Å, compared to 3.44Å of that reported for
PyC graphitic structure 9,10. Additionally, the diffraction ring pattern in Figure 8E displays characteristic
graphite rings 11. Figure 9 shows electron energy loss spectroscopy (EELS carried out using a Tachnai 200
TEM) detecting graphite-like signals for the interphase and the secondary phase (concentrated darks spots at
fiber center, observed in Figure 4 and Figure 10c of the SA3 fiber). The spectrum of pure graphite was
acquired as a fingerprint to compare to the deposited PyC. The interphase as well as the secondary carbon
phase of the fiber (dark speckles as seen in figure 3) showed a strong pi*(π*) peak at approximately 285 eV
and a sigma*(σ*) peak at about 295eV, matching the pure graphite scan as well as that in literature 12. This
indicates carbon in the interphase and the secondary phase is more graphitic than diamond-like or amorphous.
Previous literature found only sigma peak for the secondary phase in HNLS fiber 13,14.
Figure 9 - EELS spectra comparison of the pyrolytic carbon interphase and secondary carbon phase of the SA3 fiber
showing graphitic structure
Based on this evidence and qualitative visualization of the HRTEM images, it is believed that the carbon is
mostly graphitic in nature and expresses some degree of basal plane ordering with respect to the fiber
direction. Figure 8B suggests that the SA3 may exhibit more graphitic ordering compared to the HNLS
samples. However, evaluation of the HRTEM images across all three foils suggested that the arrangement of
the graphitic feature was rather random and lacks significant dependence on fiber type.
The close-up of the SA3 fiber structure is presented in Figure 10. It consists of essentially equiaxed grains,
with typical grain size in the 50 – 200 nm range, comparable to the width of the elongated grains in the
matrix. Microstructure is also radially non-uniform within the fiber – grain size is smallest in the center of a
fiber (as denoted by a dashed line) and increases towards its periphery. It is interesting to note that the grain
boundaries are often speckled with inclusions (Figure 10b, dark in this imaging mode, indicated by arrows).
Figure 10c shows a cross-section of an entire fiber, and it can be seen that the radial distribution of these
inclusions is non-uniform, with their density being highest in the center of a fiber and reducing towards the
periphery. EDS elemental analysis (performed in a line scan mode, with line crossing one of the inclusions
as denoted in the Figure 10a indicates that in comparison to the bulk of SiC grains, these dark region are
enriched in C and depleted of Si Figure 10d, which means that they are likely to be precipitates of excess
carbon.
Figure 10 - TEM images: (a) cross-section of a fiber, dashed line corresponds to the central axis of it; (b) fiber
material with the location of EDX line scan denoted, arrows indicate inclusion at the grain boundaries; (c) radial
distribution of inclusion within the fiber; (d) EDS linescane – signals of C and Si.
Obtaining crystallographic information using EBSD was difficult, due to relatively poor quality of observed
patterns. It was sufficient for pattern quality mapping, as presented above, but not for actual indexing and
orientation mapping. This was overcome by the use of TKD. This technique implements electron-transparent
samples, which can be used for TEM as well, with detected Kikuchi patterns produced by the electrons that
were transmitted through the sample. Due to the reduction of interaction volume, the lateral resolution
significantly increases. It can be further improved by the use of recently developed on-axis configuration
(on-axis TKD), where the scintillator is located directly beneath the sample in TEM-like geometry.
Crystallographic indexing with TKD confirmed that both matrix and fiber consist predominantly of cubic β-
SiC.
A sample containing the arrangement of fibers similar to the one presented in Figure 7 is not optimal for the
determination of crystallographic texture. Since grains in the matrix are positioned radially, the growth
direction of most of them doesn’t coincide with any one reference direction. For texture analysis samples of
a different kind were used: these were lifted out of the vicinity of a fiber that is close to normal to the surface.
This way, the plane of the sample contains parallel matrix grains all aligned in the direction of growth. Figure
11a presents an example of a STEM image of one such sample, containing both fiber and matrix materials.
Grain growth direction here is denoted by the horizontal arrow, which is in turn aligned to a horizontal X-
axis during TKD mapping. The insert in Figure 11 presents the corresponding TKD orientation map,
displayed in IPFX coloring (inverse pole figure showing a horizontal direction) – that is, map is color-coded
in the reference direction which is aligned with the matrix grain growth direction.
2 µm 500 nm
5 µm
Figure 11 -(a) TEM image of the microstructure and the corresponding IPFX map. The arrow represents the direction
of grain growth. The vertical dashed line in the IPFX map represents the location of the interlayer; (b) pole figure of
the matrix (top row) and fiber (bottom row).
An orientation map of this kind allows determination of the crystalloghraphic texture, which needs to be
defined separately for fiber and matrix. The corresponding pole figures for matrix and fiber regions are
presented in Figure 11b. A significant difference in texture between matrix and fiber materials is evident.
Matrix is noticeably textured in the growth direction (horizontal here), with preferred orientation being
<111>, with no texture in other directions. This is in agreement with previously reported observations for the
preferred growth direction in CVD grown SiC15. Meanwhile, there is no discernible texture in the fiber
material. TKD orientation map reveals microstructural features of a different kind, not easily resolvable by
the direct TEM imaging. Figure 12 shows the close-up of a specific region of the map that contains areas of
different crystallographic orientation, as indicated by differing colours in IPFX map. The insert shows a
misorientation profile measured along the line in the map (misorientation here is relative to the point of the
origin of this line). It is evident that high-angle grain boundaries with 60° misorientation are present between
M
aF
i
2 µm
A
B
Matrix
Fiber
the dissimilar regions; in addition they also have a common -{111} plane, and can therefore be identified as
coherent Σ3 twin boundaries, which are known to occur in SiC ]
Notably, grains in the matrix contain an additional finer structure, manifesting itself as series of parallel
fringes, running perpendicularly to the direction of grain growth Figure 12a. In fact, originally these were
assumed to be twin boundaries, however, TKD orientation mapping proved that this is not the case, and they
are not associated with twinning (length scale at which they are present, several tens of nanometers at most,
is much shorter that that corresponding to twin boundaries – hundreds of nanometers, Figure 12b. High-
resolution TEM doesn’t show any crystallographic boundaries corresponding to these fringes Figure 13– it
is evident that atomic rows are continuing uninterrupted through them. The contrast due to which they are
visible might be related to the presence of internal strain within the grains – insufficient to cause twinning,
since crystallography is everywhere the same, but still manifesting itself in TEM images.
Figure 12 - (a) Close-up of a typical matrix microstructural, with fringes within grains visible; (b)
comparison of a TKD orientation map and a TEM image of the same area. Red rectangles denote the twin
boundaries, and it is evident that the fringes do not directly correspond to these.
Figure 13 - High-resolution TEM image of an area within the matrix. The line follows is parallel to atomic
rows, and it is evident that areas of varying contrast (fringes) are not associated with different crystal
structure.
Small scale mechanical testing (SSMT) A primary goal of this NEUP was to develop and progress small scale mechanical testing techniques to
characterize mechanical properties at the constituent level. A variety of techniques were leveraged including
nanoindentation, micro-cantilever bending, micro-pillar compression, and fiber pushout testing. These
techniques proved extremely useful to probe the composite constituents at their engineered length scale; with
a b
typical fiber diameter is ~8-10 µm, and PyC layers <1 µm. Using nanoindentation it was possible to place
indents wholly within a fiber, and in the bulk matrix for elastic modulus and hardness measurment. FIB-
machined micro-cantilevers and micro-pillars were manufactured sufficiently small to place them within
individual fibers, or across the PyC interface, thus enabling measuring fracture properties – fracture stress
and toughness - of the fibers, matrix, and interface. Micro-pillar compression and fiber pushout testing enable
characterization of the shear and friction properties of the PyC interface. Identifying constituent level
properties allowed for unique insight to macroscopic behavior, and layed the ground work for the next
generation of mechanistic modelling.
2.4.2.1 Nano-indentation In order to characterize micromechanical properties over the large areas and in particular to compare the
properties of the major composite’s constituents – fibers and matrix – nanoindentation was used, performed
with the Agilent XP nanoindenter in the continuous stiffness measurement (CSM) mode with 2 nm amplitude
and 42 Hz frequency, using Berkovich tip16. Depth dependences of hardness and elastic modulus were
determined using the Oliver-Pharr method for every depth step. Calibration of the area function of the tip
was performed via calibration indents in fused silica reference sample. Here indents of 300 nm depth were
placed along the line with 1.5 µm spacing between them. Figure 14 shows an example of linear scan used,
running both through matrix and fiber. Nanoindentation measurements were performed using in CSM mode,
where hardness and elastic modulus are recorded as a function of depth, as shown below:
Figure 14 - (a) Example of a line of indents, crossing a matrix region and one of the fibers; (b) typical
depth dependence of hardness (filled symbols) and modulus (hollow symbols) as a function of depth for the
indents placed within the matrix (solid lines) and close to the center of a fiber (dashed lines); vertical lines denote
the depth range used for averaging.
It should be noted that with indents located so close to each other it is conceivable that they might influence
the results of measurements from each other. In order to clarify whether this is taking place, two rectangular
4 by 4 indent arrays were placed in the matrix, with different spacings between them – 1.5 µm in one case
(i.e, same as in the line scans) and 15 µm in another, which served as a reference. No difference in the
averaged measured values of hardness and modulus from each indent was observed. This indicates that plastic
zone is extremely confined for SiC, and that the values of hardness measured in the described manner are
indeed representative of the material’s properties.
Figure 15 below shows the coordinate dependence of the hardness and modulus in the vicinity of a fiber,
averaged over multiple line scans in the depth range 120 and 280 nm were averaged, where surface effects
are minimized. Error bars here and in all subsequent plots represent the standard deviation of measured
values. It is evident that there is a significant difference between matrix and fiber materials. Hardness and
modulus of fibers tend to be significantly lower than those of matrix. Hardness of the matrix is ~40 GPa,
modulus is ~460 GPa. In addition, a fiber itself is very non-uniform, with values changing from those close
to the ones of matrix at the periphery, to much lower near the center of the fiber, i.e. ~19 GPa hardness, ~260
GPa modulus.
10 µm
This difference in mechanical properties can be explained by the presence of residual C at the grain
boundaries, as observed in STEM images. Indeed, the amount of C within the grain is variable – it is highest
around the center of the fiber and decreases outwards; at the very periphery of the fiber it is essentially absent.
This correlates well with the observed dependences of hardness and modulus – increasing C content,
decreases property values. For the fiber periphery, where the amount of C is low and material is near-
stoichiometric SiC – like the matrix – the values are close to that of the matrix. Reduced hardness and
modulus of the composite as a whole is therefore likely due to the lower hardness and modulus of this residual
C in the fiber.
Figure 15 - Hardness (filled symbols) and modulus (hollow symbols) as a function of distance from the
fiber center averaged over multiple line scans crossing the fibers; vertical dashed lines denote the typical
dimensions of a fiber.
Measured values of elastic modulus can be compared to the literature data of the macroscopic testing.
Measurements in the matrix can be compared to the measurements on polycrystalline CVD (chemical vapour
deposition) SiC. Nanoindentation yields a value of ~460 GPa, while the values found in the literatures are
ranging between ~415 and ~460 GPa. On the other hand, given the non-uniformity of the fiber properties, in
order to compare them to the results of tensile tests of the whole Tyranno SA fibers the values of modulus
should be averaged over the cross-section of the fiber; averaged modulus is ~365 GPa, with literature values
being 375-380 GPa. Thus values of elastic modulus obtained by nanoindentation are in good agreement with
those obtained by other means, such as tensile testing or impulse-excitation technique, in macroscopic
samples. This is important for the ability to develop a microstructurally-informed model of composite’s
behaviour, since representative data can be obtained from a very small volume of material. Producing
composite samples by CVI is a lengthy process, the duration of which scales with the amount of material
produced; thus, time minimization of this methodology ensures much faster development cycle, where
multiple different composite designs (e.g., with different CVI growth conditions, interfacial structures etc.)
can be manufactured and investigated quickly.
Temperature dependence of hardness and modulus were investigated using high-temperature nanoindentation
to evaluate mechanical properties at expected operation conditions. Using NanoTest Xtreme high-
temperature nanoindenter (from MicroMaterials Ltd), nanoindentation measurements were performed at
composite SiC in vacuum in the temperature range between room temperature and 700°C, with the results
presented below. Reduction of hardness from ~45 GPa to ~20 GPa has been observed, as well as reduction
of Young’s modulus from ~450 GPa to ~300 GPa.
Figure 16 - Results of high-temperature nanoindentation – temperature dependences of hardness and
modulus.
2.4.2.2 Micro-cantilever testing
For assessment of the fracture properties of the fibers and matrix materials, as well as interphases,
microcantilever fracture tests were applied. This technique was developed by UCB17 and UO18,19. It was
found that fiber matrix interaction as well as fiber strength and matrix strength can be evaluated in this
fashion. Triangular cantilevers were manufactured using FIB milling, leading to the geometry displayed in
Figure below:
Figure 17 - Typical cantilever in the (a) matrix;(b) fiber; (c) interphases; standard triangular cross-section
is visible.
Typical dimensions of cantilevers used in this study were ~8 µm in length, ~2 µm in width, and ~ 1.5 µm in
height. The fracture testing was performed using a Berkovich tip in an Agilent G200 nanoindenter equipped
with a nano-positioning stage. The load-displacement curve is recorded and using simple beam theory, these
were converted into stress-strain curves, which were used for determination of fracture stress and strain. This
type of geometry requires significant FIB time. However, it is a simple test with a high yield allowing almost
every test to be successful. The triangle cross-section was chosen for the simplified FIB procedure relative
to other cross-sections.
The notched cantilevers were used to measure fracture toughness, shown in Figure 18. In these cantilevers,
which also had triangular cross-section, the straight notches were made by FIB milling at low current (10
pA) normally to the top surface and to the length of the cantilevers. The typical depth of these notches was
~300 nm, and typical width ~100 nm.
3 µm 3 µm 3 µm
Figure 18 - Typical notched cantilever in the matrix SiC.
In order to test the fracture behavior of different constituents of the composite, micro-cantilevers were placed
at various characteristic locations – the interphases, matrix and fibers. It should be noted that, since properties
of fibers are radially non-uniform (as revealed by nanoindentation measurements), cantilevers are oriented
axially (as shown in Figure 17b) and placed as close as possible to the axis of the fiber. This way, they do
not cross regions of different properties within a fiber (and only contain fiber material), so that non-uniformity
within of properties along a cantilever is eliminated. On the other hand, the non-uniformity across the
cantilever is inevitable since microstructure and chemical composition of a fiber is inherently non-uniform.
Having cantilevers at the axis of a fiber ensures that the region which is most different from the matrix is
probed. Despite the cross-sectional non-uniformity, since all tested cantilevers were similar both in volume
and in their location within a fiber, the results obtained from them are still comparable.
Stress-strain curves were obtained from the measured load-displacement curves using equations derived from
the simple beam theory to evaluate the fracture stress of un-notched cantilevers:
𝜎 =24𝑃𝐿
𝑤ℎ2
휀 =2ℎ𝛿
𝐿2
Here P is load, δ is displacement, L – length of a tested part of a cantilever (i.e. the distance between the
cantilever’s base and position of a loading indent), w – width of a cantilever and h – its height.
During the microstructural study it has been established that the matrix consists of elongated grains, with
multiple types of boundaries within it (i.e. twin boundaries, grain boundaries, boundaries between nested
rings and between domains of matrix originating from different fibers). It is conceivable that fracture
behaviour of the matrix material can be different, depending on how the crack propagates relative to the local
direction of the grains. In order to investigate this, cantilevers in the matrix were placed at the specific
orientations with respect to the grains in the matrix, namely, in longitudinal direction (with axis of a cantilever
being parallel to the grains) and transverse direction (with axis of cantilever being normal to the grains).
Figure 19 plots the results of the fracture tests performed on cantilevers in different constituents of the
composite.
3 µm
Notch
Figure 19 - Results of fracture tests from cantilevers in different components of a composite; two values for
the matrix are from the longitudinal (hollow symbol) and transverse (filled symbol) cantilevers. Overlay –
a stress-strain curve of a typical test.
Data points presented are averaged values over several tests (6 for interphase, 7 for fiber, 5 for transverse and
6 for longitudinal orientations in the matrix), with error bars corresponding to their standard deviations. It is
evident that they demonstrate very different fracture properties. Interphases are the weakest points, with
lowest fracture stress, ~2.6 GPa, and strain at fracture, ~3%. Material of the fibers is somewhat stronger,
fracturing at ~6.4 GPa and ~6.5% strain, and matrix material is the strongest, ~21.5 GPa and ~12%, and there
does not appear to be a fundamental difference between the longitudinal and transverse cantilevers.
Fracture toughness was calculated using the equation for stress intensity factor for straight notches
determined using finite element compliance analysis20:
𝐾 =12𝑃𝐿𝑛
𝑤ℎ2∗ 1.12√𝜋𝛼 ∗ (1 + 0.123𝛼′ + 5.456𝛼′2 + 0.073𝛼′3 − 0.023𝛼′4)
Here P is the load, Ln the distance from the notch to the load point, w and h are respectively the cross-sectional
width and height of a cantilever, α the depth of a notch and α’ defined as α/h. This equation is valid α’<0.4,
as was indeed the case for the tested cantilevers. Figure 20 presents the values of fracture toughness obtained
in such a way for the cantilevers placed on different components of the composite.
Figure 20 - Fracture toughness of different components of a composite. Two values for the matrix are from
the longitudinal (hollow symbol) and transverse (filled symbol) cantilevers.
The number of tests performed was 4 for interphase, 5 for fiber, 3 and 2 respectively for the transverse and
longitudinal directions in the matrix. Similarly, to fracture stress, the comparison demonstrates that toughness
is highest in the matrix (~4.25 MPa*m1/2), again without a systematic difference between cantilevers at
different orientations relative to the direction of grain growth, lowest at the interphase (~0.8 MPa*m1/2),
with fibers being intermediate (~2 MPa*m1/2).
Published data are available on fracture toughness of monolithic (non-composite) SiC. They show large
scatter, with the reported values, derived from various macroscopic testing techniques, ranging from ~2.3 to
~5.1 MPa*m1/2. Literature data on Tyranno SA fibers give a value of fracture toughness of ~2.7±0.4
MPa*m1/2. Thus the values of KC obtained for matrix (being the most comparable to monolithic material)
and fibers are in reasonable agreement with the bulk data literature sources. On the other hand, fracture
toughness of the interphases cannot be measured using the bulk testing, but macroscopic testing on bulk PyC
has suggested Kc ~1 MPa*m1/2, which is in the range of these findings. Additionally, this supports crack
deflection observation in macroscopic composite testing where the interface material is chosen to have a
fracture toughness < ¼ of the surround matrix material. The details around this theory are presented in the
micro-pillar compression section. Typical fracture surfaces of cantilevers in different components presented
in Figure 21.
Figure 21 - Typical fracture surfaces of the notched cantilevers in different constituents: (a) interphase
(remaining base of the cantilever, visible in the image, is in the fiber); (b) fiber; (c) matrix – longitudinal
direction; (d) matrix – transverse direction. Note the geometry of the visible straight notch.
Figure 21a presents a cantilever that was at the interphase. The remaining part is in the fiber, with the visible
surface being a surface of that fiber. On the other hand, Figure 21b result from cantilevers that was within a
fiber, so that visible fracture surface shows the internal part of it. Note the presence of dark C inclusions in
Figure 21b, corresponding to near-central region of a fiber, and their absence in Figure 21a, which shows its
outermost region of a fiber. This is in accordance to the distribution of residual C, as presented in Figure 10c–
central part contains high density of carbon inclusions, while periphery essentially none. Of note is a
significant difference in fracture surfaces between cantilevers in the matrix in longitudinal and transverse
directions relative to the grains (Figure 21c and d). Transverse cantilevers’ fracture surfaces feature the
elongated grains running across the fracture surface. This is similar geometry to what is visible in STEM
images (like Figure 12). In contrast to this, the longitudinal cantilevers contain such grains face-on. Thus, the
corresponding fracture surface doesn’t show elongated features, but rather more equiaxed instead. These
500 nm 500 nm
500 nm 500 nm
Notch
features are cross-sections of the elongated grains – which are long but rather narrow. This becomes more
evident in the STEM images presented below.
In order to rationalize the observed fracture behaviour and the crack propagation, lift-out TEM samples were
manufactured out of tested cantilevers. Making these samples requires additional steps compared to those
used for microstructural characterization as described above. The reason is that during a typical fracture test,
the cantilever that is being tested is completely broken off when fracture occurs; this makes it difficult to
observe the exact relation between the location of the crack and the grain boundaries or other microstructural
features which might influence its propagation. Therefore, in order to perform such an imaging, it is
necessary to ensure that the fracture has been initiated but at the same time the crack has not propagated
through the whole thickness of a cantilever so that the beam is attached to its base by a narrow ligament.
When such a fracture is achieved, the entire fractured cantilever is first embedded in the protective Pt layer
so that the small ligament is not broken during the lift-out; then it is lifted out of the surface, and the area
surrounding the location of the crack is thinned down to electron transparency.
In order to have the beam hanging attached to a ligament, the test must be constantly monitored and manually
stopped as soon as load drop occurs in the load-displacement curve, indicating that fracture has been initiated.
In cases of cantilevers placed within a bulk matrix or a fiber this is difficult to accomplish because once
fracture is initiated, the crack propagates fast and it is hard to react and stop the load. In order to circumvent
this obstacle, cantilevers with the chevron notch were implemented21. In this geometry, the quadrangular
chevron is introduced into the beam. This way, the stress concentration is introduced where the fracture starts;
increasing width of the material ligament through which the crack is propagating ensures that it propagates
in a stable manner, and therefore slower. Using cantilevers with chevron notches it is indeed possible to stop
the loading in a manner described above and in such a way that cantilever is not broken off and can be
therefore lifted out for manufacturing of TEM sample. For the cantilevers at the interphases notch is not
necessary for such a study since there fracture occurs slower and so test can be stopped without completely
breaking the cantilever off even without it.
Figure 22 presents the TEM images of the fractured cantilevers, concentrating on the areas surrounding the
crack tip. Figure 22a shows the propagation of fracture at the fiber-matrix interphase. The area to the left of
the crack in the image is matrix, and area to the right is fiber, the dark stripe in between is the PyC interlayer.
Small amounts of the Pt deposited on the top surface of the cantilever to form a protective layer penetrated
into the crack, forming the brighter stripes along the crack walls. It is evident that the crack propagation is
confined to the interphase region, without straying into bulk material, neither matrix nor fiber. The insert in
the Figure 22a shows an EDX elemental map of the highlighted region. It can be seen that crack
predominantly runs along the boundary between the C interlayer and the SiC fiber, although it seems to have
initiated within the volume of the interlayer. Figure 22b shows the propagation of fracture within the notched
cantilever located entirely within the fiber. It appears that fracture has both intergranular and transgranular
features. In the location X it seems to be following the boundary of a large grain. In the location Y, however,
the crack is cutting through the grain’s volume. The reader should be reminded that fiber material contains
excess C, which decorates the grain boundaries and can be seen as dark regions (some of which denoted by
arrows); at the same time, distribution of this C is non-uniform, with some grain boundaries having more
while others are essentially free of excess C. Figure 22c shows the propagation of fracture within the notched
cantilever in the bulk matrix and oriented in the transverse direction, i.e. elongated grains in the matrix are
normal to the axis of cantilever. Here it can be seen clearly that fracture is transgranular, where the crack
does not follow any associated grain boundaries. Figure 22d shows the propagation of fracture within the
notched cantilever located entirely within the matrix and oriented in the transverse direction, i.e. elongated
grains in the matrix are normal to the axis of cantilever. Similarly, to the cantilever oriented transversely,
here the fracture is transgranular.
Figure 22 - TEM images of the fractured cantilevers: (a) at the matrix-fiber interphase; insert – EDX
elemental map of the highlighted area; (b) in the bulk fiber, arrow denote the C precipitates at grain
boundaries, X denotes the instance of intergranular crack propagation, Y the instance of transgranular
crack propagation instances of crack propagation; (c) in the bulk matrix, transverse orientation; here and
in (d) the arrows are to guide the eye along the crack; (d) located in the bulk matrix, longitudinal
orientation. Cantilever in (a) is unnotched, all the others are notched.
The observed differences in crack propagation can be correlated with the results of fracture tests. The
difference between the fiber and the matrix materials can be traced back to C precipitates. It can be seen that
the cracks in the fibers are, at least at some instances, associated with the grain boundaries, where C is present,
although this does not necessarily have to be the case – transgranular crack propagation has also been
observed. On the other hand, within the matrix, where these precipitates are absent, fracture is exclusively
transgranular. The fact that in the presence of carbon intergranular fracture is preferable implies that it makes
grain boundaries weaker, which can be suggested as a reason for lower fracture toughness, as is indeed
observed in cantilever tests.
Another correlation can be drawn between the results of tests performed at different orientations of
cantilevers relative to the orientation of grains within the matrix. The primary fracture mechanism is
transgranular regardless of how cantilevers are oriented. This should result in similar fracture properties, and
indeed no systematic difference in fracture toughness can be seen in the cantilever tests. This suggests that
various kind of microstructural features in Figure 12 are in fact of no particular significance from the point
of fracture behaviour of matrix. The primary result of the existence of the complicated microstructure within
the matrix is the introduction of large number of grain boundaries. But since fracture appears to be primarily
transgranular and not dominated by grain boundaries, it can be surmised that this is inconsequential and that
for the purposes of modelling matrix can be treated as a uniform and isotropic medium with respect to fracture
– but not homogeneous, since it contains porosity, including small submicron-sized pores. On the other hand,
for the fibers the assumption of uniformity is not correct, as there the presence of carbon at the grain
boundaries might have an impact on fracture behaviour, and distribution of carbon is non-uniform. The
M
a
500 nm 500 nm
500 nm 500 nm
C Si
amount of grain boundary C is high in the center and lower or almost negligible at the periphery, and thus
fracture properties should be considered as having radial distribution within the fiber (similarly to how
hardness is radially non-uniform).
In addition it should also be noted that the values of both fracture stress/strain and fracture toughness of the
matrix material exhibit significantly larger scatter compared to those of fibers and interphases. It is likely
related to the presence of submicron-sized porosity in the matrix, as revealed by TEM analysis. It is difficult
to determine the exact internal structure of each individual cantilever, however, it is conceivable that
cantilevers without pores are more resistant to crack propagation than those without; on the other hand, such
pores can serve as initial crack locations, and hence cantilevers without pores are likely to be more resistant
to crack initiation as well.
2.4.2.3 Fiber Pushout testing Nano-indentation and micro-cantilever bending tests provided elastic and fracture characteristics necessary
for understanding crack and initiation. Once initiated, often at sites of large porosity in the matrix, these
micro-cracks propagate normal to the loading axis, and ultimately deflect into mode II shear fracture at
fiber/matrix pyrolytic carbon interface. This is observed macroscopically via fiber pullout at the fracture
surfaces. Understanding the fundamental relationship between interface properties and macroscopic behavior
opens the door to rapid optimization of these composites. However, the fundamental material properties of
the interface are very challenging to deconvolute using macroscopic test methods22. As a result, fiber pushout
testing has received significant attention over the last the decades in an attempt to probe the interface more
directly. Results and discussion around pushout testing are presented below. The next section delves into
micro-pillar compression which was applied in a novel way in an attempt to refine some complexities
associated with fiber pushout testing.
Fiber pushout testing was conducted by UO on the same SiC/SiC composite samples provided by General
Atomics. The SiC/SiC composites were polished to <100um thickness and placed over a thin gap sampled
holder. The fiber push out testing was conducted with the UCB and UO Agilent XP Nanoindenter. The high-
power optical microscope was used to assure proper alignment when placing the indenter tip over the fiber.
The test proceeds by loading an individual fiber up to the point of slipping. There are 2 main material’s
parameters that can be examined through the fiber pushout method: the interfacial shear strength (ISS) and
the sliding friction coefficient between the two once they’ve already been debonded.
Push-out tests are widely thought of as relatively standard tests for SiC/SiC composites, but those reported
in the literature are rarely done to a statistically meaningful level. In fact, as stated by F. Rebillat, “Only a
few push-out experiments could be carried out on the micro-composites owing to the difficulties involved in
micro-composite handling, preparation and testing”. The issue they have is that the sample must be extremely
thin, 500μm for push-in and 200μm for push-out, which creates difficulties in sample handling. It is also hard
to make samples of that thickness perfectly flat. Therefore, the goal under this NEUP was set to perform a
significant number of tests, on the order of hundreds, so that a statistically sound analysis of obtained number
can be performed.
Most previous studies were using sharp Berkovich tip to perform push-out. This causes extensive plastic
deformation, and even cracking, of the fiber before the debonding can occur, which changes the shape of the
fiber and load distribution, and so changes the results that would be acquired. The measured load-
displacement curve contains contributions both from deformation and interfacial sliding, making analysis
difficult. Therefore in this study a flat punch conical tip was used, with a diameter of ~5 µm. This allows
placing a tip entirely within a fiber, at the same time largely preventing its deformation, as shown below in
figure Figure 23.
Figure 23 - SEM image of the impression left by a flat-punch tip on the fiber (indicated by an arrow),
indicating minor plastic deformation during loading.
In the previously reported studies the microstructure was usually not taken into account in the analysis of
these measurements, while it is well understood as explored in previous section of this report, that the
composites microstructure can be very non-uniform. The relevance of the considered problem – difference
of interfacial properties within a tow – is due to the composite growth method. The composite is formed by
growing the SiC matrix onto a SiC fiber reinforcement architecture. While various methods are employed
for this, depending on the application, chemical vapor infiltration (CVI) is a preferable growth method for
nuclear applications, since it produces very high purity β-phase SiC with good radiation resistance23. The use
of CVI leads to a significant multi-level porosity in a resulting material, with large pores between fiber tows,
as well as much smaller pores between the individual fibers within a single tow. Figure 24 presents the
comparison between the parts of an individual tow; showing the highly porous central part of a tow, where
majority of the fibers is in contact with a pore or another fiber, and relatively monolithic periphery, where
each fiber is well embedded in a uniform layer of matrix material.
Figure 24 - Comparison of (a) highly porous central part of a typical fiber tow, and (b) relatively
monolithic periphery
Figure 25 presents a collection of the experimental load-displacement curves, obtained in a number of
successful push-out tests (i.e. the ones where the displacement of a fiber was confirmed). The overall
difference between the center and the periphery of the tows is noticeable, however, large scatter within each
dataset is evident as well.
Figure 25 - Results of a number of push-out tests performed at the fibers (a) in the center and (b) at the
periphery of the tows.
Load-displacement curves were analyzed using a simple stress model, where the interfacial shear strength is
calculated from the equation
τ=F/2πRH
where τ is the interfacial shear strength (ISS), F is the plateau load, R is the fiber radius and H is the sample
thickness. Resulting values of ISS were calculated to be 69.44 ±21.9 MPa in the central part and 119.1±19.7
MPa at the periphery of a tow. This correlates well with the distribution of internal porosity as imaged in the
micrographs. Due to higher porosity, the fibers in the center of a tow are less well bonded, and therefore the
load needed to initiate interfacial debonding and to displace a fiber is lower. However, this observation
suggests that an “effective ISS” should be defined for a fiber, and this will be different depending on where
within a tow a particular fiber is located. This will take into account both the true strength of interfacial
bonding, as well as the influence of the local environment surrounding this particular fiber.
Values of ISS measured for peripheral fibers can be considered to be close to the true ISS, since the effects
of the environment can be excluded. On the other hand, the knowledge of the distribution of the effective ISS
within a tow might be useful for rationalization of the macroscopic mechanical properties based on
fundamental material parameters, since it allows a realistic description of the non-uniformity of the properties
of fiber tows.
The measured values of ISS show significant scatter, and it is necessary to discuss the sources of errors. Some
errors were readily quantified, whilst others lent themselves more readily to qualitative analysis.
The quantifiable errors considered were those used in Equation 1: fibre radius R, push-out load F and fibre
thickness H. The error in the diameter was entirely due to the error in reading the diameter from the images,
the push-out load errors were due to systematic errors in the nanoindenter and the errors in thickness were
due to the error in matching the thickness to the push-out area as well as reading the measurement from the
image. Predominant contribution among these comes from the error in thickness. The only way to measure
the thickness is to image the side of the sample. Under this orientation, it is impossible to see the push-out
tests, so there is an element of uncertainty of where the pushed-out fibres are located. As the samples were
not completely flat, this leads to a potentially large error, so 3 measurements of width were taken and
averaged. The accuracy in measuring the radius was to the nearest 0.1 microns on each side of the fibre. The
error in the load was estimated from the nanoindenter specifications. The approximate raw values, given in
Table 1 below are estimates of the mean value for each measurement.
Table 1 - Uncertainties in the push-out measurements.
Approximate raw error Approximate raw value Approximate percentage error
Radius (R) 0.1 μm 3.5 μm 2.9%
Load (F) 1 mN 250 mN 0.4%
Thickness (t) 15 μm 150 μm - 75 μm 10% - 20%
A B
Therefore, the error in the interfacial shear strength follows the equation:
𝛿𝜏 = 𝜏 [(𝛿𝑅
𝑅)
2
+ (𝛿𝐹
𝐹)
2
+ (𝛿𝑡
𝑡)
2
]
1
2 (Equation 2)
resulting in the following errors in the determination of ISS:
Table 2 - Fiber Pushout ISS values
ISS, MPa ISS error, MPa
Center 69.44 14.04
Periphery 119.1 24.07
It should be noted that the errors for the central region is smaller than the standard deviation of the results.
The standard deviation is not only an error, but also a representation of the differing environments that fibers
exist in.
The errors that are not quantifiable are related to the direction of the fiber under the surface and potential
relaxation of the matrix (where the hole in the matrix that constrains the fiber shrinks after push-out). It is
assumed in calculating ISS that the fibers run perfectly perpendicular to the surface. If the fiber is angled,
this may lead to an overestimation of all interfacial shear strengths. If the local environment beneath the
surface contains significant porosity or touching fibers, This may lead to a underestimation of peripheral
interfacial shear strength.
After pushout testing, the surfaces of the samples were sectioned using FIB, so that the resulting cross-
sections contain the near-interphase regions. This allowed imaging of the crack path in the vicinity of the
interlayer. Figure 26 shows an example of such a cross-section. Debonding can be observed, and the visible
crack is following the fiber-interlayer boundary, shown by the yellow arrows.
Figure 26 - Cross-section of the near-interphase region following a push-out, indicating the crack path.
To summarize, the presented data indicate that there is indeed large scatter in the values of ISS, measured
using pushout method. Therefore, in order to get statistically relevant data it is imperative to perform a large
number of tests. Flat-punch tip is shown to be a requirement as well, since it has been shown to cause only a
minor deformation of a fiber. Also, there is clearly a difference between different parts of a tow, which is
linked to the difference in local environment, first of all porosity but also other fibers, which has a certain
distribution within a tow as well – maximum close to its core, where measured ISS is lowest, and decreasing
towards the periphery, where measured ISS is correspondingly highest. Unfortunately it is very challenging
to capture these true defect density beneath the surface. This suggests the introduction of a concept of
“effective ISS”, different for different fibers within composite and even within an individual tow. Knowledge
of this is useful for understand the behavior of a composite as a whole; however, for characterization of
specifically the interfacial bonding, true ISS is a relevant parameter, and this can be determined at the
periphery of the tows, where the influence of porosity and other fibers is minimal or non-existent.
Matrix Fiber
Interphase
2.4.2.4 Micro-pillar compression
As elucidated in the fiber pushout section, there are some challenging assumptions and complexities
associated with extracting fundamental shear strength properties at the fiber/matrix interface. Micro-pillar
compression was investigated as an alternative and novel approach to simplify property extraction. It is able
to uniquely probe the interface at a length scale that reduces uncertainties relating to local environment. The
concept was first applied by Shih et al in 2013, and the efforts in this NEUP developed and expanded the
methodology tremendously to capture effects of fiber type and PyC thickness on debond properties. Micro-
pillar compression probes the combined properties of shear strength and frictional resistance to mode II shear
fracture that is dependent on normal stress and interfacial characteristics. This is particularly useful as it
provides properties as a function of residual stresses (normal to the radial fiber surface) that often exist in as
fabricated composites. Characterizing these properties is critically important for understanding crack
deflection and debond length for fiber pullout that enable pseudo-ductility for bulk composite. Much of the
following section is adapted from the two of the publications that came out of this NEUP project8,24.
For this micro-pillar investigation, composite samples from GA were evaluated. For a comprehensive
understanding of the influence of fiber type and PyC thickness, composite samples from ORNL were also
characterized. Samples from GA contained Tyranno SA3 (Ube Industries, Ltd., Ube, Japan, ‘‘SA3’’
hereafter) fiber 1. Samples from ORNL contained Hi-Nicalon Type S (Nippon Carbon Co., Tokyo, Japan,
‘‘HNLS’’ hereafter) fiber 2,3. Both are nuclear-grade generation III SiC fibers that are characterized by near-
stoichiometric chemical composition with low oxygen but free carbon concentrations, as well as high
crystallinity. A second set of composites provided by GA were manufactured with non-nuclear grade,
Tyranno ZMI fibers. These were critical to develop a relationship between micro and macroscopic properties
because these composites were also tested via x-ray tomography, described later in this report. The primary
difference is that SA3 and ZMI fibers contain larger grains (d~50-400nm) and consequential increased root
mean squared (RMS) surface roughness (rRMS~8.04nm) as compared to HNLS fiber with grain size (d~10-
50nm) and (rRMS ~2.33nm) 4–6, 7. The first layer (at fiber surface) of PyC was deposited via chemical vapor
infiltration at ~1000C with varying thicknesses (~50nm-1500nm), true interface thickness values varied
across different fibers. Although several of these composites exhibited five alternating layers of PyC/SiC
before full CVI SiC infiltration, failure was always observed at this first layer. As a result, categorization of
micro-pillars is denoted by this first layer thickness, associated fiber type, and irradiation condition. The
influence of these interface conditions and fiber type on bulk composite behavior including the modulus and
ultimate strength for the SA3 and HNLS composites can be found in Katoh’s work 5,25.
These samples were cut and polished into ~2 x 2 x 1mm3 sub-samples. Each was polished such that fibers at
the edge of each sample would have different angles with respect to the polishing plane. Micro-mechanical
structures were fabricated using the FEI Quanta 3D dual beam focused ion beam (FIB) instrument via
standard FIB milling techniques. Rough and finishing cuts were performed at 30keV with 15nA and 0.1nA
currents respectively as illustrated in Figure 27 (top). Critical measurements including the interface incline
and cross-sectional area were acquired at this stage of the experiment. Pillars were fabricated on pristine
cylindrical fibers of the same incline to allow for improved statistics. A diameter of 3um was chosen to reduce
the inherent curvature of the fiber with respect to the interface plane contained within the fiber. One major
assumption made throughout this micro-pillar analysis is that the fracture plane of the micro-pillar is flat.
Modelling efforts were explored to capture the effects of the small curvature that exists, discussed later on.
The pillars are assumed pristine such that the interface is fully bonded prior to compression testing. As a
result, the extracted properties at failure are characteristic of mode II shear. Unfortunately, due to instrument
feedback sensitivity, it is common for the pillar to be destroyed following instantaneous debond, making the
extraction of dynamic friction properties difficult. Figure 27(bottom) shows a subset of scope and variety of
interface conditions that were fabricated for micro-pillar compression. The complete test matrix is outlined
in Table 3.
Figure 27 - (Top) Typical micro-pillar fabrication process using FIB milling techniques, overlays show the
extraction of and interface incline and resulting interfacial area. (Bottom) Comprehensive set of micro-
pillar interface conditions that were tested. Notably there is a a wide range of PyC interface thickness and
fiber type.
Once fabricated, the pillars were tested in situ the SEM with the Hysitron PI-85 Pico-Indenter using a 5μm
diamond tip flat punch. Tests were performed under displacement controlled loading at 10nm/s. Carrying out
the pillar compression in situ with SEM imaging allows acquisition of live testing video to ensure alignment
and contact accuracy, shown in Figure 28. This capability is particularly beneficial to the fidelity of this
characterization technique and resulting properties. A typical load-displacement curve and corresponding
interphase failure for a pillar from sample subset SA3_700nm is displayed in Figure 28. Failure was defined
as the first instantaneous load drop (Pfail). In the case of Figure 28(right), failure was observed at ~ 3.3 mN.
Figure 28 - Snapshots from live mico-pillar compression testing of SA3_700nm PyC. (Left) Diamond flat
punch applying load. (Right) Post failure with fracture surface maintained.
Pfail was extracted from the load-displacement curve for each micro-pillar test. Each sample yielded data for
its respective incline plane, which were then group averaged and plotted using the Mohr-Coulomb (MC)
criterion for interface property extraction8,26,27.
The MC criterion states that ultimate shear strength of a brittle ceramic is resisted by two components. The
first is the interface debonding strength, τdebond, which is characteristic of the chemical bonding of the material
at the plane of fracture. The second is an internal friction resistance described by the coefficient, μi, which is
fundamentally considered to be different from the static friction of classical physics. Here it describes a
resistance to debond initiation that is dependent of the tortuosity of the evolving fracture path, and the applied
normal load. The equilibrium relationship between the failure shear strength and the opposing resistances is
described below, where the critical shear stress, τfailure, is set to equal the unknown τdebond, and the unknown
μi multiplied by the applied normal stress (NR):
τ𝑓𝑎𝑖𝑙𝑢𝑟𝑒 = τ𝑑𝑒𝑏𝑜𝑛𝑑 + 𝑁𝑅 ∗ 𝜇𝑖
Experimentally, the applied uniaxial stress state is resolved into the shear and normal stress at the interface
plane. At failure, the resolved stress state is then a function of the applied failure load, Pfail, interface incline,
θ, and the inclined interface cross-sectional area; A = πD2/4*cos(θ) shown in Figure 27(top). The above
equation can then be represented in a more useful form for experimental property extraction by:
𝑃𝑓𝑎𝑖𝑙 ∗ 𝑠𝑖𝑛𝜃
𝐴= τ𝑑𝑒𝑏𝑜𝑛𝑑 +
𝑃𝑓𝑎𝑖𝑙 ∗ 𝑐𝑜𝑠𝜃
𝐴∗ 𝜇𝑖
By testing across a range of interface incline, a unique failure stress state is acquired for each micro-pillar.
When averaged and plotted in Mohr’s space (τ vs σ), applying the linear relationship formulated above allows
the extraction of τdebond and μi as the τ-intercept and slope respectively. Figure 29a shows a schematic of the
micro-pillar and the corresponding force balance is overlaid to express the stress state of the loaded structure.
Figure 29 - (a) Representative schematic of pillar fabrication. (b) Resolved stress state and force balance
of a representative micro-pillar structure.
Prior to applying the MC criterion, preliminary evaluation of the ultimate failure shear strength, τfailure, was
carried out to quickly verify a fundamental property dependence on interface condition. A case example is
shown in Figure 30a for samples HNLS_50nm, HNLS_1300nm, and HNLS_50nm_1dpa. This raw data was
plotted with respect to the cosine of the incline angle. In doing so, the data is adjusted such that the slope of
the applied linear regression represents a single normalized value for the ultimate shear strength averaged
over the range of test inclines. Analysis of Figure 30a revealed a fundamental reduction in ultimate shear
strength between HNLS_50nm and HNLS_50nm_1dpa of ~80 MPa, and of ~300 between HNLS_50 and
HNLS_1300nm. This suggests that both irradiation and increased interface thickness results in a weaker
interface. Similar trends are observed for SA3 micro-pillars. In the case of increased thickness, it may be
expected that extrinsic parameters such as fiber surface roughness contribute less to the fracture path
tortuosity within the PyC, potentially reducing the stress required to fail.
Figure 30 - Comparison of raw data for control sample HNLS_50nm, HNLS_1300nm, and
HNLS_50nm_1dpa to show fundamental reduction in strength with irradiation and thickness.
To fully understand the observed change in ultimate strength, the MC criterion was applied to investigate
and separates the contribution of strength from chemical bonding and internal friction resistance. This
provides unique dependence on resolved normal stress, which can be directly related to residual stress in
composite following CVI fabrication.
The linear regression revealed the τ-intercept and slope that identify the contribution of strength from
chemical bonding (τdebond) and internal frictional resistance (μi) respectively. This method was applied across
all fiber type and interface conditions. Another example subset comparison showing SA3 vs HNLS is shown
in Figure 31.
Figure 31 - MC criterion applied to SA3 and HNLS micro pillars with varying PyC thickness. It is observed
that for the same thickness, SA3 shows fundamentally stronger debond shear strength.
Table 3 summarizes the extracted properties for all non-irradiated conditions, including the irradiated samples
which will be discussed in further detail in the irradiation effects section 2.4.4. These values are additionally
compiled in the master data table III for comparison against push-out and bulk tensile hysteresis loop
evaluations of interfacial properties.
Table 3 - PyC interface properties from micro-pillar compression
Fiber
τdebond (MPa)
Internal friction
coefficient. (µ)
Successful pillars
SA3_140nm 280 ±85 0.25 ±0.10 20
SA3_700nm 216 ±47 0.21 ± 0.07 12
SA3_140nm_4.5dpa* 100 0.27 5
ZMI_70nm 407 ± 10 0.17 ± 0.07 11
HNLS_50nm 266 ± 23 0.25 ± 0.03 17
HNLS_50nm_1dpa 160 ± 41 0.24 ± 0.10 16
HNLS_150nm 187 ±14 0.18 ±0.04 18
HNLS_1300nm 133 ± 14 0.17 ± 0.03 8
* Estimated assuming limited change in μi observed for HNLS
From the data collected, it was possible to graphically explore an empirical relationship of interfacial shear
properties as a function of fiber type and PyC thickness. Figure 32(left) plots the τdebond and µi properties of
different fiber types as a function of PyC thickness. An exponential fit was applied to extract an empirical
relationship that can be implemented into a mechanistically informed model to define the interface property
as a function of design parameters. Figure 32(right) presents literature the results found by fiber pushout data
that identifies a similar trend. The green and red overlaid circles point out the debond shear strength values
for ~150nm PyC for SA3 and HNLS respectively. Similar values were found previously by Yang28 which
aids in the validation of micro-pillar analysis. Micro-pillar compression also provides a unique advantage by
including a dependence on residual normal stress via the internal friction coefficient, that fiber pushout testing
fails to capture. This is critically important for prospects of mechanistic modelling that introduces composite
morphology and fabricated stress states.
Figure 32 - (Left) NEUP micro-pillar property values plotted versus PyC thickness. An empirical
relationship was fitted for application in modelling efforts. (Right) Literature fiber pushout debond strength
values as a function of PyC.
Looking at the MC property extraction trends, it was found that both the chemical bonding strength as well
as the frictional resistance was reduced compared as interface thickness is increase. Additionally, the SA3
interface properties show fundamentally increased values for τdebond and µi.. Below is a representative formula
to describe ultimate shear strength with dependence on residual normal stress, fiber type, and PyC thickness
multipliers and exponents can be extracted from Figure 32(left).
τ𝑓𝑎𝑖𝑙𝑢𝑟𝑒 = τ𝑑𝑒𝑏𝑜𝑛𝑑(𝑓𝑖𝑏𝑒𝑟, 𝑃𝑦𝐶) + 𝑁𝑅 ∗ 𝜇𝑖(𝑓𝑖𝑏𝑒𝑟, 𝑃𝑦𝐶)
In an effort to better understand the mechanisms responsible for these data trends, TEM and AMF techniques
were explored. TEM foils containing the fiber/matrix interphase were fabricated before and after testing. Pre-
test foils were fabricated using standard FIB trenching and thinning techniques on a randomly selected fiber-
matrix interphase within the bulk material. Post-failure TEM foils were prepared uniquely via platinum
encasement and subsequent thinning and lift out of the failed pillar. Primary trench cuts were performed at
30keV and 15nA milling current, final cleaning cuts were made using 5keV and 30pA polishing current.
Analysis of the interphases before and after slip provided insight as to how and where the interphase failed.
Understanding these mechanisms allowed us to evaluate the applicability of the failure criterion. Figure 33
shows a representative TEM foil of post tested pillars. TEM observations were conducted using a
CM200FEG instrument at 200keV. High resolution TEM (HRTEM) and diffraction analysis was carried out
under bright field operating conditions.
Figure 33 – (Top) example fracture micro-pillar suitable TEM foil fabrication and TEM dark field image of
foil with slipped HNLS pillar. (Bottom) HRTEM image of deformed PyC layer as a result of failure. 002
graphite-like planes are visible. And suggest fracture is occurring along those weak basal planes.
Several HRTEM images were taken along the top and bottom fracture surfaces. Figure 34 compares the PyC
layer for before and after failure. The original interface had a thickness of approximately 40nm, and the post
failure surfaces showed thicknesses on the order of 20nm, predicting that failure propagated within the PyC
layer. This was consistent across all foils.
Figure 34 - A) TEM image of HNLS interphase pre-failure. B) TEM image of PyC interface post failure
suggesting cohesive failure in the PyC layer.
From this TEM analysis, it was suggested that failure occurs cohesively within the semi-ordered first PyC
deposition layer on the fibers. In Figure 33(bottom) it can be seen that the graphite-like structure exhibits
stronger disorder compared to the unfractured interface in Figure 8. This may suggest that the semi-ordered
graphite layers were bent or rolled at some point during the compressive shear failure. The degree to which
failure occurred via tortuous slip along the PyC basal planes versus mixed modes of sliding plus bending and
folding is not fully understood. Although potentially an artifact of post failure sliding surface interaction, it
is not unreasonable to postulate the graphite-like layer deformation could occur simultaneously with slip, and
thereby influence the debond properties. Ultimately, this would suggest that interphase properties are a
function of cohesive failure mechanisms in PyC. Previous TEM analysis from literature has shown that CVI
B
PyC may exhibit stronger graphite-like ordering within its thickness as a function of distance from the fiber
surface 29. If consistent, it is hypothesized that the bond strength is reduced as the lattice structure approaches
that of highly oriented pyrolytic graphite (HOPG interlayer shear strength has been found on the order of
~140 MPa 30), where weak basal plane interactions govern the strength. In this case, fracture tortuosity would
also be reduced. However, generally failure was observed very close to the fiber side, even when the PyC is
very thick, and neither SA3 or HNLS show significantly different PyC disorder/alignment as a function of
thickness. This has led us to believe that the inherent surface roughness of the fiber that is impacting crack
path tortuosity, and has a stronger influence on the observed trends. For example, the RMS roughness for
SA3 fibers was found ~8nm and resulting peak roughness at ~15nm 4–6. This roughness is on the order of 5-
10% of the total thickness of SA3_150nm, compared to <1% for SA3_700nm. It is believed that the property
values extracted from SA3_700nm are more characteristic of the true intrinsic properties of PyC.
To further explore the relationship of fiber roughness to fracture path tortuosity, SEM fractography and
atomic force microscopy measurements were performed on the fracture surface of SA3 and HNLS. Figure
35(left) shows the SEM comparison of HNLS vs SA3 fracture surface, where SA3 is clearly more tortuous.
Figure 35(right) shows an AFM measurement from the SA3 fracture surface with RMS roughness tabulated
at ~8nm. This value is very close to literature values for fiber surface roughness6, supporting the claim that
fracture tortuosity is influenced by fiber roughness, and has an impact of the observed shear properties.
Figure 35 - (Left) SEM fractography evaluation of fracture HNLS and SA3 micro-pillars. SEM images of
fiber surfaces are reproduced from Sauder et al6. (Right) AFM scan of SA3_700nm fracture surface, and
corresponding tabulated data.
Following investigation of debond shear properties, it was attempted to evaluate the fracture energy release
rate (Γ) of the micro-pillar compression test. The goal of this effort was to expand the applicability of micro-
pillars and help inform interface properties that relate to initial crack deflection as it approaches the interface.
This effort is described in detail below.
Because failure is instantaneous, it is assumed that all of the applied elastic strain energy is released via the
interface. With regard to micro-crack deflection property evaluation, the mode II fracture energy release rate
of the PyC interface can be directly extracted for each micro-pillar test. This is achieved though application
of the work of fracture criterion that states for a brittle material, the area under the load versus displacement
curve is equivalent to the work energy that was required to overcome the formation of the new fracture
surfaces 31. Applying Cedric’s relationship32 between mode II and mode I fracture energy, ΓI can be extracted
and related to the governing parameter for interface crack deflection, Γinterface/ Γmatrix <1/4 33
.
Figure 36 - Load vs displacement curve of pillar compression showing failure load and shaded area under
the curve (work of fracture) for extraction of the fracture release rate energy. HNLS_B micro-pillar with
thick PyC interphase prior to compression.
The mode II fracture energy release rate was directly extracted from several load-displacement curves across
a variet of HNLS interface conditions. A point of concern that may influence property extraction is a possible
mechanical interaction of the fiber and matrix during compression. This is likely to be a function of interface
incline as well as interface thickness. For example, a shallow inclined interface is expected to experience
more normal compression before critical shear stress is achieved. If the PyC is compressing, then a thin
interface may lead to increased fiber to matrix interaction prior to failure. Moving forward with such
considerations in mind, and assuming pure mode II fracture with echelon propagation, we can look at energy
release rate values for the deposited PyC as a function of thickness and incline. It is expected that a thin
interface would introduce increased mechanical interaction and potentially increase the work of fracture.
Figure 37 is a graphical representation of the mode I energy release rate, ΓPyC, across the range of interface
incline (25° to 65°) with a wide range of thicknesses from 30 to 1500nm (denoted by the data labels).
Figure 37 - Energy release rate as a function of interface angle grouped in three and interface layer
thickness (data call-outs in nanometers). The mode I energy release rate was calculated using Eq.3
Cedric’s relationship ΓII = 3.5 *ΓI
From this analysis, two primary observations can be made. The first is that interface thickness (represented
by the data labels of Figure 37, in nanometers) appears to have very little influence in the general trend of
the curve. This is interesting considering a clear dependence on thickness observed for shear strength and
internal friction coefficient of the interface. It is speculated that the fracture energy release rate is less
sensitive to an increased surface area from fracture tortuosity compared to its influence on internal friction
resistance. The second observation is that as the interface incline decreases, the release rate energy increases
as well as the data scatter. It is believed that this is a result of increased mechanical influence of the fiber
surface roughness and PyC compression that evolves from increased normal stress required to achieve shear
failure. Conversely, as the interface incline increases and approaches a stress state of a pure shear (90°), the
curve flattens and appears to approach a saturation value at ~2.5 J/m2, exemplified by the blue circle in Figure
37. In an effort to characterize the claim that the observed scatter and increased release rate energy was
mechanical in nature, a low angle (25°) thin interface (38nm) pillar test that exhibited initial debond as well
as frictional debonding was characterized. In the case of this sample, it was possible to reset the indenter tip
after initial debonding and re-apply the compression to the remaining micro-pillar cap. The associated load-
displacement curve and SEM image of the sheared pillar is shown in Figure 38. As a result, it was possible
to evaluate the difference in fracture energy release rate of the fully bonded interface (area under the curve
of the first load drop (blue)) to the energy release rate of the debonded interface (area under the curve of the
second load drop (red) while adjusting for the reduced interface contact area). Subtracting out the work of
the mechanical debond, denoted Γmech, from the original pristine work of fracture, Γchem+mech allows for
consideration of the un-diluted chemical bonding value for the fracture release rate energy of PyC.
Performing this energy adjustment found ΓPyC = 3.6 J/m2 which is close to that of ΓPyC at steep inclines (>60°)
= ~2.5 J/m2 as observed in Figure 37.
Figure 38- - Isolation of chemical bonding contribution to fracture energy release rate. Post interface
failure of micro-pillar from HNLS_A with pillar cap still intact.
This evaluation may explain the scatter observed at shallow inclines, where local surface roughness and
therefore increased mechanical resistance to debond may play a stronger role in property characteristics. As
a result, it is believed that steep incline interface property values are more characteristic of true PyC fracture
energy release rate. Therefore, our analysis presents 2.5 J/m2 as an additional data point for PyC mode I
fracture energy release rate. This value translates to a Kc<1 MPa-m1/2, which is in direct support of values
found via micro-cantilever bending about. Ultimately this supports the validity of this analysis method.
It can be noted that ΓPyC varies significantly from different experimental methodologies34, however the value
presented here, 2.5 J/m2, satisfies both the crack deflection criterion (compared with ΓSiC ~20 J/m2) as well
as experimental fracture observation of bulk SiC composites, where crack deflection (and subsequent fiber
pullout, is consistently observed.
The micro-pillar work enabled by this NEUP project demonstrates of the capabilities and insights that can be
provided through SSMT. It is argued that SSMT provides an excellent platform for evaluating ceramic
composites and informing mechanistic behavior of on the constituent level, opening the door to advanced
optimization and modelling opportunity for ceramic composites. This inherently raises the quesiton of how
the properties can inform the behavior of macroscopic composites. To understand this, mini-composite
tensile testing was carried out via X-ray tomography to evaluate micro-crack evolution and failure
mechanisms. Full discussion of the relationship is discussed furhter on in the report.
Spatially resolved thermal conductivity
UIUC was the university partner responsible for characterizing the thermal properties of the composite
constituents. This is a critical component of the NEUP for the advancement of SiC/SiC composite because it
provides insight the thermal and resulting stress states that will develop under reactor operation. To date,
existing models often assume uniform material properties even though different composite constituents, such
as matrix, fiber and interphase, that have different properties. In some other models, the thermal properties
of each constituent of the composite were assumed to be the same before and after incorporating them
together. But, materials tend to have different properties as they are incorporated to different structure. To
better study the thermal properties of as-fabricated composites, we used time-domain thermoreflectance
(TDTR) to do thermal conductivity mapping of the SiC composite. By knowing the microscale thermal
properties of each constituent, the macroscale model’s accuracy of SiC composites can be improved.
As part of this NEUP, SiC/SiC composites provided by GA were constructed with Hi-Nicalon Type S fibers
and a CVI matrix. Figure 39a shows the typical fracture surfaces of these composites, with fiber matrix, and
interphase shells visible. The average diameter of the fibers is 10µm. The interphase material is composed of
5 pyrolytic carbon (PyC) layers, where the first is 40 nm and each subsequent layer is 10 nm, with 50nm SiC
layers in between each (as shown in Figure 39b). Thermal conductivity of PyC and SiC coatings have been
studied previously, and the average thermal conductivity of PyC was found to be 13.5 Wm-1K-1, and SiC was
found to be 168 Wm-1K-1.35 In this study, we sought to study the PyC interphase, matrix, and fibers as it has
been incorporated into SiC/SiC composites.
Figure 39 - Detail of structure of a SiC composite and its interphase (a) SEM picture of the Hi-Nicalon
type S fibers from a SiC composite. The interphase is clearly present. (b) Schematic of structure of the
interphase material structure, consisting of 40nm pyrolytic carbon (PyC), followed by four repeating units
of 50nm SiC/10 nm PyC
Thermal conductivity mapping of SiC composites are carried out using time-domain thermoreflectance
(TDTR). Time-domain thermoreflectance (TDTR) is a non-contact metrology tool that utilizes laser light to
measure a material’s thermal properties and the thermal conductance of interfaces. Our TDTR setup, shown
in Figure 40, uses a Ti:sapphire laser to produce 80 MHz laser pulses that are mode-locked at 783 nm. The
laser pulses are then separated by a polarizing beam splitter (PBS) and by two-tint wavelength filters into
pump and probe beams36. The pump beam heats up the material and the probe beam is used to measure the
reflectance of the material. The pump beam goes through a delay stage that varies the pump’s optical path
length. The delay stage controls the time delay between pump and probe beams, which is why TDTR is called
“time-domain” thermoreflectance.
Figure 40 - This shows the TDTR system layout. The Ti:Sapphire produces laser pulses that are split by a
PBS (Polarizing Beam Splitter) and two-tint wavelength filters into pump (red line) and probe (purple line)
pulses. The pump beam is modulated by an EOM (electro-optic modulator), and the probe beam goes
through a mechanical chopper. The pump beam heats the sample, while the probe beam measures the
reflectance of the transducer layer (e.g., Al).
Thermoreflectance is also an important characteristic for this measurement because the reflectance intensity
varies linearly with temperature for small temperature excursions (<10 K). Thus, a measurement of the
reflectance intensity as a function of time delay allows us to measure the surface temperature of the metal
transducer layer at different times. Aluminum is a commonly used transducer layer. Aluminum has good
absorbance of the laser (~13%) and a sufficiently good thermoreflectance coefficient to give enough
reflectance signal37. Finally, the TDTR data is fit using an isotropic diffusive model to calculate the thermal
conductivity of the material and the interface thermal conductance between transducer layer and the sample38.
Heat capacity used in the data analysis for both matrix and fiber was obtained by first principle calculation
for 4H SiC. For the interphase, the weighted heat capacity used was 2.04 J cm-3 K-1, based on SiC having
volumetric heat capacity of 2.21 J cm-3 K-1 and PyC having a volumetric heat capacity of 1.63 J cm-3 K-1.
TDTR requires a “mirror” finish surface to minimize light scattering. So, the sample was polished using
diamond lapping films up to a 0.1 µm finish. It was then ultrasonically cleaned using acetone and ethanol to
remove any polishing residue. After it was cleaned, the sample was annealed up to 400°C at a rate of 15°C
per minute to ensure a high interfacial thermal conductance between the sample and the transducer layer.
These steps were done for both orientations, where fibers are aligned parallel and also where fibers are aligned
perpendicular to the surface.
For our TDTR set-up, we used 20x objective lens, with a spot size of 2.7-2.9 µm, thus providing
measurements with lateral spatial resolution of 1 micrometer. The thermal conductivity mapping was done
at a time delay of 150-200 ps, where the sensitivity to the interface thermal conductance was minimized. The
area mapped was 50µm x 50µm, with a step size of 1µm. This high spatial resolution enabled us to separate
the matrix, fiber, and the interphase. The steady state heating of the material was kept at 8 K, to ensure the
accuracy of TDTR analysis. Thermal conductivity mapping was done at different regions of the sample as a
function of temperature, up to 530 K. This was done to observe the temperature dependence of each material
constituents in the SiC composites.
The thermal conductivity of matrix was found to vary across different regions. However, the fiber was found
to have uniform thermal conductivity. Figure 41a shows the TDTR data and model fits of several locations
measured in this SiC/SiC composite. The thermal conductivity of matrix was found to vary between 50 W
m-1 K-1 to 130 W m-1 K-1, and the average thermal conductivity of the fibers is 21.5 W m-1 K-1.
Figure 41 - TDTR data fits for the thermal conductivity of the SiC/SiC composite components and the ratio
of the in-phase and out-of-phase voltage as a function of thermal conductivity. (a) Data and TDTR model
fits for all mapped areas of the SiC/SiC composite, using a spot size of ωo = 2.9 µm, (b) Calculated ratio
values for t = 150 ps, a spot size of ωo = 2.9 µm, a heat capacity of C = 2.21 J cm-3 K-1, and varying
interface thermal conductance
For the thermal conductivity mapping, we fixed the time delay stage so that we can measure the ratio of the
in-phase and out-of-phase voltage as a function of position. Because of the fixed time delay, we can no longer
use the full-time delay model (dashed line in Figure 41a) to fit the thermal conductivity. Therefore, we have
to measure a full time-delay measurement at the start of each mapping measurement to get the interface
thermal conductance (G). Then, we used the G value to model the relationship between ratio and thermal
conductivity at the specific time delay that we chose. This ratio-thermal conductivity relationship for various
interface thermal conductances are shown in Figure 41b.
Since the laser spot size is around 2.7-2.9 µm, we could not use the normal orientation (fibers parallel or
perpendicular to the surface) to measure the thermal conductivity of interphase. We would need an area that
is at least 4 µm, and in the normal orientation, the width of interphase on the surface would only be around
280 nm. In order to find an area this large, we need to find a region where fibers have been pulled out and
the sample has been angled at 5 degrees or smaller (shown in diagram of Figure 42a). According to
micrographs (Figure 42b) taken of the sample, many of the fibers had a slope of 5 degrees or smaller. The
interphase thermal conductivity measured was determined to be 5.9 ± 1.0 W m-1 K-1. This thermal
conductivity is smaller than the PyC thermal conductivity measured in reference35. The difference in thermal
conductivity shows that different structure of composites can result in different thermal properties.
Figure 42 - Diagram and a thermal conductivity map of an interphase of the fiber. (a) A diagram showing
a cross section of the SiC/SiC composite where a fiber has been puleed out of the matrix. According to the
geometry of these regions, the average angle of the fibers to the surface has been determined to be 5
degrees, which creates an interphase area of ~4 µm. (b) A micrograph of a pulled fiber. (c) A thermal
conductivity map of the region at the end of a fiber that has been pulled out of the matrix. Circled in the
area where full TDTR measurements were taken. Spot size (ωo) is 2.9 µm, time delay = 150 ps, and heat
capacity (C) is 2.21 J cm-3 K-1
Thermal conductivity mapping measurements were also done at different regions of the sample. Figure 43a
shows one of the 50µm x 50µm thermal conductivity map done in region where the fibers are perpendicular
to the sample’s surface. This shows that thermal conductivity of matrix vary across different regions, which
can be seen in the gradient of the matrix region in Figure 43a. While the thermal conductivity of fibers are
more uniform across different regions, which is shown in the plateau of the fiber thermal conductivity at
position ~35 µm – 40 µm in Figure 42b.
The thermal conductivity mapping was then measured at different temperatures to understand better the
mechanism of heat transport of the matrix and fiber. The mapping was done at room temperature (25°C),
90°C, 164°C, and 250°C. The changes in thermal conductivity of different regions at different temperatures
are plotted in Figure 43c and d. Figure 43c shows that even though the shape of the thermal conductivity
trend across the region is still the same, the thermal conductivity decreases as temperature increases. In Figure
43d, we compiled the thermal conductivity of different regions as a function of temperature. From this graph,
we observed that the thermal conductivity of the matrix is roughly inversely dependent on temperature, while
the fiber’s thermal conductivity is less dependent on temperature.
Figure 43 - Summary of thermal conductivity measurements of fibers and matrix as a function of
temperature. (a) A 50µm x 50 µm thermal conductivity map, taken at 1 µm steps, at room temperature
(lighter region indicates higher thermal conductivity), (b) A cross section through the map in figure5a
(dashed white line) at room temperature, (c) A cross section of the same region at different temperatures,
(d) Summary of thermal conductivity dependencies on temperatures for matrix and fiber, compared with
In summary, this NEUP supported the development and application of TDTR and we were able to map the
thermal conductivity of matrix, fiber, and interphase in SiC/SiC composites as a function of temperature
using. From the measurement results, we observed that the interphase thermal conductivity is lower than the
literature value. The matrix thermal conductivity varies across different regions, from 50 Wm-1K-1 to 130
Wm-1K-1 at room temperature, and matrix’s thermal conductivity is roughly inversely dependent on
temperature. The fiber thermal conductivity is uniform across different regions, averaging at about 21.5 Wm-
1K-1, and fiber’s thermal conductivity is not strongly dependent on temperature.
Irradiation Effects
Another key component of this NEUP was the evaluation of the neutron and ion irradiation damage effects
in SiC/SiC composites. As outlined in the project proposal, the mission of this project to integrate constituent
level properties into a mechanistically informed component scale model. A fully developed and
mechanistically informed model that has been validated by unirradiated bulk composite testing will
ultimately only require constituent level properties. This is the intended advantage of applying SSMT and
constituent level characterization, allowing rapid and safe analysis of irradiated materials without dealing
with very large and highly radioactive sample sets. The goal of this effort was to apply small scale mechanical
testing techniques to probe the evolving properties, and implement these finding into the working model.
2.4.4.1 Irradiated samples
Several characterization techniques were applied to both neutron and ion irradiated samples. Unfortunately,
shipping and inventory maintenance of radioactive material proved challenging for partner universities other
than UCB who hosts the Nuclear Materials user facility. As a result, substitute ion irradiation campaigns
were carried out by Oxford collaborators at the University of Surrey Ion Beam Center. Additionally, He ion
irradiations were carried out at Berkeley and sent to UIUC for thermal property evaluation using TDTR.
Details regarding dose and temperature are presented in the following results sections.
This NEUP provided opportunity for UCB to expand the capabilities as the NSUF Nuclear Materials User
Facility. This collaboration enabled the design and fabrication of a new sample storage unit, which can hold
up to 300 samples (depending on the dose rate) in 50 individual drawers. The lead shielding and locking
mechanism provide highest level of security and safety for long term storage of radioactive samples. Figure
44 shows the complete sample storage cabinet. UC Berkeley received and easily stored 75 SiC/SiC composite
samples from GA that were irradiated in a High Flux Isotope Reactor (HFIR) campaign prior to this NEUP.
The neutron irradiated samples were fluxed at ORNL HFIR to 2.2 and 4.5 dpa across temperatures 623-
730°C. Table 4 tabulates the list of samples delivered to UCB. CMC bar samples from pig #2 were cut and
prepared for nano-indentation and micro-pillar compression at UC Berkeley. Details regarding the
experimental techniques and results are presented in following sections.
Figure 44 - Radioactive materials storage cabinet. Source drawers are complete with lead shielding and
locking mechanisms
Table 4 - List of irradiated SiC/SiC samples received from GA
PIG 1 PIG 2 Pig 3 Pig 4
Information
Reference GSD1 GSF1 & GSF2 GST1 & GST2 GSF1 & GSF2
SpecimenType TD Discs CMC bars Torsion Joint Specimens mounted on epoxy
Material SiC-SiC SiC-SiC SiC-SiC SiC
Total specimens 16 24 24 11
Specimens
Weight (g) 1.28 0.70 0.5 21.22
Contact Dose (mrem/hr) 22.6 132 49.6 82.9
30cm Dose (mrem/hr) 0.6 4.4 3.9 1.2
Pig
Contact Dose (mrem/hr) 8 44 18.5 11.2
30cm Dose (mrem/hr) 0.6 2.4 1.1 0.6
Isotopic Analysis
Isotope Co-60 Mn-54 Mn-54 Mn-54
Activity (mCi) 3.49E+01 5.165E-01 1.02E+00 6.08E-02
Co-60 Co-60 Co-60
2.130E+02 1.067E+02 2.931E+01
Zn-65 Zn-65
1.669E+00 1.49E-01
Ta-182 Ta-182
4.590E+00 1.54E-01
Specimen ID's
GA-TD-1 GA-FB-01 GA-ST-01 GA-FB-03
GA-TD-2 GA-FB-02 GA-ST-02 GA-FB-04
GA-TD-3 GA-FB-03 GA-ST-03 GA-FB-08
GA-TD-4 GA-FB-04 GA-ST-04 GA-FB-09
GA-TD-5 GA-FB-05 GA-ST-05 GA-FB-10
GA-TD-6 GA-FB-06 GA-ST-06 GA-FB-11
GA-TD-7 GA-FB-07 GA-ST-07 GA-FB-15
GA-TD-8 GA-FB-08 GA-ST-08 GA-FB-16
GA-TD-9 GA-FB-09 GA-ST-09 GA-FB-19
GA-TD-10 GA-FB-10 GA-ST-10 GA-FB-21
GA-TD-11 GA-FB-11 GA-ST-11 GA-FB-23
GA-TD-12 GA-FB-12 GA-ST-12
GA-TD-13 GA-FB-13 GA-ST-13
GA-TD-14 GA-FB-14 GA-ST-14
GA-TD-15 GA-FB-15 GA-ST-15
GA-TD-16 GA-FB-16 GA-ST-16
GA-FB-17 GA-ST-17
GA-FB-18 GA-ST-18
GA-FB-19 GA-ST-19
GA-FB-20 GA-ST-20
GA-FB-21 GA-ST-21
GA-FB-22 GA-ST-22
GA-FB-23 GA-ST-23
GA-FB-24 GA-ST-24
Additional neutron irradiated samples were received from ORNL under a RTE NSUF proposal. Again, many
of these samples underwent similar characterization developed in this NEUP, and the results are interesting
and stimulating to discuss in this context. These data have been included to support the data of this project
as to provide a more comprehensive view of the impact of irradiation damage. NEUP GA irradiated samples
are all SA3 fiber composites w/ ~100nm PyC first layer thickness. NSUF samples are HNLS composites with
~50-100nm first layer PyC25. The applied experimental techniques include micro-cantilever testing for
fracture toughness, nano-indentation for elastic modulus and hardness, micro-pillar compression for interface
strength and friction properties, and TEM to understand degradation/evolution of the constituent
microstructure.
2.4.4.2 Neutron irradiation:
UCB Berkeley carried out nano-indentation and micro-pillar compression on irradiated composites to
evaluate the hardness, elastic modulus, and interfacial debond properties as a function of dose and irradiation
temperature. Details regarding experimental analysis were already described in previous sections, therefore
this section quickly present the results and summarizes the impact of neutron irradiation effects.
Nano-indentation:
Diamond tip Berkovich indentations were carried out using a MicroMaterials Nanoindenter on control and
neutron irradiated SA3 composites across a temperature range RT to 500°C. Below in Figure 45 show the
change in hardness and modulus for the fibers and matrix as a function of temperature. As described in section
2.4.2.1, indentation data for thin <1μm PyC interfaces has always proved challenging to gather statistically
satisfying properties. As a result, the nano-indentation campaign presents data for the fiber and matrix. The
neutron irradiated sample received 4.5dpa @630°C, and is from the exact samples used for micro-pillar
compression of the PyC interface. This data is also tabulated and compared to ion irradiation in Table 5.
Figure 45 – (Top) Plot of the hardness and (bottom) Elastic Modulus as a function of constituent,
irradiation, and temperature.
These plots show that hardness and modulus decrease for both fiber and matrix as a function of temperature.
Also, it shows fundamentally that hardness is increased and modulus is slightly decreased for the irradiated
samples. This would suggest the neutron irradiation point defect damage increases resistance to deformation
while the local disorder decreases the elastic modulus. This is consistent with literature and expected as point
defects reduce ability for plastic deformation. Though, the decrease in the modulus over temperature is larger
than expected for this temperature range, dropping nearly 100 and 50 GPa for the matrix and fiber
respectively over 500°C. It can be seen that the fibers still have a large variation in their mechanical properties
even over temperature and irradiation. This is likely due to the large concentration and variability of carbon
precipitates formed during the manufacturing process.
Looking at the RT indentations from Oxford and UC Berkeley for the unirradiated matrix, we find very
consistent hardness equal to ~37GPa, but a large difference in modulus from 460GPa to 540GPa respectively.
460GPa is consistent with literature, while 540 is likely an artifact of the data processing or system calibration
on that run. This is currently undergoing investigation, and we advise the application of 460GPa as the model
input for modulus, but still follow the slope of the hardness vs temperature curve established by Berkeley.
Micro-pillar compression:
Compression testing was carried out in-situ SEM using a Hysitron PI88 Picoindenter to evaluate the influence
of irradiation of interface debond strength and friction characteristics. Samples that were tested include ~50-
140nm PyC interface of HNLS and SA3 fiber composites respectively. The HNLS sample was exposed to
1dpa at ~350°, while the SA3 sample experienced 4.5dpa at ~630°C. An additional 12dpa HNLS fiber sample
with 180nm PyC was tested. Figure 46 compares the control and irradiated samples by applying the Mohr-
Coulomb criterion where the intercept is the cohesive shear strength and the slope the internal friction
coefficient. This linear regression, and thereby property values are displayed on Figure 46 as well as within
Table 5 in the property values section of this document. Graphics of representative pillars of each condition
are overlaid in Figure 46. The analysis revealed a significant decrease in cohesive shear strength for irradiated
samples, however without much impact on friction coefficient.
Figure 46 - Mohr-Coulomb criterion applied to the unirradiated and irradiated interfaces. A fundamental
decrease in cohesive shear strength between unirradiated and irradiated interfaces was observed.
The group averaged pillars at ~60° for each condition reveal a fundamental decrease in shear strength of
~200MPa and ~75MPa for SA3_140nm and HNLS_50nm respectively. This suggests that degradation in
shear strength scales with increase in neutron damage. With respect to irradiation effects, it can be observed
that while the chemical debonding strength decreases with irradiation, while the friction coefficient appears
remain the same. As a whole, the interface was weakened, however the relative contribution of chemical
bonding and frictional resistance has shifted. This is likely a result of irradiation induced microstructural
y =0.24x + 160
y =0.25x + 266
y =0.27x + 280
changes within the turbostatic graphite-like structures 39,40 . For thought, consider a perfect graphite lattice
for which shear failure occurs at the weakly bound basal planes. As irradiation displaces atoms from the
hexagonal frame work, it is conceived that there is an effective reduction in interacting basal plane surface
area, resulting in reduced strength. In addition, the displacements may result in new stacking and basal plane
arrangement, some of which may have less strength than the others.
The HNLS_180nm 12dpa showed very poor consistency. The group averaged interface inclines for
HNLS_180 12dpa, shown in Figure 46, reveal that the data is not applicable for the MC criterion. This is
likely a result of the relatively large porosity that evolved at this irradiation dose and temperature. Ultimately,
this porosity is expected to be inducing stress concentrations during testing, thereby introducing significant
data scatter and straying from a representative debond strength. To better understand the evolved properties
and characteristics, Figure 47 compares side-by side close up SEM images of the control and irradiated
samples. For HNLS_50nm 1dpa, major defects are not observed. HNLS_180nm 12pda shows significant
defect evolution, and SA3_140nm 4.5dpa does not show show major defects. There has been extensive
research into defect evolution and crystallite reconstruction for graphite and graphite-like materials under
irradiation 40–42. It has been found that swelling/shrinkage, dissolution and restructuring of nano-crystallites
is strongly dependent on the fabrication parameters, as well as the irradiation dose and temperature. The
porosity defects observed for HNLS_180nm 12dpa are expected to be a result of relatively low irradiation
temperature of 280°C. It has been discussed that the activation energy for graphite-like materials to undergo
significant point defect recombination during irradiation is around 300°C. Therefore, the porosity is likely an
artifact of reduced interstitial defect mobility, allowing for biased vacancy cluster formation. However, most
literature data is associated with bulk nuclear grade graphite which can exhibit substantially different
microstructure compared to the semi-oriented nano-crystallite nature of thin deposited PyC. Therefore,
drawing major conclusions and characteristics of historical nuclear graphite may be misleading.’
Figure 47 –Before and after irradiation SEM images comparing the PyC interface structure. (Top)
HNLS_50nm control vs 1dpa at 350°C. (Middle) HNLS_180nm control vs ~12dpa at 280°. (Bottom)
SA3_140nm control vs ~4.5dpa at 630°.
2.4.4.3 Ion Irradiation
The ion irradiation campaign was performed at University of Surrey Ion Beam Center. It included Si-ion
irradiation of SA3 SiC-SiC samples from GA of two different grades – one with single-layered, and one with
multi-layered interphase. In order to create a quasi-uniform damage profile, consecutive irradiations at three
different energies were accomplished – 500 keV, 1 MeV and 2 MeV. These conditions Irradiations were
performed with two peak damage levels – 0.26 and 2.6 dpa. Using SRIM software, the ion doses necessary
to achieve a specified damage level were calculated, and set as follows:
For peak damage of 0.26 dpa:
2 MeV – 6e14 ions/cm2;
1MeV – 4e14 ions/cm2;
0.5 MeV – 2e14 ions/cm2.
For peak damage of 2.6 dpa:
2 MeV – 6e15 ions/cm2;
1MeV – 4e15 ions/cm2;
0.5 MeV – 2e15 ions/cm2
Sample temperature during irradiation was set to 300°C for 0.26 dpa irradiation, and to 750°C for 2.6 dpa
irradiations. Irradiated samples were characterized using nanoindentation and cantilever fracture testing.
Nanoindentation was done using sharp Berkovich indenter and continuous stiffness measurement mode
(CSM). Two measurement modes were implemented:
Shallow indents:
In this measurement mode, shallow indents of 400 nm deep were made at 1.5 µm spacing. Lines of indents
were then going from matrix into a fiber, oriented normal to the surface. This way, hardness and modulus
could be measured both in the matrix and fiber materials; several indents at different radial position within a
fiber allowed monitoring the changes of already radially non-uniform properties. A disadvantage of this mode
is that only part of the irradiated depth can be probed. Figure 48 below presents an example of such a linescan,
for a sample of a single-layer interphase grade at 2.6dpa. In order to produce a single point corresponding to
a specific location, values of hardness or modulus from a corresponding depth dependence, as measured by
CSM method, were averaged in the depth range of 300 – 380 nm.
Figure 48 - Comparison of typical linescans of hardness and modulus across matrix and fiber for 2.6dpa
ion irradiated and reference unirradiated samples.
Deep indents:
Here deeper indents of 1000 nm deep were made, forming rectangular arrays in the matrix and irregular
arrays to probe the fibers, at 15 – 20 µm spacing. This way, most of the irradiated depth is probed, but indents
are necessarily located farther from each other, not allowing the probing of several locations within a fiber.
An example of such a measurement of depth dependence of hardness and modulus is presented in Figure 49
below:
-20 -10 0 1010
20
30
40
50
Hard
ness, G
Pa
Position relative to fiber center, m
300 nm deep indents
Unirradiated
Irraidated
-20 -10 0 10
200
300
400
500
Modulu
s, G
Pa
Position relative to fiber center, m
300 nm deep indents
Unirradiated
Irradiated
Figure 49 - Typical comparison of depth dependence of hardness and modulus in the matrix for samples
irradiated to different damage levels.
Nanoindentation measurements indicate that overall effect of irradiation on hardness and elastic modulus is
small, both in matrix and in the fibers. Hardness slightly increases, and modulus slightly decreases, similar
to that observed for neutron irradiation. Further increase of dose doesn’t make a significant effect on hardness
and modulus.
On the other hand, microscopic examination of the indents reveals that irradiation leads to the noticeable
modification of the crack patterns surrounding the indents, shown in Figure 50 below. Prominent radial cracks
appear at the corners of the indents in unirradiated matrix material. However, following the irradiation
cracking is completely suppressed.
Figure 50 - Comparison of crack patterns around the indents in the matrix of unirradiated an irradiated
samples.
This suppression of cracking can be attributed to the stress field induced in near-surface region sue to ion
irradiation. This stress would counteract the crack propagation, suppressing the cracking.
Micro-Cantilever testing:
Following the same procedures presented in section 2.4.2.2, testing was performed using straight-notched
triangular cantilevers, manufactured using FIB at the interphases, in the matrix and in the fibers. Figure 51
below presents the comparison of fracture toughness between irradiated and unirradiated samples, separately
for each composite constituent.
0 200 400 600 800 10000
10
20
30
40
Ha
rdn
ess, G
Pa
Depth, nm
1000 nm deep indents in the matrix
0.26 dpa
2.6 dpa
0 200 400 600 800 10000
100
200
300
400
500
Modulu
s, G
Pa
Depth, nm
1000 nm deep indents in the matrix
0.26 dpa
2.6 dpa
2 µm
Unirradiated
2 µm
Irradiated – 2.6 dpa
Figure 51 - Comparison of fracture toughness as measured in different constituents, for unirradiated and
irradiated samples.
It can be seen that there is a clear trend towards an increase of toughness for the interphases and fibers.
Toughness of the interphases increases after irradiation, both for single- and multi-layered ones. There is a
noticeable difference (about factor of 2) between the interphase of the irradiated and unirradiated samples;
on the other hand, toughness of the irradiated interphases is similar, regardless of the irradiation conditions.
Toughness of the fibers progressively increases with the increase of irradiation dose and temperature. On the
other hand, there doesn’t seem to be a definitive trend in the properties of matrices. Although different trends
arise with respect to constituent material, it is still the case that fracture toughness of the interface is ~<1/4
the toughness of the matrix. This is important for matrix crack deflection, and maintaining pseudo-ductility.
This supports observation of graceful failure in irradiated composites25.
To compare the effects of irradiation type, we can compare the hardness and modulus of the 760°C 2.6dpa
Si ion irradiated to the 630°C 4.5dpa neutron irradiated sample. The hardness for the matrix was 42GPa and
39GPa respectively. The modulus for the matrix is 425 GPa and 500GPa respectively. This reveals a relative
increase of ~14% and 5% for ion irradiated and neutron irradiated hardness respectively. For the modulus,
we observed a relative reduction of 5% and 7% for ion irradiated and neutron irradiated respectively. It is
unexpected that matrix properties change much beyond 1dpa as SiC reaches point defect swelling plateau.
Although relatively similar, one explanation for ion irradiation and neutron irradiation discrepancy is that
irradiation was not at the same temperature relevant dose rate, so point defect evolution and recombination
may have affected the damage state. With that in mind, the data may suggest that hardness is more strongly
dependent on the type irradiation, and that modulus is much less sensitive. The mechanical properties for
control, ion, and neutron irradiated samples are tabulated and compared below in Table 5.
Table 5 - Comparison of irradiated constituent properties tested at ambient temperature. Constituent H (GPa) E(GPa) KI
(MPa-m1/2)
τo
(MPa)
μi
Unirr.
SA3
M 37, 37* 460, 540* 4.1 - -
F Gradient with
center = 14,
25* (averaged)
Gradient with
center 190,
360* (averaged)
2.05 - -
I (140nm PyC) - - 0.8 280* 0.27*
Neutron:
4.5 dpa, 716°C
SA3
M 39* 500* - ~100* ~0.27*^
F 27* 330* - - -
I (140nm PyC) - - - - -
Unirr (NSUF):
HNLS
M - - - - -
F - - - - -
I (50nm PyC) - - - 266* 0.25*
Neutron
(NSUF):
1 dpa, 330°C
HNLS
M - - - - -
F - - - - -
I (50nm PyC) - - - 160* 0.24*
Si ion
irradiation:
0.24dpa 300°C
SA3
M 40 445 5.2 - -
F - - 2.5 - -
I (140nm PyC) - - 1.5 - -
Si ion
Irradiation:
2.6dpa 760°C
SA3
M 42 425 3.8 - -
F Gradient with
center 15
Gradient center
200
2.75 - -
I (140nm PyC) - - 1.3 - -
*UCB data ^ Estimated assuming limited change in μi observed for HNLS
TDTR on He-ion irradiated single crystal SiC:
A 6H crystal structure SiC wafer was implanted with He ions at 300°C at fluences of 0.016, 0.08, and 0.16
nC-μm-2. Following implantation, the samples were cleaned by spraying them with IPA, and were not heated
in order to not disturb the irradiated damage. Following procedures for TDTR outlined in section 2.4.3, the
He implanted areas were TDTR mapped with 1 μm resolution. In trial TDTR measurements, the aluminum
interfacial thermal conductance was calculated to be 80 MW m-2 K-1, which was slightly lower than
anticipated. However, the picosecond acoustic echo of the Al transducer had a sharp peak, indicating that the
interface was clean. It is suspected that the lower interfacial thermal conductance is only as a result of the
swelling from the He implantation, and that no foreign debris was present between the transducer and sample.
Though the interfacial thermal conductance is low compared to its values for un-implanted SiC crystals,
because the implanted area thermal conductivity is so low, the measurement is much more sensitive to the
implanted SiC area than the interface’s thermal conductance. Averaging the mapping data together, it was
found that the thermal conductivity of the sample at 0.016 nC-μm-2 was 13 W m-1 K-1 , at 0.08 nC-μm-2 it
was 2.8 W m-1 K-1 , and at 0.16 nC-μm-2 it was 2.2 W m-1 K-1 . Figure 52 shows the TLDR map of the 0.16 nC-μm-2 implantation and corresponding trend in thermal conductivity as a function of dose. This map also shows that the unirradiated thermal conductivity was approximately 100 W m-1 K-1, which is comparable to the unirradiated SiC composite matrix material (which is 3C cubic crystal structure and polycrystalline) at room temperature of 80-120 W m-1 K-1 presented earlier. Figure 52d shows the thermal conductivity trend as a function of He-ion implantation. It is clear that increasing dose rapidly reduces this conductivity, which is anticipated because of the sensitivity of thermal conductivity and phonon energy dispersion to point defect concentration. Although He implantation is not expected in fission reactor environments, this data supports finding for rapid deterioration of thermal conductivity found for neutron irradiation of monolithic polycrystalline SiC presented by Snead et al23, and may prove useful for fusion applications down the road.
Figure 52 - Thermal conductivity mapping and profile of a 20x20 μm He implanted region of a 6H SiC
wafer. The fluence of the implanted region is 0.016 nC-μm-2. (a) Photo of the He implanted region of the
SiC wafer. (b) A 20x20 μm thermal conductivity map of the He implanted region, taken at 1 μm steps. Time
delay was set to 𝑡𝑡 = 150 ps, heat capacity was C = 2.21 J cm-3 K-1, and the interfacial Al thermal
conductance for this region was 80 MW m-2 K-1. (c) Thermal conductivity profile of the implanted region.
(d) Plotted trend of thermal conductivity across fluences of 0.016 nC-μm-2, 0.08 nC-μm-2, and 0.16 nC-μm-2.
Macroscopic composite characterization In order to develop a suitable model that is able to predict composite behavior based on constituent level
properties, the macroscopic characteristics and mechanical properties must be evaluated for validation. The
first step of this effort was to capture the inherent porosity of the bulk composite that initiate micro-cracks in
the matrix. The following steps were to characterize the mechanical response and deformation of these
composites. The sections below describe and summarize these efforts in detail.
2.4.5.1 X-ray Tomography Composite materials fabricated via chemical vapor infiltration (CVI) method are prone to have some internal
porosity due to the nature of the process. The amount of internal porosity is a function of CVI process
conditions, component geometry and fiber architecture. The amount and the geometrical shape of internal
voids have a direct effect on mechanical properties of the bulk material and therefore should be considered
when evaluating mechanical performance of a composite. In this work, x-ray computed tomography (XCT)
was performed on a planar and tubular specimens with the goal to measure internal porosity. Samples
evaluated were SA3 and HNLS fiber composites produced at GA under the funding of this NEUP
All specimens were scanned at 100 kV and 100 μA x-ray beam power. Each specimen was reconstructed into
a 3D volume from 720 individual projection scans. The specimens were positioned as close to the beam
source as geometrically allowable in order to obtained the highest resolution of the scans. Volume Graphics
VGStudio Max 2.2 software was used to analyze scanned specimens and to determine internal and external
surfaces. The enclosed internal porosity was found by selecting a three dimensional boundary that was within
the sample surface and that also contained all of the internal porosity (no open porosity from the surface was
included and no internal porosity is excluded from the bounds). Once the boundary was established a simple
porosity search was conducted that finds voxels (three dimensional resolution units) within the preset bounds
that were then identified as voids. The total void voxel count, total void voxel volume, volume of contiguous
voids and their locations were all calculated within the Volume Graphics software. The resulting porosity is
represented as a void percent fraction of the total specimen volume. It should be noted that the porosity
measurement using XCT technique is ultimately limited by the resolution of a scan.
The XCT scan results for planar and tubular specimens are shown in Table 6. A three dimensional
reconstruction with the porosity maps is shown in Figure 53. It should be noted that beam hardening feature
was used during volume reconstruction of the planar composite because of higher beam scattering near flat
edges of planar specimens. Beam hardening corrections were not used for the tubular specimen. The resulting
porosity of planar specimens was measured to be less than 0.5% of total volume. The tubular specimen
showed a network of internal voids located within the composite braid, closely matching the fiber
architecture. Measured porosity for this specimen was 6.5% of the total volume.
D
Table 6 – Planar and Tubular specimen XCT scan results
Figure 53 – (left) XCT scan of a planar specimen with identified porosity map. (Right) XCT image showing
internal porosity of tubular specimen.
XCT mini-composite testing:
In situ tensile tests were performed on mini-composites to evaluate micro-crack evolution. We conducted a
24-hour beam time experiment at the Advanced Light Source (Lawrence-Berkeley Lab, University of
California Berkeley) to aid in the development and advancement of experimental methods for in situ tensile
testing at elevated temperature with simultaneous x-ray tomography imaging. A detailed description of the
beam-line and test set up was published by Bale et al43. Our work sought out to run these tests at room
temperature, 700°C and 1000°C to evaluate temperature dependence on micro-crack evolution and
deformation.
Testing was carried out on Tyranno ZMI fiber minicomposites with ~70nm PyC monolayer interfaces
fabricated by GA. The tensile test set up was comprised of ball in socket joints to allow for mini composite
self-alignment. The load train was surrounded by an inert gas chamber and directional halogen heat lamps
capable of bringing the system to 1600°C. Figure 54 is a schematic of the test chamber44. The mini-composite
samples were then mounted using OMEGABOND “700” high temperature cement in a bored and externally
threaded insert. Each insert was subsequently threaded into the high temperature brass balls for socket
mounting in the tensile fixture. Specimen temperature was regulated using a mounted thermocouple and
feedback system to adjust the current to the heat lamps. Examples of the mounted specimens are show at the
top of Figure 55. During testing, X-rays enter through a 0.5cm alumina window as the entire chamber rotates
180 degrees to capture the projection scans need for reconstruction. Because of the limited viewing window,
a notch was introduced at the center of the gauge length to promote microcrack evolution in the scan
projections. It has been discussed that the stress concentration around a notch in ceramic mini-composites
can dissipate quickly once localize micro cracking take place33. This means that micro-crack spacing away
from the notch can be representative of an unnotched specimen. Micro-crack spacing is directly related to
the interface shear properties that determine how large of a debond length will evolve along the fiber axis
once the matrix crack has deflected34. Typically, a weaker interface results in a larger debond length and
thereby larger micro-crack spacing.
Planar composite Tubular composite
Resolution, μm 5.4 5.6
Average Width, mm 7.0 8.20
Average Thickness, mm 1.0 9.68
Beam Hardening Minimal None
Porosity, % <0.5% 6.5
Figure 54 – (Left)Schematic illustrations showing a specimen mounted between upper and lower grips.
(Right) External image of chamber with halogen lamps at 700°C
Testing was successfully carried out for three ambient and three 700°C mini-composite specimens.
Unfortunately, the high temperature cement consistently failed at 1000°C. Testing was carried out by
applying incremental loads via displacement-controlled steps of 5μm. A typical load vs time curve is shown
in Figure 55 for sample T5 that was tested at 700°C. The peak load for this test was 24lbs.
Figure 55 - (Top) Sequential stage of test set up; high temperature cement casting, in situ high temperature
loading, and post fracture evaluation. Note that a thermo couple was attached to this specimen to
appropriately track the thermal state. (Bottom) Typical load vs time plot following 5um incremental lading
steps.
Alumina window for x-ray penetration
5μm disp.-
controlled steps
Peak load ~ 24lbs
At each incremental load, an x-ray projection scan was taken. This enabled real-time visualization of
microcrack evolution, shown in Figure 56, and the ability to evaluate 3D reconstructions. These
reconstructions are made by stacking individual z-slices that represent the material cross section. Figure 56
shows a reconstructed cross section. The resolution of the tomography system offered 1.6um/pixel. Using
imageJ, the cross-sectional area and porosity could be extracted. The cross-sectional area of 0.53*10-6m2
reveal failure stress equal to ~200MPa, which is within the typical range for SiC/SiC mini-composites.
Porosity was found to be ~5%, which is also typical of CVI production as will be discussed in the following
section. Looking at the normal projection of Figure 56, micro-crack spacing is clearly visible. The observed
average micro-crack spacing, taken by dividing the observable micro-cracks into the height of the observable
window, is on the order of 0.35mm for both ambient and 700°C. This translates to an interfacial sliding
strength of ~45MPa for these mini-composite configurations. Details on this value is discussed in detail in
the next section regarding hysteresis testing. Similar micro-crack spacing for both ambient and 700°C
suggests minimal change in interface characteristics across this temperature change. This is expected
considering the system is in an intern environment and both carbon and SiC material is not expected to
degrade mechanically until much greater temperatures.
Figure 56 – (top) Tomographic reconstruction showing the cross-sectional geometry of a single tow tensile
test sample. (bottom left) Normal x-ray projection of ambient temperature test. Numbers denote visible
micro-cracks. (Bottom right) Normal x-ray projection of 700°C test, numbers denote visible micro-cracks.
We see this behavior as key to understanding fiber-matrix interaction. By creating simplified representations
of fractures bridged by fibers, we can investigate the matrix, fiber, and interaction parameters that support
development of subsequent fractures and control fracture spacing. If this fracture behavior does not appear
with a particular interaction condition (e.g. simple friction) then more complex models will be investigated.
Individual parameters (e.g. friction coefficient) that underpin working models can then be cycled back
through the research group for reconciliation with existing data.
2.4.5.2 Hysteresis testing of unidirectional mini-composites Hystersis testing explores a myriad of composite behavior and performance characteristics from residual
stresses to damage tolerance22,33,45–48. The specific goal of this effort was to link interface properties from
micro- to macro-scale and to illuminate fundamental differences betweent the two. There are two analytical
relationships that have been developed to evaluate interfacial sliding strength based on macroscopic
performance. The first is related to the hystersis loop width of the stress-strain data, and the second is related
to the micro-crack evolution and specifcally the distance between these micro-cracks. Hystersis loop width
measurments requires high fidelity strain measurement. Unfortunately our experimental set up had size and
spacing limitations that eliminated the option for high precision strain gauges. Additionally the SEM field of
view was such that digital image correlation (DIC) methods were not applicable. However, the SEM enabled
unique insight to the micro-crack evolution and spacing. The analytical model for interfacial sliding strength
is related to matrix crack spacing by the following equation48
𝜏 = 𝜎𝑠𝑅𝑓𝑉𝑚
2𝑉𝑓𝐿𝑠
where σs is the crack saturation stress (usually taken to be the peak stress before failure for PyC interface
containing composites46), Rf, is the fiber radius, V is the volume fraction for matrix and fiber respectively,
and Ls is the average measured crack spacing across the gauge length. This equation is derived from the
debond length of the defelcted crack, and more specifically, the resitantce to sliding behind that crack tip.
This resistance to sliding, often refered to as the sliding strength or shear “stress” (as opposed to shear
“strength” that is the resistance to initial debond), governs the fiber-matrix load sharing as the composite is
put in tension. If the sliding strength is large, the matrix carries more load, and is therefore more likely to
crack, with dependence on defect distribution in the matrix of course. This is what gives rise to relatively
uniform crack spacing observed in experimental hysteresis testing. Increasing the sliding strength decreases
the observed matrix crack spacing. Figure 57 is a schematic that describes load sharing as function of debond
length for a given sliding strength. Where the sliding strength is dependent on the residual stress, friction
coefficient and initial debond strength. Ls represents the average matrix crack spacing that occurs based on
the stress and defect distributions.
Figure 57 – Schematic of load sharing between fiber and matrix as a function of debond length. Debond
strength as well as sliding strength are dependent on residual stress and friction characteristics.24
Thereby it is important to note that the value obtained via hysteresis testing is fundamentally diffrenet from
the τdebond described earlier by the micro-pillar and pushout testing. Both are important parameters that
influence the debond length and energy abosorbtion (micro-cracking) of the composite as a whole. However,
they are not the same property and therefore cannot be directly compared. With that, the values presented
below are lower than the values found for either micro-pillar comporession or fiber pushout. This is
reasonable as the stress required to slide an already debonded interface is expected to be less than the measure
the true bond strength.
UC Berkeley received two sets of unidirectional, single tow (~500 fiber bundle), mini-composites. One set
contained HNLS fibers with 50nm PyC monolayer fiber-matrix interphase. The other set contained SA3
fibers and the same interphase. The unidirectional fiber architecture was fully infiltrated via CVI SiC
processing at GA, then sent to UC Berkeley. Images of the mini-composites were shown earlier in Figure 1.
Figure 58 shows typical cross-sections for the SA3 and HNLS composites. The area, porosity, fiber and
matrix volume fraction were extracted for each mini-composite. Porosity for the SA3 composites was ~8%
while HNLS showed ~4%, and fiber volume fraction was around ~25% for both.
Figure 58 – Fracture surface cross-sections of SA3 (left) and HNLS (right) mini-composites. The total area
and porosity are outlined in yellow. ImageJ measurement results, in μm2, are overlaid. Small circular black
dots in the cross-section are locations of fiber pullout.
Testing was performed using the Kammrath & Weiss Tensile and Compression Module in-situ XL30 Phillips
SEM, Figure 59a. Testing was performed using a 500N load cell and displacement control at 1μm/s. A
baseline hysteresis test schedule called for 5N load steps, returing back to the first infleciton point of the load
curve (~30N) associated with flexing and final alignment of custom grips. The instrument software was very
versitle and allowed for test pausing and redefinition of load step size and lower bounds when necessary, see
Figure 60. This enabled SEM imaging of the entire gauge length mid test, as well as insight to hystersis
behavior with microcracks fully open. Special attention was paid to the gripping configuration and alignment
during mounting. A custom designed ball-and-socket gripping system was machined in house at UC
Berkeley, shown in Figure 59a&c. The mini-composites were cut to ~5cm, 2cm was used for epoxy gripping
on both ends, leaving a 1cm gauge length. The ends were mounted in bored and externally threaded studs
and set up in an alignment ficture to cure, Figure 59b.
Figure 59 – A) Kammrath and Weiss module on SEM stage with loaded sample. B) Alignment fixture for
epoxy mounting and curing. C) SA3 mini-composite test specimen with SS balls on threaded studs. Ball and
socket joints were lubricated with colloidal graphite
In total, four SA3 and four HNLS minicomposites were tested. Figure 60 shows typical stress-strain curves
for SA3 (left) and HNLS (right). Stress was evaluated via the cross-sections of each sample tested, shown
in Figure 58. As alluded to earlier, strain data was only collected via cross-head output readings. Thereby,
Total area μm2
porosity μm2
only general characteristics of the stress-strain curve are worth discussing. It is noted that the averge slope of
each loop reduces as loading and unloading takes place. This is a measure of damage tolerance and is both
expected and desired in CMCs. This damage tolerance is a result of micro-crack evolution. Secondly we
observe that loop width is increasing as the test progresses. This is a sign of increased micro-cracks and
thereby increased debonded interface. The sliding of the debonded interface, or energy absorption, is why
hystersis is observed. Finally we can point out that the average onset of matrix cracking, or proportional limit
stress, was around ~175MPa for HNLS and ~125MPa for SA3. The ultimate tensile strength was on average
about 300MPa for both. These values align reasonably with those in literature5, though are on the high end.
This may be because we accounted for and subtracted out porosity from our cross-section for stress
calculation. The lower PLS value for SA3 is attributed to the slightly less homogenous cross-section and
increased porosity, leading to more defects and therefore crack initiation sites.
Figure 60 – Stress-strain curves for SA3 (left) and HNLS (right). The SA3 curve show the versatility of the
KW software to pause, and redefine load/unload regime to explore effects of interface degradation at
matrix crack saturation.
During testing, loading was paused and images were taken along the entire gauge length to capture the matrix
crack spacing. Figure 61 shows the resulting SEM images of gauge length and micro-cracking.
Figure 61 – a) Stitched SEM image of the gauge length prior to failure for SA3 mini-composite #4. The
gauge length is saturated with matrix cracks. B) Zoomed in image from the same SA3 sample. Showing
typical matrix crack spacing. Overlaid value for Ls is the average value across the entire gauge length. C)
HNLS minicompoiste after failure with matrix crack spacing still visible. Dotted lines attributed to flaking
following abrupt crack closure of the thin conductive coating that was deposited.
Applying the observed micro-cack spacing to the equation above provided the interfacial sliding strength
values shown in the table below.
τsliding
for mini-composites with 50nm monolayer PyC interphase
τ SA3
= 62 MPa
τ HNLS
= 17 MPa
These values follow the same trend observed for micro-pillar compression. The rougher fiber results in
increased resistance to failure. In this case, it is likely that this roughness is contributing significantly to the
frictional resistance to sliding. It has also been shown that roughness can increase residual clamping stress
during fabrication49. Both the residual stress and friction can explain the increased sliding strength for the
SA3 composite. These constituent level values are critically important from a modelling perspective. The
next generation of modelling needs both the constituent properties as well as information regarding the type
of damage that is evolving. This hysteresis testing in conjunction with the micro-mechanical testing has
provided a unique platform to identify these characterstics and will prove useful for the continued modelling
efforts.
2.4.5.3 Mechanical testing of woven composites General Atomics performed a series of mechanical testing of SiC/SiC composites. Planar and tubular shaped
composite samples were tested and were all fabricated at GA using the Chemical Vapor Infiltration method
(CVI). Planar composites were reinforced with Tyranno-SA3, and tubular composites were made with Hi-
Nicalon Type S fiber. Composite tubes with a fiber ratio of hoop to axial directions at 2:1 and a pyrolytic
carbon interphase layer ~100nm were used for elastomeric expending plug test that is based on ASTM C1819,
and c-ring testing at room and 800°C test temperatures based on ASTM C1323 standard, and 4-point bed
tests. The test method and results are described below. Provided the load versus displacement data for bulk
composite test specimens provides a standard to validate code against.
Flexural 4-point bend test results for 8 specimens are shown in Table 7. The composites in planar shape were
fabricated with Tyranno SA3 fiber. All specimens with the nominal size of 52 mm by 6.8 mm by 1.1 mm
were cut from a single plate using a waterjet technique. After the initial cut, the longest sides were polished
with a 30 micron diamond disc to remove large defects. The final width of the test specimens contained about
two repeating fiber unit cells. The loading pins of the test fixture were spaced by 1/4 of the loading span
which was 42 mm. A typical flexural stress versus extension is shown in Figure 62a, small load drops are
observed upon loading, representative of microcrack formation.
Table 7 - Flexural 4-point bend test results
For the flexural testing, while the specimens underwent a significant amount of bending, and showed signs
of micro-crack evolution, the fractured fragments indicated a more brittle like behavior with minimal fiber
pull out, shown in Figure 62b. In 7 out of 8 specimens fracture occurred at one or both loading pins, which
are typically considered non-valid tests in 4-point bend testing.
For 8 specimens: Extension, mm Flexural Stress, MPa
Average Values 1.16 (±0.20) 304.84 (±29.06)
Figure 62 – (a) Typical Flexural Stress vs Extension plot (b) Fracture surface failed composite. Limited fiber pullout
is observed. (c) Macroscopic view of fractured test specimen.
To better understand the flexural test data, a dog bone shaped planar specimen, cut using waterjet method,
was loaded in tension with Digital Image Correlation (DIC) strain measurement technique. The specimen
geometry, test setup, and typical stress versus strain curve are shown in Figure 63. An example of DIC output
and a resulting stress-strain plot is shown in Figure 64. The DIC strain mapping reveals a pattern of high and
low regions of strain. This is attributed to the non-homogenous woven composite structure. This type of
stress/strain evolution is important to incorporate into the modelling effort.
Figure 63 – (Left) The dog bone specimen shape. (Center) Image of an axial tension test setup. (Right)
Stress-strain curve for a dog bone shaped specimen from a planar tension test
The tensile specimens showed pseudo-plastic behavior including micro-crack load drops and similar amount
of fiber pull-out compared to flexural testing (Figure 64).
a b
c
Figure 64 -. (Right) Example of digital image correlation strain map during test. (Right) Typical fracture
surface observed after mechanical testing of planar SiC/Sic samples, showing limited fiber pull-out.)
For the elastomeric expanding plug test the elastic modulus and the Ultimate Tensile Stress (UTS) calculated
for the outside diameter are shown for 10 specimens in Table 8. Tubular specimens with a nominal 9.58 mm
outside diameter (OD) and a 0.86 mm wall thickness were cut to approximately 25 mm in length and polished
on the cut edges with a 30 micron diamond disc. A polyurethane cylindrical insert with 95 durometer hardness
was sprayed with Teflon lubricant and then inserted inside of the test tube. The polyurethane plug was then
compressed by a universal testing machine to induce pressure on the inside wall of the composite. Thick wall
pressure vessel equations were used to calculate stress values on the outside wall. Four out of ten specimens
were setup with a Digital Image Correlation system to measure the strain maps on the outer surface allowing
to generate stress-strain curves.
Table 8 - Hoop direction mechanical properties
Test UTS, MPa
Expanding plug 330.87 (±20.31)
RT C-ring 290.56 (±65.37)
800°C C-ring 308.17 (±51.57)
A typical stress-strain curve from an expanding plug test is shown in Figure 65. The composite exhibited a
gradual transition from linear to pseudo-ductile regions. The elastic modulus of 160.65 ± 10.16 GPa was
found from a region on the stress-strain curve between 200 and 450 microstrains. Using the resulting elastic
modulus and a 0.01% offset strain method, a proportional limit strength was found to be 132.25 ± 7.29 MPa
at average strain of 782 ± 8 microstrains.
Figure 65 - A representative stress-strain curve from an elastomeric expanding plug test.
C-ring testing was performed at room temperatures (RT) and at 800°C with 10 specimen sets for each test.
The resulting ultimate tensile stress is shown for Table 8. A typical load vs displacement curve is shown in
Figure 66. Due to composite rough and stiff outside surface, a compliant material was placed between the
specimens and the compressing fixtures. For the high temperature tests, the specimens were heated in an
argon environment and dwelled at test temperatures for a minimum of 30 minutes. High viscosity grease was
used to keep the c-rings properly oriented during both tests.
Figure 66 - Load versus crosshead extension for a typical RT and High Temperature c-ring tests
Both room temperature and the high temperature testing showed a composite like behavior with a significant
amount of deformation after reaching the load peak. It should be noted that the room temperature result
between the c-ring and expanding plug tests differ, but the expanding plug data should be considered to be
more accurate because it loads a significantly larger amount of material than the c-ring test and therefore the
results are less impacted by local defects. Also, the high temperature c-ring results show no strength
degradation when compared to the room temperature c-ring testing. This strength retention would be expected
for the expanding plug test as well if it were to be performed at high temperatures.
FEA model development This effort was focused on the development of computational models that: 1) incorporate known constituent
material behavior of the SiC-SiC system, and 2) predict macroscopic behaviors observed during testing of
large sample constructs such as SiC fiber “tows” (groups of fibers in parallel strands). Material properties for
constituents were determined from nano-indentation (2.4.2.1), micro-cantilever testing (Error! Reference
source not found.), micro-pillar compression (2.4.2.3), and fiber pushout tests (Error! Reference source
not found.). Interactions between the constituents were explored through tensile testing of mini-composites
consisting of a single unidirectional SiC fiber tow.
Finite element models of micro-pillar compression tests and fiber pushout tests were developed to ensure
agreement between experimental data and implementations of constituent behavior in models. These
experimental geometries are meant to develop an understanding of constituent characteristics and to
mechanistically inform macroscale simulations. The planned geometries for macroscale component testing
are those of the flexure and hoop strength test configurations for planar and tube structures respectively.
However, the current state of this model is using mini-composite geometry and attempting to mimic behavior
observed in the in situ x-ray tomography testing section.
Microscale Models Microscale models were developed to simulate well-controlled experiments to isolate constituent properties.
Examples include micropillar compression tests and fiber pushout tests. Together, these tests contribute to
improved understanding of the pyrolytic carbon (PyC) layer and ability of the macroscopic model to capture
mechanistic behavior.
2.4.7.1 Micropillar Compression Model Micropillar compression tests at a variety of interface angles reveal the shear strength and fracture energy of
the PyC layer as a function of normal stress. Models of the micropillar compression have been developed
that reproduce the observed behavior. There are several approaches to modeling the PyC. First, it is possible
to explicitly model the PyC layer with Mohr-Coulomb material behavior. Second, the PyC layer could be
described with cohesive surface contact and frictional contact behavior. Lastly, a hybrid of the two could be
used, since it is observed that the PyC consistently fails at the interface with the fiber.
Advantages of Mohr-Coulomb material models include relatively realistic modeling of the PyC layer and
relatively low mesh sensitivity. A downside of this method is that the very thin interface layer elements break
down relatively small deformations and preclude modeling larger scale decohesion of the fiber and matrix as
the PyC layer fails. Representation of the PyC layer with cohesive and frictional surface contact allows for
complete decohesion and sliding and is more applicable to larger scale models where decohesion is important,
but early implementations neglect details of the PyC layer such as thickness, the PyC material stiffness, and
fiber roughness.
Both Mohr-Coulomb and cohesive surface models were developed and compared to test data and produce
identical behavior for predictions of the micropillar compression test, as shown in Figure 67. Cohesive
surfaces have been selected for better compatibility with larger length scale models where full decohesion
and sliding is necessary. The micropillar data shown below is from HNLS fibers with 50nm PyC, with an
observed 𝜏𝑑𝑒𝑏𝑜𝑛𝑑 of 266 𝑀𝑃𝑎 and an internal friction coefficient of 0.25.
Figure 67 – The leftmost image shows a finite element model using cohesive surfaces after decohesion has
occurred. Results from this model are shown as the solid green line on the plot at right. The center picture
shows a model with Mohr-Coulomb plasticity in a thin layer (which appears in red) just after the layer
fails. Results from this type of model are shown as circular red dots in. The remaining data points in the
plot are experimental results.
2.4.7.2 Fiber Pushout Model Fiber pushout tests are the next step increase in length scales from the micropillar compression tests. These
tests also isolate behavior of the PyC layer, but with several important differences:
1) Fiber pushout tests provide sliding stress, while micropillar compression tests generally only provide the
fracture initiation stress at the PyC interface.
2) The stress normal to the to the interface is controlled by residual stresses and fiber roughness.
Fiber pushout models were based on experiments with SA3 fibers with relatively thick (100 nm) PyC layer
with an assumed 𝜏𝑑𝑒𝑏𝑜𝑛𝑑 of 119 𝑀𝑃𝑎 as found by push-out testing presented in section 2.4.2.3. Modeling of
the fiber pushout experiments was performed with an axisymmetric model. Figure 68Figure 68 - Fiber
pushout model (left) and comparison with experiments (right). The 2D axisymmetric model was revolved
180° for visualization purposes. Model predictions are plotted at right in the orange dashed line. Note that
the load at which the PyC layer fractures and the load begins to drop is determined in part by residual stress
between the fiber and matrix, which is unknown. shows the axisymmetric model revolved 180° for
visualization.
The deformed model is shown at left in Figure 68 while a series of experimental load vs. displacement data
are shown at right for multiple fibers. The model prediction is shown in the bold orange dashed line. Note
that the model predicts that the load drops gradually as damage accumulates in the PyC layer, while the
experimental results are relatively constant or increase with additional displacement. Furthermore, the load
at which the PyC layer fails is determined by 𝜏𝑑𝑒𝑏𝑜𝑛𝑑, the internal friction coefficient, and the residual stress
state in the fiber and matrix, which is unknown. The model prediction shown below is consistent with
expectations based on Mohr-Coulomb behavior, with frictional forces decreasing as the fiber is pushed out
of the matrix and the area of the fiber/matrix interaction decreases. Overall, the model provides a reasonable
representation of the data, particularly considering that the state of residual stress in the fiber and matrix,
which determines the normal stress and ultimately cohesive strength of the PyC layer, is unknown. These
results support the use of cohesive surface contact for modeling the PyC layer.
Figure 68 - Fiber pushout model (left) and comparison with experiments (right). The 2D axisymmetric
model was revolved 180° for visualization purposes. Model predictions are plotted at right in the orange
dashed line. Note that the load at which the PyC layer fractures and the load begins to drop is determined
in part by residual stress between the fiber and matrix, which is unknown.
Mini-composite model As discussed in the tomography section Error! Reference source not found., experiments were conducted
at the ALS at LBNL. SiC/SiC mini-composites were heated and pulled in tension until failure. The fiber tow
consists of ~500 parallel SiC fibers infiltrated with SiC matrix to produce a test sample of roughly 0.75mm
x 5mm in cross-section. The beam window was focused around the notched area of the specimen as to capture
the location of failure and related mechanisms. A segment of the fiber tow with notch and micro-crack
evolution is shown in Figure 69. While these specimens contain hundreds of fibers, initial finite element
models contain only a few fibers and focus on the initiation of matrix and fiber fracture and the resulting
redistribution of stresses.
Figure 69 – Raw tomography projection scan of partially broken mini-composite with pre-machine notch
on right. Note the array of cracks through the composite matrix.
Notch
SiC Matrix cracks
Extending microscale models to the mesoscale mini-composite models is challenging for several reasons.
1) While the microscale models focused on failure or decohesion of the fiber/matrix interface, the mini-
composite models include fracture of the matrix and fiber as well.
2) Models for fracture are an active area of research in FEA and are challenging.
3) The geometry of the notched fiber “tow” prevents the use of symmetry or unit cell modeling. Because
the notch is the source of the radial cracking
4) The mini-composite must be modeled (on the mm scale), while the fibers must be modeled (on the μm
scale) to include their behavior.
Modeling large scale fracture, decohesion, and sliding of hundreds of fibers was unfortunately beyond the
time scope of this work and beyond the state of the art. Instead, this work focuses on the initial fracture of
the fiber tow. Presently, the models discussed herein are not able to predict the cracking patterns observed in
testing of mini-composites and more development is required.
2.4.8.1 Modelling Approaches First, simplified multi-fiber models were constructed with cohesive contact between fibers and matrix as well
as with predefined fracture paths with cohesive contact. While simple two-dimensional models such as those
shown at left in Figure 70 provided encouraging results. The next step was to try simple three-dimensional
models (shown at right in Figure 70), but had difficulty to converge and were much more computationally
expensive. These tests demonstrated that this technique was not suitable for expansion to comprehensive
mini-composite geometry.
Figure 70 - Simple two-dimensional single fiber model (left) and three-dimensional multi fiber model used
to explore the feasibility of a larger-scale mini-composite (right). These models used predefined cracks
with cohesive contact as well as cohesive contact between the matrix and fibers. While results from the
two-dimensional model were encouraging, convergence in the three-dimensional model was very poor and
even the simple model shown at right was computational expensive.
A survey of other techniques for modelling damage in fiber reinforced composites pointed to Hashin damage
models, which homogenize the behavior of the fiber and matrix materials to predict the system response 50,51.
Hashin damage models define a three-dimensional failure surface that includes compressive and tensile
failure modes in the fiber and matrix. Because fiber behavior is modeled at the continuum scale, these models
are appropriate for component-scale geometries that contain many fibers. A disadvantage, however, is that
the details of matrix/fiber interactions are difficult or impossible to capture. A mini-composite model using
the Hashin damage model is shown in Figure 71.
Figure 71 - Two-dimensional homogenized model of fiber "tow" with Hashin damage model. The red
elements show the damaged material and the predicted crack path
Results from the Hashin damage model are encouraging, particularly for extension to larger length-scale
models. Further work should be considered to include the toughening effects of the PyC interface between
the SiC matrix and fibers. Because the details of the fiber and interface behavior were of primary interest in
this work, the Hashin models were not pursued further.
A compromise between explicit fiber models and homogenized models may be found in embedded element
methods, which are typically used for modeling concrete with rebar, but have also been applied to fiber
reinforced composites52. With embedded elements, finite elements defining reinforcing fibers are embedded
in host (matrix) elements, and translational degrees of freedom in the fiber elements are constrained to move
with the host elements. Different material properties and damage criteria are assigned to the fiber and matrix
to model the evolution of damage within the composite. An example of the embedded element technique
applied to the mini-composite application is shown in Figure 72.
The material models selected for the fiber and matrix materials within the embedded element methods are
concrete damaged plasticity53. While these material models were developed for concrete, they are also
appropriate for modelling damage and fracture in other quasi-brittle materials. The embedded element
technique and concrete damaged plasticity material models were applied to the mini-composite geometry,
but instead of the radial array of cracks observed in the notched test, a single crack propagates across the
models. This behavior is consistent with a homogeneous brittle material, and does not characterize the
toughening observed in the SiC-SiC composites. Interface effects are typically added to concrete damaged
plasticity models through “tension stiffening”, but additional stiffening did not change the predicted cracking
behavior in these models. This suggests that this method is not fully capturing the mechanistic failure
behavior.
(
a
)
(
b
)
(
c
)
Figure 72 - Results from Embedded Element Technique with Concrete Damaged Plasticity material
models. This technique includes separate finite elements for the matrix and fibers. The matrix elements are
shown with contours for material degradation (a) and equivalent plastic strain (b). These figures show the
progression of failure in the matrix. Fiber elements are shown in (c). Here, contours have the same scale as
(a) and show degradation of the fiber material. Note that this model predicts that fracture propagates
across the specimen and does not capture the radial cracking pattern shown in testing.
Several modelling techniques were introduced and applied that take the constituent properties and failure
mechanisms and try to capture the failure behavior of macroscopic composites. The mini-composite
geometry challenges current finite element techniques. Finite element models were developed using a variety
of techniques, including cohesive surface contact, Hashin damage models, and the embedded element
technique. Exploration of these techniques revealed that models that include individual fibers with cohesive
surface contact are not extensible to larger scale geometries. As geometric complexity (number of fibers and
interfacial area) increases, computational expense and convergence of the models become untenable.
Unfortunately, the methods above were not able to link these length scales. This has been an ongoing
challenge within the community and will require continued attention and development.
To summarize the modelling work, finite element models were developed to simulate micropillar
compression experiments, fiber pushout experiments, and minicomposite tensile tests. Models of micropillar
compression experiments were used to demonstrate agreement of constitutive models of the PyC interface
layer with experimental data. Two models were explored: Mohr-Coulomb material behavior and cohesive
surface contact. Both of these models showed excellent agreement with experimental results and cohesive
surface contact was selected for compatibility with larger-scale models. Fiber pushout models were
developed to test the simulated fiber/matrix decohesion behavior on a larger scale. These models applied
cohesive surface contact between the matrix and fiber and showed reasonable agreement with the
experimental data. The fiber pushout models and experiments have limited resolution, as the failure behavior
of the PyC interface layer depends on residual stresses in the SiC-SiC composite, which are unknown and
likely not uniform. The mini-composite geometry challenges current finite element techniques. Finite
element models were developed using a variety of techniques, including cohesive surface contact, Hashin
damage models, and the embedded element technique. Exploration of these techniques revealed that models
that include individual fibers with cohesive surface contact are not extensible to larger scale geometries. As
geometric complexity (number of fibers) increases, computational expense and convergence of the models
become untenable. For extension to the macroscale, particularly when multiple mini-composites are woven
into a structure, homogenization of the matrix and fiber is required. For exploration of fundamental fracture
behavior of composites, techniques such as peridynamics or phase field, which support general and complex
fracture paths, should be considered. For more practical applications to composite behavior, both the Hashin
damage models and embedded element technique show promise.
a b c
Summary and Conclusions
From this, work, several notable advances and results have been obtained to quantify constituent properties
of SiC/SiC composites. This improved understanding of constituent micro-scale and localized properties will
enable more detailed modeling of the composite behavior and support adaptation and use of these materials
in nuclear applications. Furthermore, the experimental techniques developed in this work can be applied to a
wide range of composite materials and other materials with complex micro-structures. These localized
characterization methods can be used, and the approach of introducing these property measurements into
micro-scale and macro-scale behavior models can be applied to materials beyond SiC/SiC.
Microstructural characterization revealed that the microstructure of these composites is very complex, with
hierarchical levels of structure especially within matrix, which consists of grain elongated in the radial
direction away from the matrix (due to CVI), organized in multiple concentric rings (reflecting discontinuities
in the growth process) and into domains that are grown originating from different fibers. CVI growth process
leads to noticeable porosity as well, which has a complex size and spatial distribution – ranging from
nanometer-sized to tens of microns pores, within the matrix and in the fiber bundles, with its density highest
in the center of the bundles and lower towards the periphery and in the matrix. Using microcantilever fracture
it was shown, however, that fracture is transgranular, and therefore various boundaries within the matrix
(grain boundaries, boundaries between concentric rings and boundaries between domains) don’t appear to
influence fracture significantly. Indeed cantilevers oriented differently relatively to the direction of grain
growth produced similar values of fracture stress and toughness. These results indicate that despite its
microstructural complexity, for the purposes of modeling matrix can still be treated as a uniform medium –
however, distribution of porosity, particular spatial distribution, needs to be taken into account (as shown
more prominently by push-out testing).
Comparison of the material parameters obtained both from nanoindentation (elastic modulus) and
microcantilever testing (fracture toughness) shows that the values obtained are very close to those obtained
in macroscopic bulk measurements – where such can be done, i.e., for matrix and fiber constituents. This
validates the chosen approach, proving that micromechanical measurements can be used as a reliable testing
tool for determination of localized mechanical properties. At the same time, only micromechanical testing is
able to measure the properties of interphases. It offers other advantages as well, in particular, allowing
determination of relevant material properties from very small amounts of material (valuable for composite
design and development, since CVI growth process is slow, and thus needing only small amount of material
would allow much faster throughput if different growth procedures, interfacial structures, different fibers etc
are to be tested), and also being able to probe the near-surface material properties, which can be used for the
studies of radiation effects using ion irradiation, without needing to necessarily resort to expensive,
technically challenging and time-consuming measurements on neutron-irradiated samples.
Nanoindentation measurements performed in a linescan mode, where multiple shallow indents can be placed
within a fiber, indicated the hardness and modulus of a fiber is non-uniform, with hardness decreasing from
40 GPa at the periphery of the fiber (similar to matrix value) to 19 GPa in the center, and modulus decreasing
from 460 GPa at the periphery to 260 GPa in the center. This is correlated with the results of TEM and EDX,
showing the presence of carbon particles in the fiber material. Density of this residual carbon is maximum in
the center and decreases to almost zero at the periphery. Hence, carbon residue leads to the reduction of
hardness and modulus.
Micro-cantilever testing provided a novel approach to obtaining mechanical response data with the ability to
isolate fracture in either the fiber, interphase, or matrix. In this work, fracture at the pyrolytic carbon
interphase was observed to occur either at the interface between the interphase material and the fiber, or
within the carbon interphase layer itself, and further investigation is needed to identify what properties
contribute to this. Micro-cantilever testing allowed the measurement of toughness of the fiber, interphase,
and matrix constituents before and after irradiation, and an irradiation-toughening response was observed for
both the fiber and the interphase, again providing valuable constituent-scale properties which would not be
possible to obtain from macro-scale testing. It was found that matrix is the strongest, fibers intermediate and
interphases are the weak spots. The reason for the difference between fiber and matrix was traced back to the
presence of carbon particles decorating the grain boundaries. TEM imaging of fractured cantilevers
demonstrated that while in the matrix fracture is transgranular, in the fiber it can be transgranular but also
intergranular – depending on the local presence of carbon, with more carbon promoting intergranular fracture.
Hence it can be concluded that carbon-induced weakening of grain boundaries leads to lowering of fracture
properties of SiC fibers compared to CVI-grown SiC matrix. Since these carbon particles are non-uniformly
distributed within a fiber, radial non-uniformity of fracture properties can also be assumed (although this
wasn’t verified directly). Hence, combining the results of nanoindentation and cantilever testing one can
conclude that radial non-uniformity of fiber properties needs to be assumed for the purposes of modelling.
TEM on fractured cantilevers at the interphase indicates that fracture occurs at the fiber-interlayer boundary,
indicating that the internal structure of the interlayer is of secondary importance, and fiber-interlayer bonding
is primarily responsible for the interfacial fracture properties.
Push-out testing protocol was developed that allowed very high number of tests (in the hundreds), allowing
for statistically sound determination of interfacial shear strength. For the first time these measurements were
performed talking into account the local microstructure and its non-uniformity. As microscopic examination
reveals, fiber bundles in particular are non-uniform, with porosity high and inter-fiber distances low in the
centers of the bundles, and with low porosity and fibers being far apart at the periphery of the bundles.
Therefore, in the center of a bundle an effect of the local environment is noticeable because of the interaction
between a tested fiber and surrounding porosity and other fibers, while tests at the periphery result in a true
value of ISS, eliminating the influence of environment. A noticeable difference has been indeed observed,
with ISS measured in the bundle center being noticeably lower than that at the periphery (~70 MPa vs ~120
MPa).These results indicate that for the purposes of realistic modelling of the macroscopic behavior of the
composites the fiber bundles should not be treated as uniform, and the distribution of effective ISS increasing
from the center towards the periphery, has to be assumed.
Nanoindentation measurements following ion irradiation indicate that such irradiations do not lead to
significant changes in hardness (slight increase) and modulus (slight decrease) in the constituents of a
composite, neither matrix nor fiber. Ten-fold increase of irradiation dose, from 0.26 to 2.6 dpa, doesn’t lead
to noticeable changes in hardness and modulus either. On the other hand, cracking patterns surrounding the
indents do change, with the radial crack being very prominent in the unirradiated material and essentially
eliminated in the irradiated material, even after low irradiation dose of 0.26 dpa.
Microcantilever testing indicated that fracture toughness is somewhat impacted by ion irradiation. There isn’t
a definitive trend to increase of toughness in the matrix, with some conditions leading to increase and other’s
not. In the fiber it appears that toughens increases with the increase of dose. Toughens of the interphases
increases after ion irradiation.
Micro-pillar compression testing was a novel approach that was advanced significantly under the scope of
this NEUP. This technique was able to explore interfacial shear and fracture properties as a function of fiber
type, PyC thickness, and residual stress. It was found that rough fibers and thin interfaces increase both
chemical bonding strength and internal friction contribution. TEM analysis of fracture pillars revealed
supporting evidence that failure occurs within the PyC but near the fiber side. The examination of irradiated
SiC/SiC composite material showed two interesting effects of irradiation on the pyrolytic carbon interphase
between the fibers and the matrix. Micro-pillar compression testing shows a reduction in the debonding shear
stress of the pyrolytic carbon interphase following irradiation, without much change in the friction
coefficient. This is consistent with previous results from the literature. The evolution of this behavior as a
function of irradiation damage and irradiation temperature is important to understand the composite response
at the fiber-scale and to understand how toughening mechanisms may change. This is particularly important
as the fiber and interphase response will affect the propagation of micro-cracks as the composite is loaded
beyond the proportional limit stress, and how the formation and interconnection of micro-cracks impacts the
ability of a SiC/SiC cladding structure to remain gas-tight and retain fission gas. This is a critical function of
cladding, and understanding these irradiation effects at the micro-scale is important to model the behavior
and to correlate with macro-scale measurements. Given the different observations from this current work and
the literature, this highlights a need for additional investigation in this area, to look at other factors which
may influence this result (fiber type, processing differences, etc.). In order to better understand fracture
behavior of the constituents of the composite another set of irradiations is necessary, covering the irradiation
conditions not previously explored – low-temperature, high-dose and high-temperature, low-dose cases.
The thermal analysis performed in this project provides experimental confirmation of micro-scale properties
which are found to be consistent with what would be predicted based on macro-scale or bulk properties. This
provides an important validation of the experimental methods applied. Fiber thermal conductivity values
between 17 W/m-K and 25 W/m-K are in-line with reported macro-scale values for the Hi-Nicalon type-S
fiber measured. The variation in fiber conductivity as a function of radial position matches with the observed
radial compositional variation observed within the fibers, with higher carbon concentration (and
corresponding lower thermal conductivity) near the center of the fibers. Thermal conductivity of the CVD
SiC matrix was ~75 W/m-K with a variation of ~20%. This value is consistent with that reported in the
literature for monolithic, polycrystalline, small-grained CVD SiC. A ~20% variation was observed within
the SiC matrix, and this observation, and the potential correlation of this variation with spatial variations in
the grain structure and matrix porosity, is an area of important potential future work. The thermal conductivity
measured for the pyrolytic carbon interphase was much lower than that of the SiC constituents, at ~6.5 W/m-
K. This lower value is consistent with observations in the literature which have found macro-scale composite
thermal conductivity decreases with increasing interphase content (either through thicker carbon interphase
layers, or through the use of multi-layered interphases). This reduction in thermal properties, combined with
the reduced debonding strength observed in thicker interphases, suggests a thin interphase may offer a better
combination of mechanical and thermal performance in a SiC/SiC composite than a thick interphase.
Macroscopic composite characterization and testing was carried out applying state of the art XRT
technologies and ASTM standards. Porosity was found to range from 0.5 to 5% for planar and tubular
geometries. In situ XRT was successfully carried out at ambient and 700°C for ZMI mini-composites. Micro-
pillar compression showed interface τdebond = 407MPa and μi = 0.17. Micro-crack spacing was found to be
very similar for low and high temperature testing, on the order of 0.35mm. Suggesting that in an inert
environment, there is no fundamental change in interface characteristics up to 700°C. Woven composite
testing including 4-point bend, tensile, expanding plug, and C-ring testing provided additional load versus
displacement behavior to ultimately aid in model validations. Composite behavior showed evidence of
microcrack evolution and deflection, suggesting successful composite construction, and leaving room for
optimization through SSMT characterization. C-ring tests at ambient and 800°C showed very similar load
versus displacement behavior, supporting evidence that the interface characteristic remain constant at
elevated temperatures.
Finally, the fabrication and use of single fiber samples is also novel, and represents perhaps the most idealized
approach to testing a complete composite structure (containing a fiber, interphase, and matrix, in the simplest
possible geometry). For the micro-scale localized tests, models were developed to represent the actual test
configurations (such as micro-pillars and fiber push-out tests). However, simulation of macro-scale behavior
(including detailed fiber architecture and porosity within the matrix), can be very complex, and even the
single-tow composite model necessarily approximated the structure (fiber position and matrix). A single fiber
sample geometry may provide an ideal “bridge” between micro-scale localized tests, and macro-scale tests,
and may offer a geometry suitable for accurate representation in a model.
Overall, the work performed under this project was able to successfully obtain measurements of SiC/SiC
composite constituent (fiber, matrix, and interphase) mechanical and thermal properties through a
combination of micro-scale localized testing. The approach to incorporate these constituent properties into
micro- and macro-scale models was demonstrated, and this is an important step towards more accurate
simulation of SiC/SiC-based cladding performance based on experimentally-confirmed micro-scale
behaviors.
References
1. Ishikawa, T., Kohtoku, Y., Kumagawa, K., Yamamura, T. & Nagasawa, T. High-strength alkali-resistant
sintered SiC fibre stable to 2,200 C. 20, 773–775 (1998).
2. anaka, T., Shibayama, S., Takeda, M. and Yokoyama, A. Recent Progress of Hi-Nicalon Type S Development,
in 27th Annual Cocoa Beach Conference on Advanced Ceramics and Composites: B. Ceram. Eng. Sci. Proc.
24, 217–223 (2003).
3. Ichikawa, H. Development of High Performance SiC Fibers Derived from Polycarbosilian Using Electron
Beam Irradiation Curing-A review. ournal Ceram. Soc. Japan 114, 455–460 (2006).
4. Sauder, C. & Lamon, J. Tensile creep behavior of SiC-based fibers with a low oxygen content. J. Am. Ceram.
Soc. 90, 1146–1156 (2007).
5. Katoh, Y. et al. Continuous SiC fiber, CVI SiC matrix composites for nuclear applications: Properties and
irradiation effects. J. Nucl. Mater. 448, 448–476 (2014).
6. Sauder, C., Brusson, A. & Lamon, J. Influence of interface characteristics on the mechanical properties of Hi-
nicalon type-S or tyranno-SA3 fiber-reinforced SiC/SiC minicomposites. Int. J. Appl. Ceram. Technol. 7, 291–
303 (2010).
7. Katoh, Y., Snead, L. L., Nozawa, T., Kondo, S. & Busby, J. T. Thermophysical and mechanical properties of
near-stoichiometric fiber CVI SiC/SiC composites after neutron irradiation at elevated temperatures. J. Nucl.
Mater. 403, 48–61 (2010).
8. Kabel, J. et al. Micro-Mechanical Evaluation of SiC-SiC Composite Interphase Properties and Debond
Mechanisms. Compos. Part B Eng. 131, 1–18 (2017).
9. Dimovski, S., Nikitin, A., Ye, H. & Gogotsi, Y. Synthesis of graphite by chlorination of iron carbide at
moderate temperatures. 238–243 (2004).
10. Tsai, S. C., Huang, E. W., Kai, J. J. & Chen, F. R. Microstructural evolution of nuclear grade graphite induced
by ion irradiation at high temperature environment. J. Nucl. Mater. 434, 17–23 (2013).
11. Che, G., Lakshmi, B. B., Martin, C. R., Fisher, E. R. & Ruoff, R. S. Chemical Vapor Deposition Based
Synthesis of Carbon Nanotubes and Nanofibers Using a Template Method. 4756, 260–267 (1998).
12. Sato, Y., Bugnet, M., Terauchi, M., Botton, G. A. & Yoshiasa, A. Heterogeneous diamond phases in
compressed graphite studied by electron energy-loss spectroscopy. Diam. Relat. Mater. 64, 190–196 (2016).
13. Perez-Bergquist, A. G. et al. High dose neutron irradiation of Hi-Nicalon Type S silicon carbide composites,
Part 1: Microstructural evaluations. J. Nucl. Mater. (2014). doi:10.1016/j.jnucmat.2014.06.038
14. Yan, J. Y. et al. The investigation of crack mechanism for Tyranno-SA SiC/SiC composites with ESI method.
J. Nucl. Mater. 329–333, 513–517 (2004).
15. Niihara, K., Morena, R. & Hasselman, D. P. H. Evaluation ofK Ic of brittle solids by the indentation method
with low crack-to-indent ratios. J. Mater. Sci. Lett. 1, 13–16 (1982).
16. Oliver, W. C. & Pharr, G. M. An improved technique for determining hardness and elastic modulus using load
and displacement sensing indentation experiments. J. Mater. Res. 7, 1564–1583 (1992).
17. Frazer, D. et al. Localized mechanical property assessment of SiC/SiC composite materials. Compos. Part A
Appl. Sci. Manuf. 70, 93–101 (2015).
18. Di Maio, D. & Roberts, S. G. Measuring fracture toughness of coatings using focused-ion-beam-machined
microbeams. J. Mater. Res. 20, 299–302 (2005).
19. Armstrong, D. E. J., Wilkinson, A. J. & Roberts, S. G. Measuring anisotropy in Young’s modulus of copper
using microcantilever testing. J. Mater. Res. 24, 3268–3276 (2009).
20. Norton, A. D., Falco, S., Young, N., Severs, J. & Todd, R. I. Microcantilever investigation of fracture
toughness and subcritical crack growth on the scale of the microstructure in Al2O3. J. Eur. Ceram. Soc. 35,
4521–4533 (2015).
21. Matoy, K. et al. A comparative micro-cantilever study of the mechanical behavior of silicon based passivation
fi lms. Thin Solid Films 518, 247–256 (2009).
22. Vagaggini, E., Domergue, J. ‐M & Evans, A. G. Relationships between Hysteresis Measurements and the
Constituent Properties of Ceramic Matrix Composites: I, Theory. Journal of the American Ceramic Society
78, 2709–2720 (1995).
23. Snead, L. L. et al. Handbook of SiC properties for fuel performance modeling. J. Nucl. Mater. 371, 329–377
(2007).
24. Kabel, J. et al. Ceramic composites: A review of toughening mechanisms and demonstration of micropillar
compression for interface property extraction. J. Mater. Res. 33, (2018).
25. Katoh, Y., Snead, L. L., Nozawa, T., Kondo, S. & Busby, J. T. Thermophysical and Mechanical Properties of
Near-stoichiometric Fiber CVI SiC/SiC Composites after Neutron Irradiation at Elevated Temperatures. J.
Nucl. Mater. 403, 48–61 (2010).
26. Shih, C., Katoh, Y., Leonard, K. J., Bei, H. & Lara-Curzio, E. Determination of interfacial mechanical
properties of ceramic composites by the compression of micro-pillar test specimens. J. Mater. Sci. 48, 5219–
5224 (2013).
27. Handin, J. On the Coulomb-Mohr failure criterion. J. Geophys. Res. 74, 5343–5348 (1969).
28. Yang, W. et al. Interfacial characterization of CVI-SiC / SiC composites. J. Nucl. Mater. 311, 1088–1092
(2002).
29. Bertrand, S., Droillard, C., Pailler, R., Bourrat, X. & Naslain, R. TEM structure of ( PyC / SiC ) n multilayered
interphases in SiC / SiC composites. J. Eur. Ceram. Soc. 20, 1–13 (2000).
30. Liu, Z. et al. Interlayer shear strength of single crystalline graphite. Acta Mech. Sin. Xuebao 28, 978–982
(2012).
31. TATTERSALL, H. G. & Tappin, G. The Work of Fracture and its Measurement in Metals , Ceramics and
other Materials. J. Mater. Sci. 1, 296–301 (1966).
32. Cedric, Z. & Hutchinson, J. W. MODE II FRACTURE TOUGHNESS OF A BRITTLE ADHESIVE LAYER.
Int. J. Solids Struct. 31, 1133–1148 (1994).
33. Evans, A. G. & Zok, F. W. The physics and mechanics of fibre-reinforced brittle matrix composites. J. Mater.
Sci. 29, 3857–3896 (1994).
34. Kabel, J. et al. Ceramic composites: A review of toughening mechanisms and demonstration of micropillar
compression for interface property extraction. J. Mater. Res. 33, 1–16 (2018).
35. López-Honorato, E. et al. Thermal conductivity mapping of pyrolytic carbon and silicon carbide coatings on
simulated fuel particles by time-domain thermoreflectance. J. Nucl. Mater. 378, 35–39 (2008).
36. Kang, K., Koh, Y. K., Chiritescu, C., Zheng, X. & Cahill, D. G. Two-tint pump-probe measurements using a
femtosecond laser oscillator and sharp-edged optical filters. Rev. Sci. Instrum. 79, 114901 (2008).
37. Wilson, R. B., Apgar, B. A., Martin, L. W. & Cahill, D. G. Thermoreflectance of metal transducers for optical
pump-probe studies of thermal properties. Opt. Express 20, 28829 (2012).
38. Cahill, D. G. Analysis of heat flow in layered structures for time-domain thermoreflectance. Rev. Sci. Instrum.
75, 5119–5122 (2004).
39. Karthik, C., Kane, J., Butt, D. P., Windes, W. E. & Ubic, R. In situ transmission electron microscopy of
electron-beam induced damage process in nuclear grade graphite. J. Nucl. Mater. 412, 321–326 (2011).
40. Karthik, C., Kane, J., Butt, D. P., Windes, W. E. & Ubic, R. Neutron irradiation induced microstructural
changes in NBG-18 and IG-110 nuclear graphites. Carbon N. Y. 86, 124–131 (2015).
41. Takeuchi, M., Muto, S., Tanabe, T., Kurata, H. & Hojou, K. Structural change in graphite under electron
irradiation at low temperatures. J. Nucl. Mater. 271 & 272, 280–284 (1999).
42. Snead, L. L., Burchell, T. D. & Katoh, Y. Swelling of nuclear graphite and high quality carbon fiber composite
under very high irradiation temperature. J. Nucl. Mater. 381, 55–61 (2008).
43. Bale, H. A. et al. Real-time quantitative imaging of failure events in materials under load at temperatures
above 1,600 °C. Nat. Mater. 12, 40–6 (2013).
44. Haboub, A. et al. Tensile testing of materials at high temperatures above 1700 degrees C with in situ
synchrotron X-ray micro-tomography. Rev. Sci. Instrum. 85, 1–13 (2014).
45. Curtin, W. A. Theory of Mechanical Properties of Ceramic‐Matrix Composites. J. Am. Ceram. Soc. 74, 2837–
2845 (1991).
46. Domergue, J. M., Vagaggini, E. & Evans, A. G. Relationships between Hysteresis Measurements and the
Constituent Properties of Ceramic Matrix Composites: II, Experimental Studies on Unidirectional Materials.
Journal of the American Ceramic Society 78, 2721–2731 (1995).
47. Kerans, R. J., Parthasarathy, T. A., Rebillat, F. & Lamon, J. Interface properties in high-strength nicalon/C/SiC
composites, as determined by rough surface analysis of fiber push-out tests. J. Am. Ceram. Soc. 81, 1881–
1887 (1998).
48. Lamon, J. & Bansal, N. Ceramic Matrix Composites: Materials, Modeling and Technology. (2015).
49. Buet, E. et al. Influence of surface fibre properties and textural organization of a pyrocarbon interphase on the
interfacial shear stress of SiC / SiC minicomposites reinforced with Hi-Nicalon S and Tyranno SA3 fibres. J.
Eur. Ceram. Soc. 34, 179–188 (2014).
50. Hashin, Z. Failure Criteria for Unidirectional Fiber Composites 1. J. Appl. Mech. 47, 329–334 (1980).
51. Hashin, Z. & Rotem, A. A Fatigue Failure Criterion for Fiber Reinforced Materials*. J. Compos. Mater. 7,
448–463 (1973).
52. Tabatabaei, S. A., Lomov, S. V. & Verpoest, I. Assessment of embedded element technique in meso-FE
modelling of fibre reinforced composites. Compos. Struct. 107, 436–446 (2014).
53. Lubliner, J., Oliver, J., Oller, S. & Oñate, E. A plastic-damage model for concrete. Int. J. Solids Struct. 25,
299–326 (1989).
List of Figures
Figure 1 - Single tow mini-composites with ZMI SiC fibers (left) and SA3 fibers (right), fabricated by GA
that were provided for characterization. .......................................................................................................... 7 Figure 2 - Six planar SiC-SiC panels fabricated for this work (top row are baseline panels with a single PyC
layer at the fiber matrix interface; bottom row are panels with multi-layer interphase). ................................. 7 Figure 3 – Examples of single-fiber mico-composites. Initial fabrication produced a non-continuous SiC
coating (left) while subsequent fabrication produced individual fibers coated with a uniform CVD SiC later
(right). .............................................................................................................................................................. 8 Figure 4 - SEM images of the polished surface of the composite: (a) general appearance, (b) close-up of the
vicinity of fiber bundle, (c) close-up with individual fibers, denoted by arrows. ............................................ 8 Figure 5 - (Left) SEM image of an edge polish required for pillar fabrication. (Right) Height image of a SiC
fiber in the SiC matrix. It can be seen that the graphite containing regions were removed preferentially during
polishing. ......................................................................................................................................................... 9 Figure 6 - STEM image of a typical region containing fibers and surrounding matrix, arrow indicates
submicron-sized pores. .................................................................................................................................... 9 Figure 7 – (Left) Image quality map of a large area, including several fibers and a surrounding region of the
matrix, artifact CVI rings are visible. (Right) STEM image of the matrix material. ..................................... 10 Figure 8 - A) SA3 TEM foil. B) SA3 first layer PyC/fiber interface with oriented graphitic structure. C SEM-
STEM image of HNLS foil. D) HNLS first layer PyC/fiber interface with oriented graphitic structure. E)
Diffraction ring pattern at HNLS first layer PyC with characteristic graphite rings (002).8 ......................... 11 Figure 9 - EELS spectra comparison of the pyrolytic carbon interphase and secondary carbon phase of the
SA3 fiber showing graphitic structure ........................................................................................................... 11 Figure 10 - TEM images: (a) cross-section of a fiber, dashed line corresponds to the central axis of it; (b)
fiber material with the location of EDX line scan denoted, arrows indicate inclusion at the grain boundaries;
(c) radial distribution of inclusion within the fiber; (d) EDS linescane – signals of C and Si. ...................... 12 Figure 11 -(a) TEM image of the microstructure and the corresponding IPFX map. The arrow represents the
direction of grain growth. The vertical dashed line in the IPFX map represents the location of the interlayer;
(b) pole figure of the matrix (top row) and fiber (bottom row). .................................................................... 13 Figure 12 - (a) Close-up of a typical matrix microstructural, with fringes within grains visible; (b) comparison
of a TKD orientation map and a TEM image of the same area. Red rectangles denote the twin boundaries,
and it is evident that the fringes do not directly correspond to these. ............................................................ 14 Figure 13 - High-resolution TEM image of an area within the matrix. The line follows is parallel to atomic
rows, and it is evident that areas of varying contrast (fringes) are not associated with different crystal
structure. ........................................................................................................................................................ 14 Figure 14 - (a) Example of a line of indents, crossing a matrix region and one of the fibers; (b) typical depth
dependence of hardness (filled symbols) and modulus (hollow symbols) as a function of depth for the indents
placed within the matrix (solid lines) and close to the center of a fiber (dashed lines); vertical lines denote
the depth range used for averaging. ............................................................................................................... 15 Figure 15 - Hardness (filled symbols) and modulus (hollow symbols) as a function of distance from the fiber
center averaged over multiple line scans crossing the fibers; vertical dashed lines denote the typical
dimensions of a fiber. .................................................................................................................................... 16 Figure 16 - Results of high-temperature nanoindentation – temperature dependences of hardness and
modulus. ........................................................................................................................................................ 17 Figure 17 - Typical cantilever in the (a) matrix;(b) fiber; (c) interphases; standard triangular cross-section is
visible. ........................................................................................................................................................... 17 Figure 18 - Typical notched cantilever in the matrix SiC.............................................................................. 18 Figure 19 - Results of fracture tests from cantilevers in different components of a composite; two values for
the matrix are from the longitudinal (hollow symbol) and transverse (filled symbol) cantilevers. Overlay – a
stress-strain curve of a typical test. ................................................................................................................ 19 Figure 20 - Fracture toughness of different components of a composite. Two values for the matrix are from
the longitudinal (hollow symbol) and transverse (filled symbol) cantilevers. ............................................... 19 Figure 21 - Typical fracture surfaces of the notched cantilevers in different constituents: (a) interphase
(remaining base of the cantilever, visible in the image, is in the fiber); (b) fiber; (c) matrix – longitudinal
direction; (d) matrix – transverse direction. Note the geometry of the visible straight notch. ...................... 20
Figure 22 - TEM images of the fractured cantilevers: (a) at the matrix-fiber interphase; insert – EDX
elemental map of the highlighted area; (b) in the bulk fiber, arrow denote the C precipitates at grain
boundaries, X denotes the instance of intergranular crack propagation, Y the instance of transgranular crack
propagation instances of crack propagation; (c) in the bulk matrix, transverse orientation; here and in (d) the
arrows are to guide the eye along the crack; (d) located in the bulk matrix, longitudinal orientation. Cantilever
in (a) is unnotched, all the others are notched. .............................................................................................. 22 Figure 23 - SEM image of the impression left by a flat-punch tip on the fiber (indicated by an arrow),
indicating minor plastic deformation during loading. ................................................................................... 24 Figure 24 - Comparison of (a) highly porous central part of a typical fiber tow, and (b) relatively monolithic
periphery ....................................................................................................................................................... 24 Figure 25 - Results of a number of push-out tests performed at the fibers (a) in the center and (b) at the
periphery of the tows. .................................................................................................................................... 25 Figure 26 - Cross-section of the near-interphase region following a push-out, indicating the crack path. .... 26 Figure 27 - (Top) Typical micro-pillar fabrication process using FIB milling techniques, overlays show the
extraction of and interface incline and resulting interfacial area. (Bottom) Comprehensive set of micro-pillar
interface conditions that were tested. Notably there is a a wide range of PyC interface thickness and fiber
type. ............................................................................................................................................................... 28 Figure 28 - Snapshots from live mico-pillar compression testing of SA3_700nm PyC. (Left) Diamond flat
punch applying load. (Right) Post failure with fracture surface maintained. ................................................ 28 Figure 29 - (a) Representative schematic of pillar fabrication. (b) Resolved stress state and force balance of
a representative micro-pillar structure. .......................................................................................................... 29 Figure 30 - Comparison of raw data for control sample HNLS_50nm, HNLS_1300nm, and
HNLS_50nm_1dpa to show fundamental reduction in strength with irradiation and thickness. .................. 30 Figure 31 - MC criterion applied to SA3 and HNLS micro pillars with varying PyC thickness. It is observed
that for the same thickness, SA3 shows fundamentally stronger debond shear strength. .............................. 30 Figure 32 - (Left) NEUP micro-pillar property values plotted versus PyC thickness. An empirical relationship
was fitted for application in modelling efforts. (Right) Literature fiber pushout debond strength values as a
function of PyC. ............................................................................................................................................ 31 Figure 33 – (Top) example fracture micro-pillar suitable TEM foil fabrication and TEM dark field image of
foil with slipped HNLS pillar. (Bottom) HRTEM image of deformed PyC layer as a result of failure. 002
graphite-like planes are visible. And suggest fracture is occurring along those weak basal planes. ............. 32 Figure 34 - A) TEM image of HNLS interphase pre-failure. B) TEM image of PyC interface post failure
suggesting cohesive failure in the PyC layer. ................................................................................................ 32 Figure 35 - (Left) SEM fractography evaluation of fracture HNLS and SA3 micro-pillars. SEM images of
fiber surfaces are reproduced from Sauder et al6. (Right) AFM scan of SA3_700nm fracture surface, and
corresponding tabulated data. ........................................................................................................................ 33 Figure 36 - Load vs displacement curve of pillar compression showing failure load and shaded area under
the curve (work of fracture) for extraction of the fracture release rate energy. HNLS_B micro-pillar with
thick PyC interphase prior to compression. ................................................................................................... 34 Figure 37 - Energy release rate as a function of interface angle grouped in three and interface layer thickness
(data call-outs in nanometers). The mode I energy release rate was calculated using Eq.3 Cedric’s relationship
ΓII = 3.5 *ΓI .................................................................................................................................................. 34 Figure 38- - Isolation of chemical bonding contribution to fracture energy release rate. Post interface failure
of micro-pillar from HNLS_A with pillar cap still intact. ............................................................................. 35 Figure 39 - Detail of structure of a SiC composite and its interphase (a) SEM picture of the Hi-Nicalon type
S fibers from a SiC composite. The interphase is clearly present. (b) Schematic of structure of the interphase
material structure, consisting of 40nm pyrolytic carbon (PyC), followed by four repeating units of 50nm
SiC/10 nm PyC .............................................................................................................................................. 36 Figure 40 - This shows the TDTR system layout. The Ti:Sapphire produces laser pulses that are split by a
PBS (Polarizing Beam Splitter) and two-tint wavelength filters into pump (red line) and probe (purple line)
pulses. The pump beam is modulated by an EOM (electro-optic modulator), and the probe beam goes through
a mechanical chopper. The pump beam heats the sample, while the probe beam measures the reflectance of
the transducer layer (e.g., Al). ....................................................................................................................... 37 Figure 41 - TDTR data fits for the thermal conductivity of the SiC/SiC composite components and the ratio
of the in-phase and out-of-phase voltage as a function of thermal conductivity. (a) Data and TDTR model fits
for all mapped areas of the SiC/SiC composite, using a spot size of ωo = 2.9 µm, (b) Calculated ratio values
for t = 150 ps, a spot size of ωo = 2.9 µm, a heat capacity of C = 2.21 J cm-3 K-1, and varying interface thermal
conductance ................................................................................................................................................... 38 Figure 42 - Diagram and a thermal conductivity map of an interphase of the fiber. (a) A diagram showing a
cross section of the SiC/SiC composite where a fiber has been puleed out of the matrix. According to the
geometry of these regions, the average angle of the fibers to the surface has been determined to be 5 degrees,
which creates an interphase area of ~4 µm. (b) A micrograph of a pulled fiber. (c) A thermal conductivity
map of the region at the end of a fiber that has been pulled out of the matrix. Circled in the area where full
TDTR measurements were taken. Spot size (ωo) is 2.9 µm, time delay = 150 ps, and heat capacity (C) is 2.21
J cm-3 K-1 ....................................................................................................................................................... 39 Figure 43 - Summary of thermal conductivity measurements of fibers and matrix as a function of temperature.
(a) A 50µm x 50 µm thermal conductivity map, taken at 1 µm steps, at room temperature (lighter region
indicates higher thermal conductivity), (b) A cross section through the map in figure5a (dashed white line)
at room temperature, (c) A cross section of the same region at different temperatures, (d) Summary of thermal
conductivity dependencies on temperatures for matrix and fiber, compared with ........................................ 40 Figure 44 - Radioactive materials storage cabinet. Source drawers are complete with lead shielding and
locking mechanisms ...................................................................................................................................... 41 Figure 45 – (Top) Plot of the hardness and (bottom) Elastic Modulus as a function of constituent, irradiation,
and temperature. ............................................................................................................................................ 43 Figure 46 - Mohr-Coulomb criterion applied to the unirradiated and irradiated interfaces. A fundamental
decrease in cohesive shear strength between unirradiated and irradiated interfaces was observed. .............. 44 Figure 47 –Before and after irradiation SEM images comparing the PyC interface structure. (Top)
HNLS_50nm control vs 1dpa at 350°C. (Middle) HNLS_180nm control vs ~12dpa at 280°. (Bottom)
SA3_140nm control vs ~4.5dpa at 630°. ....................................................................................................... 45 Figure 48 - Comparison of typical linescans of hardness and modulus across matrix and fiber for 2.6dpa ion
irradiated and reference unirradiated samples. .............................................................................................. 46 Figure 49 - Typical comparison of depth dependence of hardness and modulus in the matrix for samples
irradiated to different damage levels. ............................................................................................................ 47 Figure 50 - Comparison of crack patterns around the indents in the matrix of unirradiated an irradiated
samples. ......................................................................................................................................................... 47 Figure 51 - Comparison of fracture toughness as measured in different constituents, for unirradiated and
irradiated samples. ......................................................................................................................................... 48 Figure 52 - Thermal conductivity mapping and profile of a 20x20 μm He implanted region of a 6H SiC wafer.
The fluence of the implanted region is 0.016 nC-μm-2. (a) Photo of the He implanted region of the SiC wafer.
(b) A 20x20 μm thermal conductivity map of the He implanted region, taken at 1 μm steps. Time delay was
set to 𝑡𝑡 = 150 ps, heat capacity was C = 2.21 J cm-3 K-1, and the interfacial Al thermal conductance for this
region was 80 MW m-2 K-1. (c) Thermal conductivity profile of the implanted region. (d) Plotted trend of
thermal conductivity across fluences of 0.016 nC-μm-2, 0.08 nC-μm-2, and 0.16 nC-μm-2. .......................... 50
Figure 53 – (left) XCT scan of a planar specimen with identified porosity map. (Right) XCT image showing
internal porosity of tubular specimen. ........................................................................................................... 51 Figure 54 – (Left)Schematic illustrations showing a specimen mounted between upper and lower grips.
(Right) External image of chamber with halogen lamps at 700°C ................................................................ 52 Figure 55 - (Top) Sequential stage of test set up; high temperature cement casting, in situ high temperature
loading, and post fracture evaluation. Note that a thermo couple was attached to this specimen to
appropriately track the thermal state. (Bottom) Typical load vs time plot following 5um incremental lading
steps. .............................................................................................................................................................. 52 Figure 56 - (top) Tomographic reconstruction showing the cross-sectional geometry of a single tow tensile
test sample. (bottom left) Normal x-ray projection of ambient temperature test. Numbers denote visible
micro-cracks. (Bottom right) Normal x-ray projection of 700°C test, numbers denote visible micro-cracks.
....................................................................................................................................................................... 53 Figure 57 – (a) Typical Flexural Stress vs Extension plot (b) Fracture surface failed composite. Limited fiber
pullout is observed. (c) Macroscopic view of fractured test specimen. ......................................................... 58 Figure 58 – (Left) The dog bone specimen shape. (Center) Image of an axial tension test setup. (Right) Stress-
strain curve for a dog bone shaped specimen from a planar tension test ....................................................... 58 Figure 59 -. (Right) Example of digital image correlation strain map during test. (Right) Typical fracture
surface observed after mechanical testing of planar SiC/Sic samples, showing limited fiber pull-out.) ....... 59 Figure 60 - A representative stress-strain curve from an elastomeric expanding plug test. .......................... 59
Figure 61 - Load versus crosshead extension for a typical RT and High Temperature c-ring tests .............. 60 Figure 62 – The leftmost image shows a finite element model using cohesive surfaces after decohesion has
occurred. Results from this model are shown as the solid green line on the plot at right. The center picture
shows a model with Mohr-Coulomb plasticity in a thin layer (which appears in red) just after the layer fails.
Results from this type of model are shown as circular red dots in. The remaining data points in the plot are
experimental results. ...................................................................................................................................... 61 Figure 63 - Fiber pushout model (left) and comparison with experiments (right). The 2D axisymmetric model
was revolved 180° for visualization purposes. Model predictions are plotted at right in the orange dashed
line. Note that the load at which the PyC layer fractures and the load begins to drop is determined in part by
residual stress between the fiber and matrix, which is unknown. .................................................................. 62 Figure 64 – Raw tomography projection scan of partially broken mini-composite with pre-machine notch on
right. Note the array of cracks through the composite matrix. ...................................................................... 62 Figure 65 - Simple two-dimensional single fiber model (left) and three-dimensional multi fiber model used
to explore the feasibility of a larger-scale mini-composite (right). These models used predefined cracks with
cohesive contact as well as cohesive contact between the matrix and fibers. While results from the two-
dimensional model were encouraging, convergence in the three-dimensional model was very poor and even
the simple model shown at right was computational expensive. ................................................................... 63 Figure 66 - Two-dimensional homogenized model of fiber "tow" with Hashin damage model. The red
elements show the damaged material and the predicted crack path .............................................................. 64 Figure 67 - Results from Embedded Element Technique with Concrete Damaged Plasticity material models.
This technique includes separate finite elements for the matrix and fibers. The matrix elements are shown
with contours for material degradation (a) and equivalent plastic strain (b). These figures show the
progression of failure in the matrix. Fiber elements are shown in (c). Here, contours have the same scale as
(a) and show degradation of the fiber material. Note that this model predicts that fracture propagates across
the specimen and does not capture the radial cracking pattern shown in testing. .......................................... 65
