Sie sind auf Seite 1von 123

Copyright by Raushan Kumar 2010

The Thesis committee for Raushan Kumar certifies that this is the approved version of the following thesis

Effect of Chemical Environments on Subcritical Crack Growth in Geological Materials

Approved by Supervising Committee:

Jon E. Olson, Supervisor Jon T. Holder

Effect of Chemical Environments on Subcritical Crack Growth in Geological Materials

by Raushan Kumar, B. Tech

Thesis
Presented to the Faculty of the Graduate School of The University of Texas at Austin in Partial Fulfillment of the Requirements for the Degree of

Master of Science in Engineering

The University of Texas at Austin May 2010

Dedication

To my parents: for their unconditional love. To my brother, Ravi and sister, Rashmi: for their presence and support.

Acknowledgements

I would like to extend my sincerest gratitude to Dr. J.E. Olson for his invaluable support, patience and guidance throughout my time at The University of Texas at Austin. He had been a wonderful supervisor and provided intellectual stimuli for this work. I would also like to offer sincere thanks to Dr. J. Holder for the assistance and inputs during experiments. He was very instrumental in solving issues related to hardware electronics and mechanical. I am also thankful to Glen Baum and Gary Miscoe for their input on safety practices. And thanks to Autumn Kaylor for helping me with point counts of my samples. I would like to thank the sponsors of my work - the member companies of FRAC (Fracture Research and Application Consortium) as well as a grant from ExxonMobil Research and Engineering Company in New Jersey. Thanks to Gareth Block of

ExxonMobil for taking interest in my fracture mechanics research. Special thanks to Libsen Castillo and Vinay Sahni for being wonderful friends during my stay at Austin. Lastly, I want to thank my father Prof. Suman Kumar, mother Smt. Bibha Sinha, brothers Ravi and Subodh and sister Rashmi for their support.

Abstract

Effect of Chemical Environments on Subcritical Crack Growth in Geological Materials

Raushan Kumar, MSE The University of Texas at Austin, 2010

Supervisor: Jon E. Olson

Subcritical crack propagation experiments were performed using the double torsion test methodology on sandstone and granite samples. The control environments consisted of air, water, aqueous cationic surfactant solutions and aqueous caustic solutions of high pH. The expected reduction in fracture toughness and subcritical index going from air to water environments was observed, a result of water being a more reactive environment. The reduction in subcritical index correlated strongly with the clay content of the sandstones, but no other mineralogic parameters seemed to affect fracture results significantly. Caustic solutions reduced strength and subcritical index more than did water, but surfactant solutions performed approximately the same as water, even though they were expected to be more reactive. Activation volume values were estimated from velocity versus stress intensity factor curves using published values of crack-tip vi

radius of curvature, giving results on the order of 15.0 x 10-7 m3/mol for sandstone tested in air, about 5 times higher than previously reported values for quartz. The values dropped about 50% when measured in aqueous environments. A systematic study of variability in subcritical index values in air showed that subcritical index correlated inversely with the load drop observed during crack propagation. Subcritical indices decreased as load drop increased, with stable values usually resulting at load drops of 0.5 lbs or more for the 1.5 mm thick test samples. Tests performed in more reactive environments (i.e., aqueous solutions) typically resulted in higher load drops and less variable subcritical indices. High rate data acquisition techniques were implemented to try to capture crack propagation results beyond the stage I power-law velocity versus stress intensity factor behavior. The stage II constant velocity regime was observed, but only in one sample tested in air. Fracture roughness of experimentally created crack paths was also analyzed using fractal analysis. The degree of crack path tortuosity (actual rough path length divided by straight-line length, a measure of roughness) was clearly dependent on the scale at which it was measured. Using the modified divider technique, crack-path tortuosity exhibited a clear fractal nature, with results for sandstone, shale and limestone exhibiting fractal dimensions from D1.04 to D1.09. The fractal dimension indicated a negative

correlation with subcritical index, but the results showed a significant scatter.

vii

Table of Contents
List of Tables ...........................................................................................................x List of Figures ........................................................................................................ xi Chapter 1: Introduction ............................................................................................1 1.1 Objective ................................................................................................1

Chapter 2: Mechanical and Thermodynamical Basis ..............................................3 2.1 Rock Fracture Mechanics ......................................................................3 2.1.1 Critical Crack Growth...................................................................3 2.1.2 Modes of Crack Surface Displacement.........................................4 2.1.3 Subcritical Crack growth ..............................................................5 2.2 2.3 Thermodynamical Basis.........................................................................7 Effect of Chemically-active Environment ...........................................13 2.3.1 Rebinder Effect ...........................................................................14 2.3.2 Westwood Approach...................................................................15 2.3.3 Stress Corrosion ..........................................................................18 2.3.3.1 2.3.3.2 2.3.3.3 2.3.4.1 2.3.4.2 2.3.4.4 Chemical Reaction Rate Theory ..................................19 The Lawn-Cook Model................................................22 Power Law vs. Exponential Law .................................25 Diffusion ......................................................................29 Dissolution ...................................................................29 Microplasticity .............................................................30

2.3.4 Other Mechanisms ......................................................................29

2.3.4.3 Ion Exchange ...............................................................30

Chapter 3: Measurement of Subcritical Crack Index.............................................31 3.1 Test Description ...................................................................................31 3.1.1 Mathematical Description of Stress Intensity Factor for Double Torsion Test ................................................................................33 3.1.1.1 Expression for Crack Velocity for Constant Displacement Method ...............................................................................36 viii

3.1.1.2 Data Analysis and Reduction Method .........................37 3.2 Sample Preparation ..............................................................................41

Chapter 4: Experimental Results and Analysis......................................................44 4.1 4.2 Petrographic Information .....................................................................44 Results Subcritical indices and Fracture Toughness.........................49 4.2.1 Dependence of Subcritical Indices on Loading Magnitude........52 4.2.2 Effect of Rock-types and Environmental Control ......................61 Chapter 5: Velocity vs. Stress-intensity Curves ....................................................68 5.1 Data Noise Reduction ..........................................................................68 5.1.1 Spline Average of Recorded Data-points ...................................68 5.1.2 Numerical Differentiation Technique .........................................69 5.2 5.3 Region-II Crack-growth.......................................................................70 Activation Volume Calculation ...........................................................73

Chapter 6: Fracture Roughness Analysis...............................................................76 6.1 6.2 6.3 6.4 Background ..........................................................................................76 Fractal and Fractal-dimension..............................................................76 Fractal Applied to Fractures.................................................................78 Measuring Fracture Surface Roughness ..............................................79 6.4.1 Crack Imaging.............................................................................80 6.4.2 Crack-path Digitization...............................................................81 6.4.3 Data Analysis ..............................................................................83 6.5 Fractal Analysis ...................................................................................85 6.5.1 Application of Modified Divider Technique ..............................85 6.5.2 Fractal Analysis Random Path.................................................88 6.5.3 Fractal Analysis Subcritical Test Samples...............................90 Chapter 7: Conclusions ..........................................................................................96 Bibliography ..........................................................................................................99 Vita .....................................................................................................................108

ix

List of Tables
Table 2.1: Expressions for the crack velocity from literature................................27 Table 2.2: Expressions for time to failure from literature. ....................................28 Table 4.1(a): Point-count data. Numbers in each column denote raw count out of the total number of counts in the right-most column..............................46 Table 4.1(b): Summary of point-count data reported percentages. .......................47 Table 4.2: A summary of the dual torsion tests performed for all samples. n is the subcritical index, KIMAX is the stress-intensity factor corresponding to peak load, and P is the load drop observed after 300 seconds. ......50 Table 4.2: Contd.....................................................................................................51 Table 4.3(a): Mean and standard deviation of subcritical indices data for different samples..............................................................................................64 Table 4.3(b): Youngs modulus and fracture toughness of different specimens in different control environments..........................................................65 Table 5.1: Activation volume and subcritical index for three different sandstone samples..............................................................................................75 Table 6.1: Comparisons of mean, standard deviation and skewness parameters of an actual crack-path and a randomly generated crack-path with segment lengths in a Gaussian distribution. ....................................................88 Table 6.2: A summary of fractal analysis results and sub-critical indices for given samples..............................................................................................94

List of Figures
Figure 2.1: Three basic modes of loading for a crack: a) Mode I, b) Mode II, and c) Mode III (adapted from Whittaker et al., 1992)..................................5 Figure 2.2: Schematic crack-velocity/stress-intensity diagram. K0 is the stress corrosion limit and KIC is the fracture toughness; experiments yet to confirm K0 in rocks and minerals (Atkinson, 1984). ..........................6 Figure 2.3: a) Griffith crack system b) energetic of Griffith crack in uniform tension, plane stress (adapted from Lawn, 1993). ............................................8 Figure 2.4: Irwin-Orowan extension of the Griffith concept: small-scale zone model ...........................................................................................................10 Figure 2.5: Thermodynamically admissible kinetic relations: a) thermally activated crack growth in vacuum; lattice trapping, b) environmentally assisted crack growth, c) three typical stages, d) environment is surface active but unable to reach separating crack tip bonds, e) strongly surface active environment makes negative but unable to reach tip (Rice, 1978)...12 Figure 2.6: Interaction between water molecules and strained crack-tip bond in glass: (A) adsorption, (B) reaction and (C) separation (Michalske and Freiman, 1984). ................................................................................................19 Figure 2.7: Schematic of potential rate-limiting phenomena (Lawn, 1993)..........23 Figure 3.1: Loading configuration of the Double Torsion test (Nara and Kaneko, 2005). ................................................................................................32 Figure 3.2: Schematic diagram of a point loaded rectangular torsion bar (from Williams and Evans, 1973). ..............................................................34

xi

Figure 3.3: Load vs. relative time from Double Torsion test; blue triangles represent actual data acquired and red diamond represents the power-law fit to calculate the subcritical index...........................................................40 Figure 3.4: Load vs. relative time from Double Torsion test. Sharp drop in load (dashed blue box) denotes critical failure. Maximum load is used to calculate fracture toughness of the specimen....................................41 Figure 3.5: Schematic representation of the compliance relationship for DT test method (adapted from Fuller, 1979). ................................................42 Figure 3.6: Evolution of the crack front with crack extension (Chevalier et al., 1996). ...........................................................................................................43 Figure 4.1: Q-F-L diagram for the point-count data for different specimens........48 Figure 4.2: Load-decay data of sample 5950_1A (Test a) tested in air shows bumps during initial time..............................................................................56 Figure 4.3: Subcritical index vs. load-drop for Sandstone A1, A2 and A3 combined (A) air, (B) water, and (C) caustic. ...................................................57 Figure 4.4: Subcritical index vs. load-drop for Canyon I, II and III combined (A) air, (B) water, and (C) caustic. ................................................................57 Figure 4.5: Subcritical index vs. load-drop for MesaVerde (A) air, (B) water, and (C) caustic. ..............................................................................................58 Figure 4.6: Subcritical index vs. load-drop for Granite (A) air, (B) water. ...........58 Figure 4.7: Load-drop vs. maximum-KI for Sandstone A1, A2 and A3 combined (A) air, (B) water, and (C) caustic...........................................................59 Figure 4.8: Load-drop vs. maximum-KI for Canyon I, II and III combined (A) air, (B) water, and (C) caustic. ......................................................................59

xii

Figure 4.9: Load-drop vs. maximum-KI for MesaVerde (A) air, (B) water, and (C) caustic. ..............................................................................................60 Figure 4.10: Load-drop vs. maximum-KI for Granite (A) air, and (B) water........60 Figure 4.11: Crack-velocity vs. load for sandstone sample. Leftward shift of the curve indicates lower energy requirement for crack-growth in water and surfactant...........................................................................................66 Figure 4.12: Percentage change in subcritical index from air to water, surfactant and caustic vs. clay-content. ....................................................................67 Figure 5.1: Comparison between actual experimental data-points and spline data-fit. Blue diamonds represents the spline-smoothed curve and the red squares are the actual data..............................................................................69 Figure 5.2: (a) (a) Load-decay curve, and (b) velocity stress-intensity factor curve for specimen MV3 (Test b). The rollover of velocity at higher KI (earlier in the load decay) represents Stage II behavior. Start of RegionIII can also be observed (for first few load-decay data). ..................72 Figure 5.3: v-K curve for the specimen Sandstone-A1_3A (Test a) - test conducted in caustic (a) power-law fit, and (b) exponential fit. ............................74 Figure 5.4: Activation volume vs. subcritical index for three sandstone types. ....75 Figure 6.1: Illustration of self-similar and self-affine fractals (Kulatilake, 1995).77 Figure 6.2: Patterns of a Microcrack coalescence process. (a) just before fracture, and (b) a newly nucleated microcrack triggers the catastrophic failure (adapted from Lu et al., 1994.) .........................................................78 Figure 6.3: Environmental Scanning Electron Microscope (ESEM) facility at JSG, UT Austin..........................................................................................80

xiii

Figure 6.4: A screenshot of the profile generated from the crack-path in a sample from Canyon Sand tested in air, (a) BSE image, and (b) discrete digitized image..................................................................................82 Figure 6.5: Effect of scale-length on estimation of length with the original divider method...............................................................................................84 Figure 6.6: A screenshot of Excel VBA Tool developed for fractal analysis........84 Figure 6.7: Variation of log (Normalized length) with log (scale) for crack-roughness data of Canyon samples from 6071 ft. depth for with the magnification factor of (a) 100, and (b) 1000; Sub-critical test carried out in an aqueous environment of pH 13. ........................................................86 Figure 6.8: Variation of fractal dimension with magnification factor for crack-path roughness data (Canyon samples from 6071 ft. depth).....................87 Figure 6.9: Variation of fractal dimension with magnification factor for fracture surface roughness data (adapted from Kulatilake, 2006)..................87 Figure 6.10: A comparison of fractal analysis from the actual crack-path and randomly generated crack-path; red triangles are from actual crack-path follow power law systematically; blue diamonds are from randomly generated crack-path continuously curving. ..................................89 Figure 6.11: BSE images of crack-path in Canyon Sandstone and Barnett Shale, both tested in air, at magnification 1000x.................................................91 Figure 6.12: Variation of log (Normalized length) with log (scale) for crackroughness data of Canyon samples from 6071 ft. depth with the magnification factor of 100x in (a) air, and (b) caustic with pH 13. 92 Figure 6.13: A comparison of fractal analysis for Canyon sand (D = 1.092) and Barnett shale (D = 1.045)..................................................................93 xiv

Figure 6.14: Fractal analysis of a Yates sample; fractal dimension is 1.0849, which is in the same range as Canyon sandstone; visual impression also gives the same order of roughness for the samples..........................................93 Figure 6.15: (a) fractal dimension vs. subcritical index, (b) y-intercept vs. subcritical index..................................................................................................95

xv

Chapter 1: Introduction
1.1 OBJECTIVE Understanding the mechanics of rock fracture has been the continued interest of engineers and scientists. In the oil and gas industry, fracture mechanics has conventionally been applied to: 1) characterize naturally fractured reservoirs for better modeling and prediction of flow-behavior, 2) stimulate oil and gas wells by hydraulic fracturing, and 3) resolve issues related to wellbore integrity and stability. Applications are being extended to fragmentation of oil shale beds for in-situ retorting, and stability of thermal-stress induced fracturing of geothermal wells. Understanding the mechanics of rock-fracture has helped in formulating laws to explain how stresses remotely applied at the outer boundaries of a specimen transmit to the crack tip (Tada, Paris and Irwin, 2000), and in defining criteria for extension of cracks in terms of some parameter which characterizes the intensity of locally concentrated stress fields at the crack-tip (Irwin and Wells, 1965). Understanding the mechanism needs investigation on a micro-scale to understand the processes involved in fracture at the crack-tip and the process-zone. Lawn (1983) stated: For those who concern themselves primarily with the question of when fracture occurs, as engineers do, the methodology of fracture mechanics appears to be totally adequate as a predictive tool. However, if we ask ourselves why fracture occurs, things start to go wrong. Therefore, understanding the mechanism is vital to resolve the issues related to engineering application of fracturing in a more effective way. In oil and gas engineering, it can further be applied to explain the observations and to resolve the issues related to increased rate of drilling in the presence of surface-active media, increased injectivity during matrix-acidization with acid-surfactant mixture, and possible loss of 1

ground stability after secondary oil recovery both during injection of water and surfactant-water system. During critical (irreversible catastrophic) failure, crack propagate unstably with a velocity of the order of the shear wave velocity (103 ms-1). The most important parameter based on the premise of linear elastic fracture mechanics to characterize this type of rapid fracture growth is fracture toughness or critical stress intensity factor. Subcritical, slow or stable crack growth at velocities less than 10-1 ms-1 can occur at stress intensities far below the critical stress intensity factor. The presence of aggressive chemical environment and elevated temperature, under sustained or residual loading condition, further aggravates the kinetics of subcritical crack growth. The purpose of this study is to document the mechanism of crack-growth (critical and subcritical) in the presence of various chemical environments and to subsequently characterize the observations in terms of fracture mechanics parameters. An attempt has been made to couple the experimental observations with pertinent theoretical and empirical models available in the area of material science of metals, glass, ceramics and rocks to understand the processes involved at the crack-tip. An attempt has also been made to relate sub-critical crack growth index to fractal dimension of the crack-path roughness.

Chapter 2: Mechanical and Thermodynamical Basis


2.1 2.1.1 ROCK FRACTURE MECHANICS Critical Crack Growth Linear Elastic Fracture Mechanics (LEFM) has been applied to provide insights into the fracture of rocks and minerals (Atkinson, 1984; Meredith and Atkinson, 1985; Swanson, 1984). Two alternative approaches to fracture analysis have been used: a) the energy criterion, and b) the stress-intensity factor approach. The energy criterion (Griffith criterion) states that crack-extension or fracture occurs when the energy available for crack-growth is sufficient to overcome the resistance of the material. At fracture of a linear elastic material, the crack driving force, G equals a critical value Gc, which is a function of the fracture toughness. Gc = f ( E , , K IC ) (2.1)

The stress intensity factor KI for a sharp crack at in an infinite body is given by:
KI = I c

(2.2)

where c is the fracture half-length and I is the applied remote stress. Near crack-tip stresses, that cause crack growth, are directly proportional to stress-intensity factor and are given by a generalized equation:

i =

KI f i ( ) 2 r

(2.3)

where KI depend on the outer boundary conditions (i.e., on the applied loading and specimen geometry). The remaining factors depend on the spatial coordinates about the

tip, and determine the distribution of the field: the coordinates factors consist of a radial component (r-1/2) and an angular component (f()). Thus, KI characterizes the crack-tip conditions, and therefore the fracture of the material must occur at a critical stress intensity factor, KIC, also called the fracture toughness of the material. Irwin (1957) showed that the energy approach is equivalent to the strength approach and the stress-intensity factor KIC is related to Gc by: Gc = where E '
2 K IC , E'

(2.4) in plane-stress, and in plane strain,

=E = E / (1-2)

= Youngs modulus of the material, = Poissons ratio

2.1.2

Modes of Crack Surface Displacement Different loading configurations lead to different modes of crack surface

displacement. Generally, three different modes are defined (Figure 2.1): 1) Mode I, the tensile opening mode corresponds to normal separation of the crack walls where displacements of crack surfaces are perpendicular to the crack-plane 2) Mode II, the sliding (in-plane shear) mode corresponds to longitudinal shearing of the crack walls where displacements of the crack surfaces are in the crack plane and perpendicular to the crack front and 3) Mode III, the tearing (out-of-plane) mode corresponds to lateral shearing of the crack walls where displacements of the crack surfaces are in the crack planes and perpendicular to the crack-front. Throughout this thesis, only Mode I fracture has been taken into consideration while studying different fracture parameters. 4

(a)

(b)

(c)

Figure 2.1: Three basic modes of loading for a crack: a) Mode I, b) Mode II, and c) Mode III (adapted from Whittaker et al., 1992). 2.1.3 Subcritical Crack growth Subcritical crack growth was first observed in glass by Grenet (1899). Gurney (1947) and Gurney and Pearson (1949) studied the effect of environment on delayed failure of glass and used thermodynamic concepts to explain moisture enhanced crack growth. It has been subsequently investigated using glass (Charles, 1958a, b; Wiederhorn, 1967; Kies and Clark, 1969; Wiederhorn and Bolz, 1970; Bhatnagar et al., 2000), ceramics (Evans, 1974; Wu et al., 1978; Wiederhorn et al. 1980), metals (Williams and Evans, 1973; Pabst and Weick, 1981), cement (Beaudoin, 1985a, 1985b, 1987), and rocks and minerals (Henry et al., 1977; Atkinson, 1979, 1980, 1981, 1984; Atkinson and Meredith, 1981; Meredith and Atkinson, 1983; Holder et al., 2001; Rijken, 2005, Adefashe, 2006). Stress-corrosion has been considered the main mechanism for subcritical crackgrowth in brittle rocks. Crack-velocity (v) stress intensity factor (KI) diagram (Wiederhorn, 1967) has been used to study the subcritical crack growth and the effect of 5

different control environments on it. Wiederhorn (1967) identified trimodal behavior of log v - KI plot (Figure 2.2) and classified it into three distinct regions of crack-growth behavior: 1) Region I, characterized by a strong dependence of crack-velocity on stress intensity factor and considered to be controlled by the rate of stress-dependent chemical reaction, 2) Region II, where the crack-velocity is almost independent of the stress intensity factor and is limited by transport of reactive species to the crack-tip, and 3) Region III, where crack-growth is controlled mainly by mechanical rupture and is relatively insensitive to the chemical environment.

Corrosive + mechanical cracking Diffusion controlled cracking

Log10 crack-velocity

KO KC Reaction controlled cracking

stress- intensity factor

Figure 2.2: Schematic crack-velocity/stress-intensity diagram. K0 is the stress corrosion limit and KIC is the fracture toughness; experiments yet to confirm K0 in rocks and minerals (Atkinson, 1984). 6

These diagrams have been used to: a) predict time-dependent fracture properties (Evans, 1972; Wiederhorn, 1974) and b) understand the micro-mechanism of subcritical crack growth under different environments (Wiederhorn, 1974; Atkinson, 1979a). 2.2 THERMODYNAMICAL BASIS Contemporary theories of the strength of brittle materials stem from the Griffith energy balance description of fracture processes (1920). Griffiths idea to model a crack as a reversible thermodynamic system (Griffith, 1920) coupled with his energy-balance concept (pertaining to crack propagation) and flaw hypothesis (pertaining to crack initiation) laid a strong foundation for a general theory of fracture. Griffith considered a static crack as a thermodynamical system and defined a criterion for crack propagation so as to minimize the total free energy of the system based on the first law of thermodynamics (law of energy conservation). The system energy (U) during crack propagation can be considered as a sum of mechanical-energy (UM) and surface-energy (US) terms. Mechanical energy consists of strain potential energy (UE) in the elastic media and the potential energy (UA) of the outer applied loading systems. Surface energy (US) is the free energy consumed in creating new surfaces by overcoming cohesive forces of molecular attraction. These can be mathematically expressed as U = U M + US , U = (U E + U A ) + US , Differentiating Equation 2.5 with crack length, we get (2.5) (2.6)

dUS dU dU M = + = G + 2 . dc dc dc where G is the mechanical-energy release-rate,


7

(2.7)

= free surface-energy per unit area. Mechanical energy terms generally decrease during crack propagation (dU M / dc < 0) and thus favor it. The surface energy term shall generally increase during crack growth and thus oppose it. Griffiths concept assumes crack-growth as a reversible process and proposes equilibrium condition as a criterion for predicting fracture behavior:
dU / dc = 0, Gc = 2 . (2.8) (2.9)

The crack should propagate or not according to whether the (dU/dc) term is negative or positive, respectively. Equation (2.4) provides an equivalence between mechanical-energy release rate and critical stress-intensity factor, KIC. Equation (2.4) and Equation (2.9) combined can be used to calculate free-surface energy or fracture-energy, as a function of KIC

US = 2 c
Equilibrium

Energy

c0

Crack length, c

UM = -

2
2E

c2

U=U M + US

Figure 2.3: a) Griffith crack system b) energetic of Griffith crack in uniform tension, plane stress (adapted from Lawn, 1993).

2 K IC E'

(2.10)

Simpson (1973) proposed that in a porous material, the stress intensity in the solid will be elevated because of the reduction in load-bearing area. KIC should be divided by (1 ) to take porosity into account. Thus, the fracture energy normalized for porosity should be

N =

(1 2 )

(2.11)

where N is normalized fracture energy. By invoking linear elasticity theory and using Inglis solution of the stress and strain fields, Griffith showed that the critical failure for an elastic plate with crack of length 2c subjected at infinity to the action of uniformly distributed stress tension, 1, occurs at

I = (2 E ' / c)1/ 2
where I is remotely applied stress

(2.12)

Irwin (1957) and Orowan (1949) independently proposed to add the work of plastic deformation, p to the specific surface energy related to a unit area of the newly formed surfaces, in the Equation (2.3). Plastic deformation zones are small zones in front of the crack. Poncelet (1965) challenged the basic Griffith Criteria (1920) assumption of isothermal behavior of freshly cleaved surface. From the statistical thermodynamic viewpoint, Poncelet (1965) argued that the surface energy of bodies is a free energy and not a potential energy as stated by Griffith (1920). The breaking of bonds which are not reformed, supplies the heat of sublimation of the solid to the surface and kinetic energy

terms. But still it didnt take the thermodynamics of slow-crack growth into consideration.
Zone boundary

crack

dc

Figure 2.4: Irwin-Orowan extension of the Griffith concept: small-scale zone model Two major works to advance the understanding on thermodynamical aspects of crack propagation came from Pollet and Burns (1977) and Rice (1978). Pollet and Burns (1977) expressed the crack extension force, neglecting thermodynamic surface energy of the material, as sum of two components: an athermal and a thermal component. Pollet and Burns (1977) proposed that the athermal component corresponds to the value of the applied crack extension force below which crack propagation does not occur and is required to overcome long-range, high-energy obstacles that cannot be activated by thermal fluctuation, whereas the thermal component is used to assist thermal activation of short-range barriers. The value of athermal component should correspond to the value of crack extension force at which the crack velocity is zero. Rice (1978) included the concept of irreversible thermodynamics in the formalism of crack-propagation and proposed an extended Griffith criteria. Basing his arguments on the second law of thermodynamics, i.e. non-negative entropy production, Rice extended Griffith criterion to be

10

(G 2 ) c 0 where G = elastic energy release rate, 2 = reversible work to separate the fracture surfaces,
.

(2.13)

= crack speed.

It was shown that the reversible work of separation (in vacuum) must lie in the range between the critical levels of G for rapid growth (G+) and rapid healing (G-), i.e.
G 2 G + .

(2.14)

Thus, it paved the way for a thermodynamically unstable condition, which described the crack propagation as stable within two extremes of applied stress-condition. It was further shown that for crack growth in a reactive environment, which can adsorb on the newly-created fracture surfaces, inequality remains valid when is interpreted as a surface energy as altered appropriately to account for adsorption. Thus, Rices extension of the Griffith criterion (Figure 2.5) helped to explain lattice trapping of atomistic model (Thomson, Hsieh and Rana, 1971) and stress corrosive effects (Charles and Hillig, 1962) within the theoretical formalism of thermodynamics. Rice criterion provided a fundamental rationale for the construction of v-G or v-K diagrams.

11

Figure 2.5: Thermodynamically admissible kinetic relations: a) thermally activated crack growth in vacuum; lattice trapping, b) environmentally assisted crack growth, c) three typical stages, d) environment is surface active but unable to reach separating crack tip bonds, e) strongly surface active environment makes negative but unable to reach tip (Rice, 1978).

12

2.3

EFFECT OF CHEMICALLY-ACTIVE ENVIRONMENT

Understanding the effect of chemically-active environment on mechanical properties of rocks is paramount to application in oil and gas industry. Operations on subterranean reservoir rocks include use of acidic media (matrix acidization), basic media (Alkaline-Surfactant-Polymer Flooding) and surfactants (ASP Flooding, Hydraulic fracturing and Drilling). The effect of chemistry on polycrystalline-polyphase rocks is difficult to analyze because of the complexity imposed due to the presence of microstructures, cements, fabrics etc. There are various reversible and irreversible mechanisms at work during fast and slow crack growth and it is imperative to understand the role of each of them, before extrapolating the experimental results to macro-scale application. The interdisciplinary nature of this field of study warrants integrating the complicated interplay of variables arising out of chemistry of the rock constituents and environment, physics of the surface of the rock and mechanics of rock-engineering. Different phenomena contributing to critical and subcritical crack growth have been identified and various models have been proposed to incorporate these mechanisms at work. Rebinder (1928; cited in Rebinder and Shchukin, 1973) and Orowan (1944) were among the earliest to identify the role of chemistry in brittle fracture. Rebinder (1928) proposed reduced strength and plastic flow during deformation and failure of solids under a definite stress-state in presence of surface-active media. The reversible physicochemical action of the media lowers the specific free surface energy of the solid and therefore, the work of formation of new surfaces during deformation and failure processes. Orowan (1944) observed a reduction by a factor of about three in the strength of glass specimens in moist air relative to that in a vacuum under sustained loading. He proposed that environmental molecules enter the crack and get adsorbed onto the walls in 13

the adhesion zone, lowering the surface energy of the solid. Another approach to explain reversible effect of the presence of surface-active agent on mechanical strength of materials came from Westwood (1974). He proposed that the environments which tend to bring zeta-potential to zero have the highest effect on lowering the mechanical strength of materials. In particular, Mills and Westwood (1980) cited that the addition of 10-3 to 10-4 moles l-1 of cationic surfactant DTAB (dodecyl trimethyl ammonium bromide) to cutting fluids increases drilling rate in quartz and westerly granite.
2.3.1 Rebinder Effect

The reversibility of physico-chemical effect and obligatory participation of mechanical stress distinguishes Rebinder Effect from other chemical or electrochemical processes such as stress-corrosion or dissolution of the solids in the surrounding media, which are typically irreversible. It incorporates the idea of thermodynamically stable interfaces between the given solid phase and the medium and therefore partial cancellation of the intermolecular forces on the newly produced surfaces. This effect is manifested both due to an adsorption monolayer and a liquid-phase layer, which can lead to still stronger changes in the mechanical properties corresponding to very low values of the inter-phase energy. Rebinder Effect proposes reduction in materials strength due to combined action of a reduction in surface energy of the material, a reduction in the bonding forces (causing healing of the fracture) across the developing crack or fracture and a chemomechanical wedging pressure exerted by the surfactant on the flanks of existing cracks in the form of osmotic pressure. Parameters affecting these actions can be expressed in terms of contact angle , wetting energy w, wedging pressure Pw and capillary pressure Pc in order of importance. The contact angle is defined as, 14

cos =

SV SL , LV

(2.15)

where SV = Interfacial tension between solid-vapor interfaces,

SL = Interfacial tension between solid-liquid interfaces, LV = Interfacial tension between liquid-vapor interfaces.
Contact angle is a measure of wetting (ease of spread of liquid over solid surface). Complete wetting occurs at a contact angle of zero; non-wetting means that the angle is greater than zero and complete non-wetting occurs at a contact angle of 1800. Wetting energy is defined as the difference between the solid-vapor and solid-liquid surfacetensions. Wetting energy is also a measure of wetting and the ability of liquid to penetrate irregularities. Wedging pressure is related to the liquid-vapor surface tension and the crack dimension. The effect of wedging pressure is to increase the tensile stress perpendicular to the tip of a crack or fracture. Capillary pressure is the pressure that drives a wetting fluid into a crack or fracture.
2.3.2 Westwood Approach

Westwood (1974) proposed that the chemo-mechanical effects in the presence of surface-active agents are result of electrostatic interaction between the adsorbed monolayer of species from the chemical environment and the bulk fluid. Westwood (1974) further proposed that, even if propagation of fast moving cracks in ceramic solids is not affected by presence of surface active agents, slow-crack growth is affected. The nucleation and mobility of dislocation near the crack tip are affected by a type of surface electrostatic potential called the zeta-potential (). The zeta potential is the electrostatic potential between the monolayer of adsorbed ions and molecules from the surfactant and the bulk-fluid. The sign and magnitude of the zeta-potential are dependent on the concentration of the surfactant in the fluid medium. 15

If the zeta-potential is zero ( = 0), the mobility and nucleation of near-surface dislocations are minimized. Westwood (1974) noted that the stresses required to propagate a crack are reduced because all energy is directed toward breaking the bonds at the tip of the crack. In a highly positive or negative zeta-potential environment, the nucleation and mobilization of near-surface dislocations are enhanced. Energy is consumed in the process of nucleation and mobilization of dislocations, which blunt the tip of the crack or fracture and inhibit propagation of the crack Any deformation that involves plastic deformation would be inhibited at = 0 and enhanced at 0. Plastic deformation invariably involves the processes such as surfacedislocations in the deformation-zone, which result in crack-tip blunting. The presence of dislocations would result in stored elastic strain energy around the crack and therefore an increase in the amount of work required propagating the crack. Westwood (1974) noted that dislocation mobility in ionic solids such as MgO is considerably influenced by dislocation extrinsic point defect interactions and the state of ionization of these defects will be influenced by surface potential. For covalent solids such as alumina and certain crystalline silicates, dislocation-lattice interactions dominate dislocation mobility and therefore, they might be immune to any chemo-mechanical effects. Wu et al. (1978) showed that the linking of microcracks is a significant mechanism of crack growth in crystalline brittle materials and therefore, application of Westwood approach alone to explain the effect of dislocation-mobility during crack propagation might be insufficient. The electrical potential (potential) is measured at the slipping plane, i.e. the plane at which relative motion takes place. Dunning (1984) postulated possibility of enhanced penetration of a liquid into cracks or flaws in a material at zero-zeta ()

16

potential due to absence of any shearing resistance between the bulk fluid and the adsorbed species on the solid material. Westwood (1974) related an increasing rate of penetration during drilling through silicate materials in presence of cationic surfactants to zero potential as compared to silicates in water, which exhibit negative potential. Dunning, Lewis and Dunn (1980) have linked changes in hydrofracture strength and microfracturing rate in orthoquartzite to potential. Ishido and Mizutani (1980) showed that maximum reduction of strength in quartz diorite is around zero potential. However, Dunning et al. (1980) showed that Westwood (1974) and Rebinder et al. (1944) do not accurately predict the crack-propagation behavior in quartz. Freiman (1984) raised question whether zeta potentials measured on bulk solids or powders will be similar in any way to those within the small confines of a crack tip. During subcritical crack growth in rocks, various other mechanisms are also at work, which are irreversible in nature, as opposed to the reversible effects of the environment discussed in preceding paragraphs. Atkinson (1984) identified stress corrosion, dissolution, diffusion, ion-exchange and microplasticity as the major mechanisms, all of which are effected by chemical effects of the pore fluid. Various phenomenological approaches using the theory of reaction rates in conjunction with continuum mechanics description of tip geometry (Hillig and Charles, 1965; Wiederhorn et al., 1980) have been taken to incorporate variables such as temperature, pressure and concentration (activity). Yet, the fundamental understanding of chemical interactions at the crack-tip is not complete. A complete description is yet to be formulated so that data from one system could be used to predict the response of another.

17

2.3.3

Stress Corrosion

Stress corrosion presumes that the chemical reaction between strained bonds and an environmental agent produces a weakened state which can be broken at lower stresses than the unweakened states. The general expression for weakening for silicate glasses and quartz in water environments is proposed as (Scholz, 1968, 1972; Martin, 1972; Atkinson, 1979; Martin and Durham, 1975):
H O H + [ Si O Si ] Si OH HO Si 2 [ Si OH ] .

Charles (1958) proposed following expression for corrosion of silica glasses in basic environments: Si O Si + OH Si O + Si OH . Michalske and Freiman (1982) approached the molecular interaction problem from an electron orbital viewpoint. They postulated that for crystalline silicates and silicate glasses, the strained Si-O bonds at crack tips can react more readily with the environmental agents than unstrained bonds because of a strain induced reduction in the overlap of atomic orbitals. It was envisaged that the incoming water molecule interacts with the Si-O-Si crack-tip bond in three stages (Figure 2.6): 1) Step A involves attachment and alignment of water molecule with the bridging bond, 2) Step B involves reaction where water molecules donate an electron to the silicon and a proton to the oxygen, in the stretched linkage unit and 3) Step C involves rupture of a weak hydrogen bond and creation of a fracture surface saturated with hydroxyl groups.

18

Figure 2.6: Interaction between water molecules and strained crack-tip bond in glass: (A) adsorption, (B) reaction and (C) separation (Michalske and Freiman, 1984). Stress-corrosion of Si-O-Si bonds has been attributed to both ionized water (Wiederhorn et al., 1980) and to molecular water (Michalske and Freiman, 1982). Freiman (1984) correlated the rate of stress-corrosion to activity of the corrosive agent. Stress-corrosion reactions in calcite rocks are even less understood. Possible chemical reactions that accompany stress corrosion crack growth in the complex silicates biotite and feldspar has been indicated by Barnett and Kerrich (1980), but these complex reactions were also not well characterized.
2.3.3.1 Chemical Reaction Rate Theory

The thermodynamic formulation of reaction rate theory is based on the assumption that reactants are in a state of equilibrium with an activated complex formed during the reaction and which in turn decomposes to form the reaction products:

A + B ( A B )* C
The reaction rate, kr for such a single-step reaction is given by: 19

(2.16)

kr =

kT [ A] [ B] exp ( G * / RT ) h

(2.17)

where k is Boltzmanns constant, T is the absolute temperature, h is Plancks constant, R is the gas constant, [A] and [B] are the concentrations of the reactants, and G* is the change in partial molar free energy between the initial and activated state of reaction. G* can be expressed as: G * = T S * + E * + PV * (2.18)

where S* is the activation entropy, E* is the activation energy, V* is the activation volume and P is the pressure at the crack-tip. Entropy (S) is a measure of the extent of randomness or disorder in a system. The difference between the entropy of the transition state and the sum of the entropies of the reactants is activation entropy S*. The temperature dependence of the rate constant, kr is characterized by the activation energy of the experiment. The pressure dependence of the chemical reactions is characterized by the activation volume V* (d (ln kr)/P = V*/RT). The Charles-Hillig model (Charles and Hillig, 1962; Hillig and Charles, 1965) proposes that catastrophic delayed failure is triggered by the interaction between thermodynamics and chemical kinetics of a chemical corrosion process on a glass that is subjected to a tensile stress. Wiederhorn et al. (1980), on the basis of the Charles-Hillig model, treated the rupture process at crack tips as a chemical reaction. By assuming that 1) the crack tip can be modeled as an elastic continuum, 2) the crack tip has an elliptical shape with a curvature equal to , 3) the pressure term in the Equation (2.17) can be replaced by negative crack-tip stress (I = 2KI/()), and 4) the chemical potential of reactants should be modified to include surface curvature, Wiederhorn et al. (1980) obtained the following expression for the free energy of activation:
20

2 K * ( *V * V ) I G = T S + E V ( )1/ 2
* * *

(2.19)

where KI is the applied stress intensity factor and V is the partial molar volume of the material undergoing reaction. Wiederhorn et al. (1980) further suggested that the experimental crack-growth data in region I can be expressed by the following empirical relationship
E * + bK I v = vo exp RT

(2.20)

where vo, E*, and b are empirical constants of the fit. Here, vo is approximately proportional to the activity of the reactive species, E* is the related to the stress-free activation-energy and b is related to the activation volume. For chemical reaction rate theory to be consistent with the empirical relationship, the term containing the stress intensity factor in Equation 2.19 must be equal to the terms containing the stress intensity factor in Equation 2.20 V * = b . 2 (2.21)

Therefore by assuming a crack-tip radius and by measuring the slope of the v-K curve, the activation volume of the reaction can be determined. The limitations of the above model are the approximation of the the crack-tip profile as a smooth, rounded ellipse. Further, only linear term for crack-tip stress has been introduced into the energy barrier for activated crack-growth. Thus, the linearity expressed between the logarithmic stress-intensity factor and the crack-velocity loses sound physical basis. Further, there is no provision for crack-healing.

21

To overcome these limitations, Lawn (1975) developed a model to explore crackmotion kinetics at atomistic level, based on the lattice trapping theory of Thomson and coworkers (Thomson 1973, Thomson et al. 1971). This model was further developed by Cook et al. (1993) to derive atomistic variables from macroscopically measured variables.
2.3.3.2 The Lawn-Cook Model

Lawns atomistic model (Lawn, 1975) considers an ideally brittle fracture crack in which sequential bond rupture occurs via the lateral motion of atomic kinks (considering each atomic-scale jump as an energy-barrier), which are enhanced by thermal fluctuations. Chemically enhanced subcritical cracking is a two stage process: transport and reaction. Reactive species must be transported to the crack-tip before reactions can occur that facilitate crack extension. The slower of these two steps will control the rate of the overall process. By considering as the sum of a linear reversible surface energy term and a nonlinear trapping term approximated by a harmonic function with atomic periodicity, including statistical thermodynamics of a Maxwell-Boltzmann distribution to characterize bond-rupture frequency and incorporating chemical potential term to modify surface energy function, Lawn (1975) attempted to explain crack-growth on the atomic scale and while incorporating the effect of some key control variables such as applied loading, chemical concentration of reactive species and temperature on crack-velocity.

22

Mass Flow (viscous fluids)

Free molecular flow (dilute gases)

Activated diffusion

Adsorptive reaction
Solute diffusion (liquids)

Figure 2.7: Schematic of potential rate-limiting phenomena (Lawn, 1993). The fundamental basis of this model is that the frequency of bond-rupture and bond-healing in a reactive environment is modified by the magnitude of the energy release rate. The net frequency of bond-rupture is represented by Maxwell-Boltzmann statistics (Lawn, 1975) as:
U * U * f = f o exp + exp kT kT f o = kT / h where fo = characteristic lattice vibration frequency, k = Boltzmanns constant, T = absolute temperature, h = Plancks constant.
* * and U terms The activation energies for kink advance and retreat are represented by U +

(2.22) (2.23)

which are modulated by the mechanical energy release rate to promote macroscopic crack motion. For a simple chemical reaction of the form

X + B B* ,

23

where X is the reactive environmental species, B and B* represent the unbroken and activated complex state, the rate of change of surface potential or the fracture resistance R, for an increment in crack area A (Cook and Liniger, 1993). dU s R = ( B* B ). N . X . N + R sin ( 2 NA ) dA where X = chemical potential of the environmental specie X, (2.24)

B = chemical potential of the reactant, B, B* = chemical potential of the activated complex, B*,
N = number of bonds per unit area, NA = total number of broken bonds. The last term is included to account for trapping and gives a periodic fracture surface energy term. On integrating the above equation, surface potential is obtained, which is periodic in bond-separation: U s = uo ( NA) u1 cos ( 2 NA ) , 2 (2.25)

where uo represents the energy required to break the bond in the reactive environment
uo = ( B* B ) . A ,

and u1 the intrinsic energy for bond rupture, R . u1 = N Cook et al. (1993) showed that for small departures from the equilibrium Griffith condition, the activation energy barriers can be expressed as a function of mechanical energy release rate, G as

G / N + uo * = u1 1 U 2u1
24

(2.26)

Further, the rate of increase of crack area, dA/dt can be expressed as f/N where f is given by the Equation (2.22). An expression for crack velocity can be obtained by combining Equation (2.21) and Equation (2.25): G 2 v = vo sinh (2.27)

Equation (2.26) can be used to fit the experimental v-G curves and the macroscopic crack velocity parameters (vo, , ) obtained can be related to the atomistic bond rupture parameters (fo, N, w, uo, u1) as

vo =

2 fo u exp 1 , Nw kT

(2.28) (2.29) (2.30)

2 = uo N ,

= 2 NkT .

Here, it can be noted that Equation 2.27 contains a zero and sign reversal at the equilibrium point G = 2. As discussed earlier, the energy per unit area 2 is the surface energy that is sensed experimentally. Cook et al. (1993) noted that Equation 2.27 has two terms: a local, kinetic term vo, containing information about the mechanism of bondrupture in the intrinsic activation barrier u1; and a global, thermodynamic constraint, the sinh function, containing information about the departure from equilibrium in (G-2). As noted earlier, uo represents the energy required to break the bond in the reactive environment (global term) and u1 the intrinsic energy for bond rupture (local term). Cook model is to be extended to the porous rock material.
2.3.3.3 Power Law vs. Exponential Law

Charles (1958) fit experimental static fatigue data of glass to a power law and proposed the following expression for subcritical crack growth velocity
25

' v = vo exp ( H * / RT ) K n

(2.31)

where v is crack-velocity, H* is activation enthalpy, R is the gas constant, T is temperature, vo and n are constants and n is called the subcritical index. Charles and Hillig (Charles and Hillig, 1962; Hillig and Charles, 1965) favored reaction rate theory for constitutive modeling of slow crack-growth. It was subsequently developed by Wiederhorn et al (1980). A brief synopsis of their model is discussed in Section (2.3.3.1). Equation (2.20) is a simple form of the exponential law. The Lawn-Cook model (Lawn, 1975; Cook and Liniger, 1993) proposed a hyperbolic sine functional relationship (Equation 2.27) between velocity and fracture-mechanics parameter. All these models fit quite well with the experimental data but diverge significantly outside the range of observations (Atkinson, 1987). Despite the sophistications involved in the reaction-rate theories and atomic theories, Charles power law is the most used method to characterize subcritical growth, particularly for geological materials. Costin (1987) noted that Equation 2.20 predicts a finite stress corrosion threshold, whereas Equation 2.31 does not. No stress-corrosion threshold has been observed for rock (Adefashe, 2006; Atkinson, 1984). The other advantage of using power-law model is that time to failure can be calculated more easily, as integration operation on a power-law model is easier. Various authors have presented v-K relationship in different mathematical forms. The bases of all these models are extension of empirical power-law relationship (Charles, 1958), exponential relationship from chemical reaction-rate theory (Charles and Hillig, 1962) or, relationship from atomistic model (Thomson, 1973; Lawn, 1975). A compendium of various models from literature is presented in Table 2.1.

26

Table 2.1: Expressions for the crack velocity from literature.


Reference Expression

Charles, 1958a Charles and Hillig, 1962 Hillig and Charles, 1965 Wiederhorn, 1967 Kies and Clark, 1969 Wiederhorn and Bolz, 1970 Evans, 1972 Evans and Wiederhorn, 1974

v = k ( m ) n exp( A / RT )

= o exp ( E * +bK I ) / RT

= o exp E * +V *

VM / RT

V = exp(H + K I ) / RT
3 dx d o = exp ( f * / kT + G / N o kT ){( p + )G N o f s } dt h

= o exp( E * +bK I ) / RT
V = ' K n exp(H / RT )

V = Vo K InT / To exp(H / RT )

Wiederhorn et al., 1974 Atkins et. al, 1975 Pletka and Wiederhorn, 1982 Cook and Liniger, 1993 Lockner, 1998

= o exp( E * +bK I ) / RT
= AI exp ( (U R ) / kT ) a

K v=v I K0
*

G 2 v = vo sinh
ln o E* + f ( Pc , ) g ( i ) = co RT

27

Loading conditions present in the subsurface are generally favorable for subcritical crack-growth. Rocks are loaded for long periods of time below their fracturetoughness value (Anderson and Grew, 1977; Atkinson, 1976; Atkinson, 1984; Segall, 1984). Many authors have used subcritical fracture growth theory to explain observations of natural fractures (Anderson and Grew. 1977; Olson, 1993, 2004). As noted earlier, v-K diagram and associated relationship from the theories of stress-corrosion has been used in the prediction of tensile failure. The most direct method for doing this is by a simple integration of the v-K diagram (Evans, 1972). Various authors have used different v-K relationship to deduce expression for time to failure. A brief summary of their final results has been presented in Table 2.2.

Table 2.2: Expressions for time to failure from literature.

Reference Zhurkov, 1965 Scholz, 1968 Wiederhorn and Bolz, 1970 Kranz, 1980

Expression

= o exp (U o ) / kT
t f = to .exp ( E / kT + b( S * ) )
2 t0.5 = (4 / ) ( RTK IC / bv0.5 ) N , for Griffith type crack 2 t0.5 = ( RTK IC / bv0.5 ) N , for penny-shaped crack

t f = to . ( ) ; t f = to .exp ( 2.303b )
m

28

2.3.4

Other Mechanisms

2.3.4.1 Diffusion

Experimental evidence (Lewis and Karunaratne, 1981) shows that the dominant mechanism of subcritical crack growth in ceramics at high homologous temperature can be mass transport. Theoretical analysis of the observation has been done by Stevens and Dutton (1971) and Dutton (1974). Potential diffusion paths identified are 1) lattice or bulk diffusion, 2) surface diffusion, 3) vapor phase transport and 4) grain boundary diffusion. For diffusion-controlled crack-growth, sub-critical index, n is often in the range of 2-10, whereas for stress-corrosion crack-growth, n may be much higher (~40 or higher).
2.3.4.2 Dissolution

Dissolution of quartz in aqueous environment is given by reaction: y SiO2 + x H 2O y SiO2 .x H 2Oaq. Quartz shows an increase in solubility with increase in temperature (Fyfe et al. 1978). Solubility of quartz is largely unaffected by dissolved salts or changes in pH until a pH of 9, when there is a large increase in solubility. Dissolution of calcite is described by CO2 + H 2O + CaCO3 Ca 2+ + 2 HCO3 Solubility is greater in NaCl solutions and sea water than in fresh water. Solubility increases with either an increase in partial pressure of carbon dioxide or a decrease in temperature (retrograde temperature dependence).

29

2.3.4.3 Ion Exchange

If the chemical environment contains species which can undergo ion exchange with species in the solid phase and if there is a gross mismatch in the size of these different species, then lattice strain can result from ion-exchange which can facilitate crack extension, e.g. exchange of H+ for Na+ in silicate glasses (Atkinson, 1987a). Another effect of ion-exchange is to modify the chemistry of the crack-tip solution. For glass-water system, the exchange of hydrogen ions for alkali ions increases pH near the crack-tip because of the restricted volume of fluid at the crack-tip. If cracktip pH exceeds 9, reaction becomes very fast.
2.3.4.4 Microplasticity

At high homologous temperatures and low strain rates, in the presence or absence of chemical environment, a damage zone may develop in the stress field ahead of macrocrack tip. In the damage zone, microcracks first get nucleated by inhomogeneous plasticity and subsequently link up to allow macrocrack extension. Available experimental evidence from electron microscopy of quartz suggests that chemically enhanced subcritical crack growth is not accompanied by any significant plastic flow at crack tips up to a temperature of 250 C (Dunning et al. , 1980; Lawn 1983). Galena and calcite show micro-plasticity at low stresses even at room temperature.

30

Chapter 3: Measurement of Subcritical Crack Index


The double-torsion load-relaxation test method (Evans, 1972; Williams and Evans, 1973) has been the most widely technique for measuring subcritical crack growth, particularly in opaque polycrystalline rocks, where crack-length measurements are difficult to make. The wide applicability for this method lies in its low-cost setup with simple test-specimen geometry and loading configuration, the simplicity in the acquisition and analysis of data-sets, relative ease in providing specific environmental controls and a complete K-v diagram from a single load decay measurement for a given environmental condition. In this study, this technique was used to study both critical (fracture-toughness measurement) and subcritical (subcritical-index measurement) crack growth on rock samples from reservoir and outcrops. This chapter provides an overview of the test method, the theoretical basis of data analysis and the underlying assumptions, the techniques used and the precautions taken during sample preparation.
3.1 TEST DESCRIPTION

Double-torsion testing techniques were initially proposed by Kies and Clark (1969) to determine crack-velocity as a function of the driving force, and were subsequently developed by Outwater et al. (1974). Various authors (Evans, 1972; William and Evans, 1973; Evans, 1974; Evans and Johnson, 1975) subsequently developed the complete description of the theoretical basis of this method. Excellent critical studies were further done by Fuller (1979), Pletka el al. (1979), and Tait et al. (1987) to focus on analytical, experimental and practical aspects of this test-technique respectively. 31

The test-specimen of DT test consists of a thin rectangular plate which is loaded in bending via load-cell at one end across a notch or crack in double torsion (Figure 3.1). Load is applied at one end of the specimen, which is supported at the outsides of the same end of the specimen. The crack propagates along the center of the bottom of the specimen under vertical loading from the end where the load is applied. The elegance of this simple technique lies in the fact that the stress-intensity factor is independent of the crack length over a substantial portion of the length of the specimen (Fuller, 1979).

Figure 3.1: Loading configuration of the Double Torsion test (Nara and Kaneko, 2005). The fracture toughness of the test-specimen can be determined from the maximum applied load without need for corrections for the applied load from plot of load vs. crack mouth opening displacement (CMOD; Williams and Evans, 1973). Cracklength does not enter into the equation for the derivation of fracture-toughness or, stressintensity. Three loading methods for DT Tests constant load method (Kies and Clark, 1969), incremental displacement method (Evans, 1972), and load relaxation method 32

(Evans, 1972; Williams and Evans, 1973) have been described in the literature, but the load relaxation method is most commonly used. It has an advantage over other methods because a range of stress intensity factor-velocity (K-v) data-points can be obtained in one given test, whereas other methods provide for the determination of a single data-point during a given experimental run.
3.1.1 Mathematical Description of Stress Intensity Factor for Double Torsion Test

Williams and Evans (1973) provided a complete description of the theoretical development of the stress intensity factor for this test method. The double torsion specimen is considered as two elastic torsion bars with a rectangular cross-section subjected to load 2P as shown in Figure 3.2. For small deflections and for bars where width is much greater than specimen thickness, it has been shown (Novozhilov, 1961) that the torsional strain, , is given by:

y 6Ta , wm Wd 3G

(3.1)

where T = torsional moment, (P/2) wm, P/2 = total load applied to one bar, G = shear modulus of the material,
a = crack length,

d = bar thickness, W/2 = bar width, wm = moment arm. Equation 3.1 can be rearranged such that
2 a y 3wm S, 3 P Wd G

(3.2)

33

where, S is the elastic compliance. The strain-energy release rate for crack extension G, and the specimen compliance are related by (Irwin and Kies, 1954)
G= P 2 dS , 2 dA

(3.3)

where, A is the area of the crack.

Figure 3.2: Schematic diagram of a point loaded rectangular torsion bar (from Williams and Evans, 1973). If the shape of the crack front is independent of crack length, then Equation 3.3 becomes,
2 3EP 2 wm K = , 3 2 2 Wd d G (1 ) n

(3.4)

where dn is the plate thickness in the plane of the crack.

34

For the plane-strain conditions, the stress-intensity factor, K, is related to the strain energy release rate by (Paris and Sih, 1965)
EG 2 K = , 2 1
1

(3.5)

where E is the Youngs modulus and is the Poissons ratio. Thus substituting Equation 3.6 into 3.5 gives
2 3EP 2 wm K = , 3 2 2Wd d n G (1 )

(3.6)

E and G are related by, G= E , 2(1 + ) (3.7)

Equation 3.7 becomes


K = Pwm 3 Wd d n (1 )
3

(3.8)

Equation 3.9 is the expression for stress intensity factor under plane-strain condition for double torsion load relaxation method (Pletka et al., 1979). Fuller (1979) and Swanson (1981) analyzed the assumptions used in this derivation: a) Mode-I failure, b) crack profile independent of the crack-length, c) no frictional constraints along the sides of the torsion arms (i.e. the sides of the crack), d) the elastic strain energy which provides the driving force for crack extension is derived only from the strained torsion arms with a negligible amount of deformation occurring ahead of the crack tip, e) plane-strain or plane stress, and f) the elastic constants of the

35

sample are independent of the test environment. Swanson (1981) attributed the scattering in the experimental observation to possible violations of some of these assumptions.
3.1.1.1 Expression for Crack Velocity for Constant Displacement Method

An empirical compliance calibration for the double torsion specimen shows that the specimen compliance, S, is linearly related to the crack length, a as (Evans, 1972; Williams and Evans, 1973), S= y = So + Ba, P (3.9)

where y is the displacement of the loading point, P is the load, So is the elastic compliance of the intact specimen and B is an experimental constant. Differentiating Equation 3.10 as a function of time at constant displacement gives an expression for the crack propagation velocity, v as (Williams and Evans, 1973) ( S + Ba) P a v= = o , BP t y t y From Equation 3.10, it can be seen that for constant displacement, P ( So + Ba) = Pi ( So + Bai ) = Pf ( So + Ba f ), (3.11) (3.10)

where, Pi and ai are the initial values of load and crack length and, Pf and a f are the corresponding values at the end of load relaxation. Combining Equations 3.11 and 3.12 gives,
Pi , f ( So + Ba) P a v= = = BP 2 t y t y Pi , f (ai , f + P2 So ) B P , t y

(3.12)

36

In general, except for very low modulus materials (such as polymers), or for very S small crack lengths, o < a (Williams and Evans, 1973) so that B

P a P a v = = i , f 2i , f P t y t y

(3.13)

Equation 3.13 is the expression for the crack propagation velocity determined from the load relaxation curve for a constant displacement double torsion test. Combining Equation 3.9 and Equation 3.13 provides a unique description of the dynamics of subcritical crack growth for a given specimen from the plot of crack velocity, V, versus stress intensity factor, K. From the equations discussed above, two major advantages of double-torsion load relaxation techniques can be cited: a) the stress intensity factor is directly related to the applied load over much of the specimen length (Equation 3.9), and b) crack velocity can be determined without the need for multiple crack length measurements (Equation 3.14).
3.1.1.2 Data Analysis and Reduction Method

A number of equations (Table 2.1) have been proposed to describe subcritical crack-growth. Some of these are very complex because of their tendency to include a comprehensive description of the competing mechanisms. The two most popular are power law (Charles, 1958b) and exponential (Wiederhorn and Bolz, 1970) relationships between crack-velocity and stress intensity factor. The Charles(1958b) power law is the most popular equation (Atkinson, 1984) to describe subcritical crack growth in rocks and minerals because of its ability to describe the whole range of K-v data with appropriate changes in the empirical constants (Atkinson and Meredith, 1987a). Test data for the load relaxation method is normally recorded in the form of applied load versus time. Calculating the crack velocity (Equation 3.14) requires 37

numerical differentiation of individual data points on the load-time curve. Because of the inherent scatter in the acquired experimental data, direct numerical computation is not suitable. Various schemes have been adopted to smooth the experimental data before attempting any numerical computation. Swanson (1984) fitted load-time data to a sixth order polynomial for glass. The drawback of this method is that before a good fit can be obtained, the load-time curve may have to be segmented into separate regions with polynomials of the same or different degrees fitted to each region. This method is not only time consuming but also unreliable as it often computes data that deviates from the well established power law and exponential relations. Using a power law assumption, Holder (personal communication, 2001; described by Rijken, 2005) developed an approach to obtain crack velocity from raw experimental data by calculating a smooth load decay curve that permits numerical differentiation of the load-time data. At the start of the test, the displacement is given as

y=

S , Pi

(3.14)

Combining the above equation with the compliance Equation 3.10 yields
P= Pi B 1 + So a ,

(3.15)

Differentiating Equation 3.16 with respect to time gives

P a = = V= 2 2 t B t B 1 + a 1 + a So So

B Pi So

B Pi So

B 2 P So V, Pi

(3.16)

38

Using power-law dependence of crack-velocity on the load, V = A( P ) n Equation 3.17 can be written as

P = t
A and

B 2 B P A So So A( P) n = ( P)n2 , Pi Pi

(3.17)

B in Equation 3.18 can be related to the initial values of load Pi and its time So

derivative Pi ' as

P' B = ( ni+1) , So Pi

(3.18)

Integrating equation 3.18 gives


P= Pi Pi ' 1 (n + 1) t Pi
1 ( n +1)

(3.19)

Equation 3.19, in principle, does provide a direct determination of subcritical index by a least square fit of load and time to this power-law expression. An iterative procedure to determine the parameters in Equation 3.19. Crack velocity is subsequently determined from Equation 3.14. A typical load-decay data is shown in Figure 3.3.

39

Figure 3.3: Load vs. relative time from Double Torsion test; blue triangles represent actual data acquired and red diamond represents the power-law fit to calculate the subcritical index. The load relaxation becomes noisy at long times (low values of crack velocity) (Beaumont and Young, 1975; Adefashe, 2006). Therefore all data-points, which give a crack-velocity lower than ~10-7 ms-1, were removed during data reduction. This reduces the noise in the collected data-set and therefore, yields a better estimation of sub-critical index from the given experimental run.

40

Figure 3.4: Load vs. relative time from Double Torsion test. Sharp drop in load (dashed blue box) denotes critical failure. Maximum load is used to calculate fracture toughness of the specimen. Figure 3.4 represents a typical load-decay for fracture-toughness measurements. The load is ramped up to a point, where the stress intensity factor exceeds the critical stress-intensity factor resulting in catastrophic failure of the rock sample. Maximum load observed during such a catastrophic failure is used to calculate fracture toughness of the given specimen.
3.2 SAMPLE PREPARATION

Test-specimens were cut into rectangular slab using an oil-cooled saw. It has been shown that for rock sample fracture-toughness is independent of the specimen thickness (Schmidt, 1980). However, Atkinson (1979) showed that to avoid erroneous estimation of KI, width should be greater then twelve times thickness of the sample, and Pletka et

41

al.(1979) suggested that the specimen length L should be greater than twice width, W, i.e.,
12d W L / 2

(3.20)

A typical dimension for the test-specimens used in this study was 4 x 1.1 x 0.07 (L x W x d).

Crack Length, c Crack length, a

Figure 3.5: Schematic representation of the compliance relationship for DT test method (adapted from Fuller, 1979). The double-torsion load relaxation method assumes that the stress intensity factor, KI, is independent of crack length, a, for the double torsion test (Evans, 1972), Pletka et al. (1979) showed that KI varies along the specimen. Pletka et al. (1979) proposed that in order to ensure cracks are in the constant KI regime for a specimen of length L and width W; the crack length should be between W and L-W. Therefore, pre-cracking a DT specimen is necessary before performing any fracture mechanics studies (Pletka et al., 1979). 42

COMPLIANCE,

Trantina (1977), using finite element analysis, showed that the range for which KI is independent of the crack length, a is
0.55W a L 0.65W

(3.21)

All the samples used in this study were precracked in a controlled manner. The precracking of the sample was allowed to continue to ensure that the crack has reached the steady-state front (Chevalier et al., 1996). For fracture-toughness measurement, blunt crack give higher value because for a given applied load the stress intensity at a blunt crack tip is smaller than for a sharp tip.

Figure 3.6: Evolution of the crack front with crack extension (Chevalier et al., 1996). Pabst and Weick (1981), performing tests on commercial alumina specimen, showed that the level of reproducibility was highest for the specimen without guide grooves. However, specimens without a guiding groove require a well-finished surface as well as an extremely balanced loading device. Nara and Kaneko (2005) showed that the shape of the guide groove affects experimental results. They conducted tests on rectangular, semi-circular and rectangular grooves and observed the highest linearity and least scattering for the specimens with a rectangular guiding groove. A rectangular groove, approximately 0.02 inches deep, centered with respect to width, was cut along the length of the test-specimens to provide a guide for the crack- propagation. The dual torsion apparatus was assembled of stainless steel (SS 316) parts to make the equipment amenable to hostile environments. 43

Chapter 4: Experimental Results and Analysis


In the oil and gas industry, surface active agents at high pH are used in stimulation operations (hydraulic fracturing) and EOR activities (chemical flooding). pH of the fracturing fluid can go up to 12 (Economides and Nolte, 2000) depending upon the nature of the crosslinker used in fracturing fluid. During Alkaline flooding and AlkalineSurfactant-Polymer flooding, pH of the injected fluid can go up to 13 (Green and Wilhite, 1998). Tests were carried out in a surfactant solution of 0.3% to 0.5% (w/w). This represents the typical strength of an ASP solution used during chemical EOR processes for enhanced oil recovery. As discussed earlier (Chapter 2), fluid environments with surface active agents (surfactants) or alkali have potential to alter sub-critical cracking behavior. During geological time, different chemical controls - which have potential to alter sub-critical crack-growth behavior of the rocks - can exist. The focus of this work is to understand the effect of different environmental controls on sub-critical crack growth in rocks. Tests were conducted in air, water (brine and de-ionized), aqueous surfactant solution and, aqueous sodium hydroxide solution using measurements with a double torsion beam test technique. Subcritical indices and fracture toughness were measured for these different environments for rock samples of sandstone.
4.1 PETROGRAPHIC INFORMATION

Seven different sandstone samples were tested along with one sample of granite. The sandstones were from subsurface cores from various locations in North America. The Canyon I, II, and III samples are from the Canyon Sands Formation of Texas (Miller

et al., 1994). The Canyon I and II samples are from the Sun Oil Dunbar well. Canyon-I is
from depth of 5950.5 ft., whereas Canyon II is from depth of 5949 ft. Canyon III is from 44

the Phillips Ward "C" No. 11 well. Sandstones A1, A2, and A3 are from wells in western North America, and the Mesa Verde core is from a well in the Piceance Basin in Colorado. Petrographic analysis for the 7 sandstone samples (but not the granite) consisted of 250 count point-counts on thin-sections. Table 4.1(a) and Table 4.1(b) summarize the point-count data from these analyses. Figure 4.1 shows the Q-F-L diagram based on Folk (1980) composition diagram.

45

Table 4.1(a): Point-count data. Numbers in each column denote raw count out of the total number of counts in the right-most column.
Quartz Quartz Plagioclase overSandstone Calcite Calcite (poly growths (mostly pore Sample Names and Grains Cement sericitized) and Mono) cement

clay lithic

mica other lithic (usually (vrf, srf, sericite mrf) form)

chert unknown

other

Total Counts

Canyon-I

91

58

36

11

38

0 1 (opaque) 0

250

Canyon-II

172

26

10

24

243

Canyon-III

120

59

31

25

250

Sandstone-A1

131

17

52

22

12

250

Sandstone-A2

139

15

40

22

13

250

Sandstone-A3

140

63

13

10

12

250

MesaVerde

109

14

84

33

249

46

Table 4.1(b): Summary of point-count data reported percentages.


Sandstone Samples Canyon I Canyon II Canyon III Sandstone A1 Sandstone A2 Sandstone A3 MesaVerde Total Cement 23.2 0.8 23.6 2.0 6.0 1.2 34.5 Total Pore 4.4 4.1 2.0 2.8 3.2 5.2 0.0 Total Framework grains 72.4 95.1 74.4 95.2 90.8 93.6 65.5
Name of rock

Lithic 16.5 10.8 10.4 10.3 4.6 4.9 0.0

Feldspar 15.7 11.3 12.9 23.3 18.4 28.0 22.0

Quartz 67.8 77.9 76.8 66.4 77.0 67.1 78.0 feldspathic litharenite subarkose subarkose arkose subarkose arkose subarkose

47

C-I C-II C-III SA-I SA-II SA-III MV

Canyon-I Canyon-II Canyon-III Sandstone-A1 Sandstone-A2 Sandstone-A3 MesaVerde

Figure 4.1: Q-F-L diagram for the point-count data for different specimens.

48

Most of the samples fall on the arkosic side of the QFL diagram (Figure 4.1), but are higher in quartz than feldspar grains. The samples vary in quartz framework grains percentage from 66% to 78%, with 13% to 28% feldspar grains and 0% to 16% lithics. The lithics are predominantly clay-rich, chert and carbonate grains. Total cement varies from 1% to 35%. The Mesa Verde sample (a subarkose) has the highest framework grain percentage of quartz at 78%, with 22% feldspar and no lithics (Table 4.1). The Mesa Verde sample has the highest amount of intergranular cement at 34.5%, most of which is calcite (the only sample tested with calcite cement). The point count porosity for this sample was 0%, but all the sandstones counted at 5% porosity or less. The Canyon samples vary from subarkose to feldspathic litharenite. They represent the most lithic-rich of the sandstones tested. In Canyon II, the total cement was very small (1%) whereas in Canyon I and III total cement was almost 20%, mostly quartz. The remaining sandstones, A1, A2 and A3, vary in composition from arkose to sub-arkose. Sandstones A2 and A3 have considerable amounts of mica grains (22% and 12%, respectively), while A1, A2 and A3 all have around 10% chert grains.
4.2 RESULTS SUBCRITICAL INDICES AND FRACTURE TOUGHNESS

Measurements of subcritical indices and fracture-toughness were carried out in ambient air, de-ionized water, brine, surfactant and alkaline fluids. Cationic surfactant (Quaternary Ammonium Compound M-Quat-32) and Sodium Hydroxide (NaOH) were used as proxies to create equivalent environmental control that might exist in the subsurface. During the tests, each specimen remained submerged in the respective chemical environment. Results from all the tests are summarized in Table 4.2 including the stress-intensity factor corresponding to peak load applied (KIMAX) and the load drop observed after 300 seconds (P) of crack propagation. 49

Table 4.2: A summary of the dual torsion tests performed for all samples. n is the subcritical index, KIMAX is the stress-intensity factor corresponding to peak load, and P is the load drop observed after 300 seconds.
Air Formation Sample name 5950_1B 5950_4A 5950_6A n 51 42 51 74 79 80 73 102 KIMAX 3.56 3.70 2.14 2.15 3.89 4.01 3.26 3.22 P 0.435 0.548 0.267 0.146 0.415 0.415 0.347 0.211 46 62 73 46 42 45 2.71 2.74 2.94 4.20 4.13 3.86 0.360 0.224 0.181 0.559 0.765 0.528 46 54 54 67 3.06 0.302 65 51 46 53 42 73 57 57 2.57 2.49 2.87 2.87 1.409 0.754 0.422 0.424 55 47 63 62 61 59 61 71 51 60 1.94 0.279 60 51 101 122 55 78 79 66 2.06 2.07 3.42 2.31 2.63 1.64 0.191 0.153 0.468 0.352 0.346 41 0.247 30 36 34 2.82 2.73 3.30 0.757 0.504 1.956 4.62 0.980 23 3.39 0.883 29 34 3.98 3.80 1.325 1.007 49 3.64 0.670 2.07 1.98 0.275 0.263 1.75 2.02 1.97 1.95 1.95 3.17 2.97 3.01 2.45 0.195 0.256 0.208 0.212 0.191 0.430 0.353 0.298 0.336 1.71 2.16 2.19 2.12 0.162 0.263 0.370 0.300 2.50 2.80 2.68 0.393 0.439 0.401 n Water KIMAX P n Surfactant KIMAX P n Caustic KIMAX P

47 48

4.58 3.93

1.357 0.621

Canyon I

5950_6B 5950_8C

5950_9A

5949_3A 5949_4A 5949_4B 5949_5A 5949_6A 5949_7B

Canyon II
5949_8A

5949_8B 5949_9B 5949_10A 5949_10B 5949_11B 6071_3A 6071_3B

Canyon III

6071_9A 6071_10A 6071_10B 6071_11B

50

Table 4.2: Contd.


Air Formation Sample name X22_1A Water P 0.813 0.811 0.720 27 31 36 39 77 87 74 84 83 3.47 3.55 3.70 3.73 3.84 0.512 0.456 0.538 0.407 0.443 50 3.23 0.844 39 31 27 61 65 69 63 3.93 4.32 4.42 4.53 0.691 0.802 0.699 0.678 32 33 77 86 59 270 149 69 82 308 149 172 4.38 4.59 4.77 3.67 4.18 4.28 2.80 2.17 2.44 2.88 0.741 1.313 1.054 0.246 0.349 0.785 0.460 0.126 0.261 0.268 24 29 34 27 32 3.33 3.28 3.61 4.68 4.60 1.924 1.240 1.352 2.131 1.899 30 26 45 44 90 294 346 66 68 62 96 3.77 3.28 2.78 4.04 4.85 4.00 3.31 0.616 0.230 0.190 0.734 1.472 1.495 0.485 52 4.70 1.328 3.49 4.28 1.052 1.381 4.59 2.313 4.71 1.887 3.39 3.29 1.136 1.222 31 3.49 1.128 3.49 1.075 3.61 1.346 3.76 1.945 3.96 3.94 1.281 1.109 40 3.98 0.956 3.94 1.731 4.44 1.804 28 3.30 0.512 n KIMAX P n Surfactant KIMAX P n Caustic KIMAX P

n
64 64 64

KIMAX 4.20 4.34 4.59

Sandstone A1

X22_2B X22_3A X22_4A

X301_1A

Sandstone A2
X301_2A X301_3A

X326_1A

Sandstone A3
X326_2A X326_3A

MV3

27 35

4.03 1.182 4.61 1.321

MV4

MesaVerde
MV6 MV7 MV8 G2

G3

Granite
G4 G6 G10

51

4.2.1

Dependence of Subcritical Indices on Loading Magnitude

A number of tests have bumps (Figure 4.2) in the load-decay data. These bumps can be attributed primarily to heterogeneity in the test specimen. All the data with discernible bumps into the load-decay data were deemed not reportable, as it is difficult to fit a reliable curve through bumpy data. Of the remaining tests, there was considerable variability in the subcritical indices (n) derived from different test-specimens from the same depth-interval. Different test-specimens from a small core-sample are from too close a separation to attribute this high variability to heterogeneity. Therefore, further analysis was done on the loading conditions of different tests on specimens from each of the depth-intervals. The first step was to analyze the load decay within a given time-interval (300 seconds) on different tests from same depth-interval. Subcritical indices from every reported data-set was plotted against the load-decay in that test (Figure 4.3 through Figure 4.6) for different rock-types in air, water, and caustic. All the plots show a distinct behavior of decreasing subcritical index, n with increasing load drop, with many of the curves approaching a stable value at load drop of 0.5 lbs or more. The variability of subcritical indices with load-drop was systematically analyzed for tests in air. The control on the magnitude of the load drop during any crack propagation event seems to be related to the load level at which crack propagation was initiated. Figure 4.7 to Figure 4.10 plots KI at the peak load versus load drop during over 300 seconds for the different formations. The higher the initial KI, the greater was the load-drop during propagation. A number of tests were carried out in air and different load-drops were observed for different peak-loads. This can be seen as a wide range of load drops in air over a 52

broad span of KIMAX in Figures 4.3(A) Figure 4.10(A). Subsequently, for water and caustic environments, tests were carried out at higher loads. Therefore the spread of data for tests in water and caustic are smaller than that for test in air. The implication of these experiments is that for tests in water and caustic, peak-load was closer to the fracturetoughness value equivalent for many of the tests. In Figure 4.3 (A), subcritical indices for sandstones A1, A2 and A3 are more than 75 for values less than 0.5 lbs of load drop. At load drops higher then 0.5 lbs, subcritical indices are clustered around 65. For tests in water, load-drops are relatively higher, around 1.0 lbs. For tests in caustic, load-drop range varies from 0.5 lbs. to 1.25 lbs. Increasing load-drop does not affect the subcritical indices for the tests in water and caustic and a flat trend of subcritical indices was observed for a wide range of load-drops higher than 0.5 lbs. For Canyon-samples, multiple tests in air gave load-drops less than 0.4 lbs resulting in a wide scatter of n (Figure 4.4 (A)). Much smaller variability (637) in subcritical indices is evident for load-drops more than 0.4 lbs. For tests in water,

subcritical indices flattens to a value around 50 for load-drops 0.4 lbs or higher. Loaddrops for tests in water lie between 0.4 to 0.75 lbs for the plateau. For tests in caustic, n flattens for load-drops between 0.5 to 2.0 lbs. A load-drop as high as 2.0 lbs. doesnt alter the subcritical index value substantially. Mesa Verde samples show a very big span of subcritical indices from 175 to 75 in air (Figure 4.5 (A)). The high values of subcritical indices however correspond to a very small load drop of ~0.25 lbs. Load-drops higher than 0.50 lbs. give consistent measurements around a mean value of 75. In water and caustic, load-drops are much higher, 1.0 lbs to 2.25 lbs (Figure 4.5 (B) and (C)). A consistent plateau around 30 was observed for tests in both water and caustic. 53

Granite also shows a trend similar to Mesa Verde for tests in air (Figure 4.6 (A)). However, for smaller load-drops ~0.25 lbs, n values are consistent with a mean of 77. Load-drops from all the tests in water range from 1.0 lbs to 1.5 lbs (Figure 4.6 (B)) and very consistent values of n with a mean of 47 are observed. Figure 4.7 plots load-drops vs. initial peak-load for Sandstone A1, A2 and A3 for tests carried out in air, water and caustic. Figure 4.7 (A), (B) and (C) show that KIMAX values for all the tests reported are close to 4 MPa-m. Fracture-toughness value for sandstones A1, A2 and A3 are from 4 MPa-m to 4.6 MPa-m. Thus, all the tests were done at loads close to fracture-toughness levels (Table 4.2). Therefore, ideal loading conditions were being satisfied for specimens from Sandstone A1, A2 and A3. This resulted in optimum load-decay for tests on samples from these formations and therefore, more consistent values of subcritical values with std. deviation less than 5 (Table 4.3(a)) were obtained from multiple tests. Also, for similar ranges of peak-load, load-drops in water and caustic are higher than that in air. Variability of load-drop for small change in KIMAX for tests carried out in caustic is large (Figure 4.7 (C)). However, as discussed earlier, even very high load-drops dont alter the estimated subcritical indices. In Figure 4.8 (A), for values of KIMAX less than 3.5 MPa-m, there is an increasing trend of load drop with increase in load, but the load drop is higher than 0.50 lbs at KIMAX values close to 4 MPa-m. Table 4.3(b) shows that the Fracture toughness value, KIC for Canyon I is 4.23 MPa-m. So, at values more than 80% of the KIC, load drop is higher than 0.500 lbs. Also, similar to sandstones A1, A2 and A3, between a KIMAX of 3 MPa-m and 4 MPa-m, variability in load-drop is large. Again, load drops higher than 0.50 lbs show less variability in results. The Mesa Verde (Figure 4.9 (A)) and the granite (Figure 4.10 (A)) show a wide range of load drops in air over a broad span of KIMAX (2 MPa-m to 5 MPa-m). For 54

those same rocks in water and caustic, there are no small load drops. For similar ranges of peak-load, load-drops for tests in water and caustic are comparable or higher as compared to those in air. This trend is very similar to that observed with Canyon samples and sandstones A1, A2 and A3. This suggests good fracture propagation in the environments of water and caustic are easier than in air. It also reflects bigger stress drops for tests in more active control environments. This discussion shows that tests carried out at loads closer to fracture-toughness values give consistent values of subcritical indices from multiple tests. Otherwise, erroneously high values of subcritical indices are obtained. Load-drop in first 300 seconds can be used as another yardstick to ensure correct measurement of subcritical indices. Tests with load-drops less than 0.50 lbs are generally not representative. However, since load drop is a function of sample size, this conclusion needs to be applied carefully under different circumstances. A good approach would be to perform a spread of experiments to establish acceptable load-drops and then average the measured values of subcritical indices which lie on the plateau of subcritical index vs. load-drop curve.

55

Figure 4.2: Load-decay data of sample 5950_1A (Test a) tested in air shows bumps during initial time.

56

100

100

100

(A)
subcritical index, n subcritical index, n 75 75

(B)
subcritical index, n 75

(C)

50

50

50

25

25

25

0 0.0 0.5 1.0 load-drop, lbs. 1.5 2.0

0 0.0 0.5 1.0 load-drop, lbs. 1.5 2.0

0 0.0 0.5 1.0 load-drop, lbs. 1.5 2.0

Figure 4.3: Subcritical index vs. load-drop for Sandstone A1, A2 and A3 combined (A) air, (B) water, and (C) caustic.

200

200

200

(A)
suncritical index, n suncritical index, n 150 150

(B)
suncritical index, n 150

(C)

100

100

100

50

50

50

0 0.0 0.5 1.0 load-drop, lbs. 1.5 2.0

0 0.0 0.5 1.0 load-drop, lbs. 1.5 2.0

0 0.0 0.5 1.0 load-drop, lbs. 1.5 2.0

Figure 4.4: Subcritical index vs. load-drop for Canyon I, II and III combined (A) air, (B) water, and (C) caustic.

57

200

100

100

(A)
subcritical index, n subcritical index, n 150 75

(B)
subcritical index, n 75

(C)

100

50

50

50

25

25

0 0.0 0.5 1.0 load-drop, lbs. 1.5 2.0

0 0.0 0.5 1.0 1.5 2.0 2.5 load-drop, lbs.

0 0.0 0.5 1.0 1.5 2.0 2.5 load-drop, lbs.

Figure 4.5: Subcritical index vs. load-drop for MesaVerde (A) air, (B) water, and (C) caustic.

400 subcritical index, n

100

(A)
300 subcritical index, n 75

(B)

200
`

50

100

25

0 0.0 0.5 1.0 load-drop, lbs. 1.5 2.0

0 0.0 0.5 1.0 1.5 2.0 2.5 load-drop, lbs.

Figure 4.6: Subcritical index vs. load-drop for Granite (A) air, (B) water.

58

2.0

2.0

2.0

(A)
Load-drop (lbs) Load-drop (lbs) 1.5 1.5

(B)
Load-drop (lbs) 1.5

(C)

1.0

1.0

1.0

0.5

0.5

0.5

0.0 0 1 2 3 4 5 6 KIMAX, MPa-sqrt(m)

0.0 0 1 2 3 4 5 6 KIMAX, MPa-sqrt(m)

0.0 0 1 2 3 4 5 6 KIMAX, MPa-sqrt(m)

Figure 4.7: Load-drop vs. maximum-KI for Sandstone A1, A2 and A3 combined (A) air, (B) water, and (C) caustic.

2.0

2.0

2.0

Load-drop (lbs)

Load-drop (lbs)

1.0

1.0

Load-drop (lbs)

1.5

(A)

1.5

(B)

1.5

(C)

1.0

0.5

0.5

0.5

0.0 0 1 2 3 4 5 KIMAX, MPa-sqrt(m)

0.0 0 1 2 3 4 5 KIMAX, MPa-sqrt(m)

0.0 0 1 2 3 4 5 KIMAX, MPa-sqrt(m)

Figure 4.8: Load-drop vs. maximum-KI for Canyon I, II and III combined (A) air, (B) water, and (C) caustic.

59

2.0

2.0

2.5

(A)
Load-drop (lbs) Load-drop (lbs) 1.5 1.5

(B)
Load-drop (lbs) 2.0 1.5 1.0 0.5 0.0 0 1 2 3 4 5 6 0 1 2 3

(C)

1.0

1.0

0.5

0.5

0.0 0 1 2 3 4 5 6 KIMAX, MPa-sqrt(m)

0.0 KIMAX, MPa-sqrt(m)

KIMAX, MPa-sqrt(m)

Figure 4.9: Load-drop vs. maximum-KI for MesaVerde (A) air, (B) water, and (C) caustic.

2.0

2.0

(A)
Load-drop (lbs) Load-drop (lbs) 1.5 1.5

(B)

(C)

1.0

1.0

0.5

0.5

0.0 0 1 2 3 4 5 6 KIMAX, MPa-sqrt(m)

0.0 0 1 2 3 4 5 6 KIMAX, MPa-sqrt(m)

Figure 4.10: Load-drop vs. maximum-KI for Granite (A) air, and (B) water.

60

4.2.2

Effect of Rock-types and Environmental Control

The analysis of the previous section with regard to load drop and subcritical index illustrated some of the variability in mechanical behavior for the different rock types and environmental conditions. Table 4.3 (a) is a reassessment of the subcritical index values only using data from the flatter portion of the n versus load drop plots (essentially excluding the data for load drops less than 0.5 lbs). Subcritical indices for all the samples in air spanned a narrow range of 64 to 81. Canyon samples had average subcritical indices 59 to 70. Sandstone A1, A2 and A3 were not very different in their composition except for slightly higher cement content in Sandstone A2 (Table 4.1(b)). This reflects higher subcritical index value for Sandstone A2 (n = 81) as compared to that of A1 (n = 64) and A3 (n = 65). Mesa Verde sample, which had predominantly calcite cement (Table 4.1(a)) as cementing material, measured an average subcritical index value of 75. Granite showed an average value of 77 for subcritical index in air. Fracture-toughness values for all the samples tested in air showed a very narrow range of 3.93 to 4.78 (Table 4.3(b)). Consistent with the results reported in the literature (Rijken, 2005; Wiederhorn, 1967), the subcritical index and the fracture toughness in air are higher than those obtained in water. Figure 4.11 shows that there is a left-ward shift of the curve on velocity-load (log-log) diagram. It has been widely reported in literature that presence of water increases the rate of reaction at the crack-tip and thereby reduces the energy required to break the bonds. Another factor, which may play a critical role in weakening of bond in sedimentary rocks, is calcite cement. Calcite cement has a higher tendency to weaken in the presence of water and therefore a higher decrease in subcritical indices. The highest reduction (61%) of subcritical indices from air to water was observed in 61

MesaVerde samples. This can be explained by the presence of high amount of calcite cement (Table 4.1(a)) in MesaVerde samples. Rijken (2005) observed that samples containing more than 15% carbonate exhibit a larger drop in subcritical index as compared to other samples in identical aqueous environments. Granite sample also shows a decrease in its subcritical indices from air to water. Rock samples showed a consistent reduction in subcritical index measured in surfactant solution as compared to that in water (Table 4.3(a)). A decrease in subcritical index is prompted by adsorption of surface-active agents on the newly created fracture surface. Adsorption of cationic surfactants on the rock-surface shall largely be controlled by the surface charge. In presence of water, at neutral pH, sandstones have a predominantly negatively charged surface (Schramm et al., 1991). The presence of negative charge on the surface increases the tendency of adsorption of cationic surfactants and thereby reduces the specific surface energy. However, the extent of reduction shall depend upon the extent to which these surface active materials get adsorbed on the surface and reduce the surface energy. Rocks are complex polyphase materials and might have different surface charges positive or, negative. These local variations in surface charges might affect adsorption of the cationic surfactants, so less adsorption shall correspond to less reduction in subcritical crack-growth. Noticeably, Mesa Verde samples didnt show any further reduction in subcritical index from water to surfactant (Table 4.3(a)). This can be explained by the presence of calcite cement in these samples (Table 4.1(b)). Calcite cement tends to have positive charge on their surfaces and shall, therefore, be expected to adsorb lower quantities of cationic surfactants. Fracture-toughness value showed a consistent reduction in strength from air to surfactant (Table 4.3(b)). A leftward shift of the curve on velocity-load diagram (Figure 4.11) indicates a decrease in strength. This observation is consistent with the Rebinders 62

theory which predicts decreased strength in presence of surface active agents. However, comparison of fracture-toughness for all the rocks for tests in water and surfactant remained inconclusive. Results in caustic show slightly higher decrease in subcritical index compared to water (Table 4.3(a)). The pH range (about 13) was within that of the chemical solutions used during hydraulic fracturing and chemical flooding (Economides and Nolte, 2000; Green and Wilhite, 1998). Irreversible chemical reaction between high pH aqueous solutions with quartz-rich sandstone samples is the predominant mechanism that alters the mechanical properties of these rock-types (Charles 1958; Scholz, 1972; Wiederhorn, 1982; Michalske and Freiman, 1982; Barnett and Kerrich, 1980). Atkinson (1979) observed that at a pH higher than 9, solubility of quartz increases. This is in contrast to the reversible effect of surfactant adsorption on rock surface. Presence of caustic didnt change subcritical indices of MesaVerde samples compared to water. It may be attributed to the fact that high-pH doesnt weaken the calcite cement as much as the quartz. The reductions in subcritical indices from air to water, air to surfactant and air to caustic correlated inversely with the clay content of the sandstones. Increase in claycontent corresponded with a smaller reduction in values of subcritical index for the specimens tested during this study. Figure 4.12 shows the data-set and the plot. Canyon I had the highest count (38) of clay lithic, and showed lowest percentage reduction for water (19%) and surfactant environments (27%). No other mineralogic parameters seemed to affect fracture results significantly other than clay and calcite cement.

63

Table 4.3(a): Mean and standard deviation of subcritical indices data for different samples.
Air Mean Canyon I Canyon II Canyon III Sandstone A1 Sandstone A2 Sandstone A3 MesaVerde Granite 64 59 70 64 81 65 75 77 Std. deviation 16 11 11 0 5 4 11 14 38 50 32 29 47 1 4 4 31 2 Mean 52 Water Std. deviation 12 Mean 47 33 41 29 40 56 34 30 35 31 28 3 6 8 9 Surfactant Std. deviation Mean Caustic Std. deviation

64

Table 4.3(b): Youngs modulus and fracture toughness of different specimens in different control environments.

Young's Modulus GPa Air Canyon I Canyon II Canyon III Sandstone A1 Sandstone A2 Sandstone A3 MesaVerde 4.67 4.56 6.28 6.21 8.96 5.52 Air 4.23 3.93 4.10 4.60 3.90 4.60 4.78

Fracture Toughness, KIC MPa m-1/2 Water Surfactant 3.96 3.39 4.60 3.98 4.44 3.98 3.40 4.07 4.25 3.35 3.80 3.49 4.05 Caustic

65

Figure 4.11: Crack-velocity vs. load for sandstone sample. Leftward shift of the curve indicates lower energy requirement for crack-growth in water and surfactant. 66

70%

Change in subcritical index from air

60%

50%

40%

Sandstone Samples
30%
Canyon-I

clay lithic
38 24 25 22 6 10

Percentage change from Air Water


19%

Surfactant
27% 44% 41%

Caustic
5% 51% 53% 57% 52%

20%

Canyon-II

10%

n(Air) - n(Water) n(Air) - n(Surfactant) n(Air) - n(Caustic)

Canyon-III Sandstone-X1 Sandstone-X2 Sandstone-X3

41% 38% 51%

55% 51%

0% 0 10 20 30 40

Clay-content (counts)

Figure 4.12: Percentage change in subcritical index from air to water, surfactant and caustic vs. clay-content.

67

Chapter 5: Velocity vs. Stress-intensity Curves


5.1 DATA NOISE REDUCTION

Crack-velocity is determined from a derivative of the load vs. time decay curve (Equation 3.13). Direct numerical computation of load-time slopes from raw data-points is generally very noisy. Red squares in Figure 5.1 represent a typical scatter in raw load data. Sharp fluctuations in raw data-set deem it completely unsuitable for numerical differentiation. Holder (2001) proposed a method which power-law fits to the whole dataset, and calculates velocity on the load-decay data obtained from this fit. However, a power-law fit sometimes obscures signal in the dataset that may be important for crack propagation interpretation. Consequently, in an effort to better explore early time crack propagation from the dual torsion tests, and to try to detect stage II subcritical behavior (see Fig. 2.2), the load-time curves were re-analyzed using spline curve smoothing before plotting velocity versus stress intensity factor curves.
5.1.1 Spline Average of Recorded Data-points

Power-law curve fits are inadequate for early time data, where the load versus time may deviate from power-law, and at late time, where the signal to noise ratio can be very low. Noise reduction was accomplished by smoothing the data using the spline method, which piecewise fits the data with polynomials of order N (degree N-1) that have continuous derivatives up to order N-2. Another advantage of spline is that the method guarantees smooth derivatives where the adjacent polynomials meet. Figure 5.1 gives a comparison between typical low-velocity noisy data and smoothed out spline data for a given DT test data-set.

68

7.40

Actual Spline
7.39 Load (lb)

7.38

7.37

7.36 100

105

110 Relative Time (sec)

115

120

Figure 5.1: Comparison between actual experimental data-points and spline data-fit. Blue diamonds represents the spline-smoothed curve and the red squares are the actual data
5.1.2 Numerical Differentiation Technique

Even with data smoothing, the calculated derivatives had significant scatter at late time because of the poor signal to noise ratio. Some additional improvement was

achieved by using a central difference scheme for the derivative (Clark and van GolfRacht 1985; Bourdet et al. 1989). It uses one point before and one point after the point of interest i, calculates the corresponding derivatives, and places their weighted mean at the point considered

69

X1 X2 t2 + t1 t2 dX t1 = t1 + t2 dt i
where X1 = Xi - Xi-1,

(5.1)

X2 = Xi+1 - Xi,
t1 = ti - ti-1, t2 = ti+1 - ti, where 1 = point before i, and 2 = point after i. The order of relative truncation error is less for these central difference schemes than for forward or backward difference schemes. The central-difference schemes were been adopted to calculate crack-velocity from load-decay data.
5.2 REGION-II CRACK-GROWTH

As discussed in Chapter 2 (section 2.1.3), subcritical crack-growth velocity in glass shows trimodal behavior (Figure 2.2). Most of the studies on rocks observe only region I behavior. However, various authors (Wiederhorn, 1968; Atkinson, 1984) have also reported region II behavior on geological materials such as sapphire, marble and micrite. In region I the velocity of the crack growth is controlled by the rate of stresscorrosion reactions at crack tips. In region II, the crack-velocity is almost independent of the stress intensity factor and is limited by transport of reactive species to the crack-tip. This is reflected as a flat plateau on velocity-KI curve (Figure 2.2). Figure 5.2b illustrates region II behavior in specimen MV3 tested in air. Figure 5.2a shows the load-decay curve of the test. As discussed earlier, spline data was fit on the noisy section of the load-decay curve. Initial load-decay data were not included into the spline-fit to avoid any artificial manifest of the velocity behavior similar to region II. Then the central difference scheme was used to differentiate the load-decay data. 70

Region-II is of the interest because of the potential of subcritical propagation of hydraulic fractures. Subcritical index for region-II are much lower than that for region-I. Therefore, subcritical crack-propagation can happen at a relatively faster rate at stressintensity factor less than its critical values. Fracture propagation rate in this region for this particular test is of the order of millimeters per second.

71

Region II

Region I

Figure 5.2: (a) (a) Load-decay curve, and (b) velocity stress-intensity factor curve for specimen MV3 (Test b). The rollover of velocity at higher KI (earlier in the load decay) represents Stage II behavior. Start of Region-III can also be observed (for first few load-decay data). 72

5.3

ACTIVATION VOLUME CALCULATION

Charles and Hillig (1962) considered activation energy as a function of tensile stress , and used Taylor expansion to express activation energy:

E * ( ) = E * (0) +

E * | = 0 + ...

(5.2)

They noted that the term E * /

has the dimensions of volume and termed it an

activation volume, V*. The Arrhenius energy or, activation energy E*(0) includes terms which express both the chemical potential difference driving the reaction as well as the energy barrier for the reacting species. The activation volume, V* term does not distinguish between these two types. Equation (2.19) (2.21) provide a basis for calculation of activation volume from a double-torsion test. The exponent obtained from an exponential fit on stage-I of the v-K diagram, along with an assumed radius of curvature of crack-tip, can be used to calculate activation volume.

d ln v b = dK I RT

(5.3)

Observations from measurements on silica glass (Bando et al. , 1984) indicate a crack tip radius of curvature of 1.5nm. Using this value, given an experimental temperature of ~25 degC (298K) and using R = 8.314 JK-1mol-1, activation volumes were computed for sandstones A1, A2, and A3 (Table 5.1). Liang and Davis (2002) reported activation volume of quartz cement as 3.1 x 10-7 m3/mol, while the results in this study were on the order of 15.0 x 10-7 m3/mol. It is evident from the results that more reactive environments cause a reduction in activation volume. The activation volume for water is 73

approximately 50% less than that for air, and similar differences were observed between air and surfactant.

1.E-02

1.E-03

Velocity, m/sec

y = 1.116E-209x 2 R = 9.887E-01
1.E-04

3.153E+01

1.E-05

1.E-06 1000000

10000000

(a)
1.E-02

KI, Pa-sqrt(m)

1.E-03

Velocity, m/sec

y = 3.013E-19e 2 R = 9.910E-01
1.E-04

1.056E-05x

1.E-05

1.E-06 2000000

2500000

3000000

3500000

(b)

KI, Pa-sqrt(m)

Figure 5.3: v-K curve for the specimen Sandstone-A1_3A (Test a) - test conducted in caustic (a) power-law fit, and (b) exponential fit. 74

Table 5.1: Activation volume and subcritical index for three different sandstone samples.

Activation volume (m3mol-1) Air Sandstone A1 Sandstone A2 Sandstone A3 1.404E-06 1.982E-06 1.266E-06 Water 9.112E-07 1.423E-06 9.014E-07 Caustic 8.982E-07 1.069E-06 8.681E-07 Air 64 77 63

Subcritical Index Water 36 50 32 Caustic 32 39 31

100

80

Subcritical Index

60

40

20

0 0.000E+00 1.000E-06 2.000E-06


3 -1

3.000E-06

Activation Volume (m mol )


Sandstone - A1 Sandstone - A2 Sandstone - A3
Air Water Caustic

Figure 5.4: Activation volume vs. subcritical index for three sandstone types.

75

Chapter 6: Fracture Roughness Analysis


In addition to measuring the fracture mechanics properties of rocks under various environmental conditions, the roughness of the created crack paths was also analyzed. The motivation was to determine whether fracture surface roughness could be linked to the micro- and macro-mechanics.
6.1 BACKGROUND

Smooth, regular geometric forms such as spheres, cylinders, smooth planes or their derivatives are described by Euclidian geometry. Euclidean geometry is defined as a geometry in which Euclid's fifth postulate if two lines are drawn which intersect a third in such a way that the sum of the inner angles on one side is less than two right angles, then the two lines inevitably must intersect each other on that side if extended far enough

holds. The shapes of many natural objects, such as mountains, coast-lines, clouds, etc.
are not regular, smooth curves or surfaces. Description of such irregular, complex shapes using Euclidian geometry is difficult. Fractal geometry introduced by Mandelbrot (1967) may allow description of such irregular shapes.
6.2 FRACTAL AND FRACTAL-DIMENSION

Fractal was defined by Mandelbrot (1982) as a set for which the HausdorffBesicovitch dimension (D) strictly exceed the topological dimension (DT). This definition, although correct and precise, is too mathematical and restrictive. For this study, another definition by Mandelbrot (1986) which is more useful has been used: a

fractal is a shape made of parts similar to the whole in some way. Barton (1995) defined
fractal geometry as a branch of mathematics that can identify and quantify how the geometry of patterns repeats from one size to another.

76

The concept of fractals can be understood by considering the existence of a family of mathematical functions that are continuous but nowhere differentiable. Linear profiles across a rough surface belong to this class of mathematical functions. The fractal dimension, D, has been used to characterize a feature having a fractal property. The fractal dimension is a fraction lying between the topological and Euclidian dimensions and describes the degree to which the fractal function fills up the Euclidian space. A linear profile across a rough surface may have a D between 1 (the topological dimension of a line) and 2 (the dimension of a Euclidian plane). Similarly, a rough surface may have a D between 2 and 3. Fractals can be either self-similar or self-affine. A self-similar fractal is a geometric feature that retains its statistical properties through various magnifications of viewing. A self-affine fractal remains statistically similar only if it is scaled differently in different directions (Kulatilake et al., 2006).

Figure 6.1: Illustration of self-similar and self-affine fractals (Kulatilake, 1995).

77

6.3

FRACTAL APPLIED TO FRACTURES

Previously, attempts have been made in different areas of earth sciences to understand temporal and spatial fractal character of geological features and processes. In the oil and gas industry, a fractal approach to characterize natural fractures has a great potential for its application in reservoir characterization. After the Mandelbrot et al. (1984) paper on the fractal character of fracture surfaces in metals, various authors have attempted to apply fractal geometry to describe rock-fracture surfaces objectively. Miller et al. (1990) studied basalt, gneiss and quartzite and proposed the use of the y-intercept of modified divider method along with fractal dimension to quantify roughness of fracture surface. Poon et al. (1992) showed that naturally fractured rocks can be classified as self-affine fractals and proposed to use fractal dimension and topothesy to characterize surface roughness. Weiss (2001) studied the link between fractal dimension of fracture surface and nominal fracture toughness.

Figure 6.2: Patterns of a Microcrack coalescence process. (a) just before fracture, and (b) a newly nucleated microcrack triggers the catastrophic failure (adapted from Lu et al., 1994.) 78

Barton (1995) used fractal geometry to characterize the outcrop scale heterogeneity of fracture network patterns over nearly 10 orders of magnitude in length scale, and to examine the relevance of acquiring fracture data from bore-holes. Lu et al. (1999) showed that the cumulative distribution of coalescence events in the vicinity of critical fracture follows a power law and the fracture profile has self-affine fractal characteristic.
6.4 MEASURING FRACTURE SURFACE ROUGHNESS

Roughness of the crack paths created during these tests was analyzed to determine whether fracture surface roughness can be linked to the micro features. Fractal analysis was used to quantitatively analyze fracture roughness of the crack-path. After the DT tests were performed, crack-paths were observed using an environmental scanning electron microscope (ESEM) utilizing backscattered electron (BSE) images. About 30-35 BSE images from a single DT specimen were collated and then transformed into digitized values. Digitized values were subsequently used to determine fractal dimension of the given crack-path. Tortuosity of the crack-path is related to the surface area created during cracking. Therefore, an increase in tortuosity corresponds to an increase in the newly-created surface area. As surface area increases, the energy required to propagate the crack also goes up. Therefore, the tortuosity of a crack-path should be related to fracture-mechanics parameters. However, tortuosity in itself is not a unique number for a crack-path and is a strong function of the length-scale by which it is measured. Fracture surfaces created in the dual torsion apparatus were characterized using the divider method described by Mandelbrot (1967, 1985). This method helps to determine tortuosity at different scale length and gives a characteristic parameter, called fractal dimension, for a given crack 79

path. In subsequent section (section 6.5), it has been shown that the fractal dimension is also a function of magnification, at which the crack-path is being mapped.
6.4.1 Crack Imaging

An environmental scanning electron microscope (ESEM) permits observations of uncoated samples at moderate to low vacuums. In the ESEM, the primary electron beam and backscattered and secondary electrons from the sample interact with gas molecules in the specimen chamber to produce positively charged gaseous ions and an amplified cascade of secondary electrons that are accelerated toward a positively biased gaseous secondary electron detector (GSED). Once the gas is ionized, it migrates toward the sample where it acts as a charge-neutralizing agent. The ESEM is normally operated at high enough gas pressures to produce sufficient positive gaseous ions to neutralize charge at the sample surface and, consequently, most ESEM images show no charging. At lower gas pressures, however, it is possible to implant a limited amount of charge in an uncoated sample (Danilatos, 1993). Water vapor is the most commonly used chamber gas, as it has a relatively low ionization potential (producing maximum signal amplification) and a high charge neutralization capacity.

Figure 6.3: Environmental Scanning Electron Microscope (ESEM) facility at JSG, UT Austin 80

A Philips XL 30 ESEM has been used in this study. This is a flexible scanning electron microscope that can be used for conventional high vacuum imaging or in the environmental mode to examine wet, oily or non-conducting samples. This is particularly suited to this study, as images can be captured without any carbon or gold coating, generally required for imaging from conventional SEM. Crack-paths were imaged by GSE (Gaseous Secondary Electron) and BSE (Backscattered Electron) detector.
6.4.2 Crack-path Digitization

Image slides captured for a given crack-path were collated into mosaics with Adobe Photoshop CS3. Manual digitization of crack-path was done using digitizing software Didger / WinDIG 2.5. A typical profile generated is shown in Figure 6.4. Adjacent data-points are connected through line-segments. The minimum feature size and the maximum feature size of a profile can be defined, respectively, as the minimum segment length and maximum segment length out of all the segment lengths between two adjacent data points on the profile (Kulatilake et al., 2006). Both the estimated maximum and the estimated minimum feature size depend upon the resolution of the image. To deal with this issue, images were captured at two resolutions, 100x and 1000x.

81

Figure 6.4: A screenshot of the profile generated from the crack-path in a sample from Canyon Sand tested in air, (a) BSE image, and (b) discrete digitized image.

82

6.4.3

Data Analysis

Data analysis was done with the original divider method (Mandelbrot, 1967; Kulatilake et al., 2006) and modified divider method (Mandelbrot, 1985). The original divider method is best visualized by considering a pair of dividers set to a particular span and then walked along the profile. The number of divider spans required to cover the entire profile is counted, and then multiplied by the divider span, r, to give an estimate of the profile length, L. The divider span is set to another value and the process is repeated several times to produce a discrete relation between r and L (Kulatilake, 2006). Feder (1988) showed that for self-similar fractals, the r and L are related linearly in loglog space according to the following equation: log L = log a + (1- D ) log r (6.1)

where log a is the intercept of the log L log r plot and the slope of the log-log plot equals 1 D in which D is the fractal dimension. For divider spans that are considerably shorter than the minimum feature resolution, the returned length L will be more or less the same. Accordingly, the log L log r curve gradually flattens, as r decreases beyond the minimum feature size. When the divider span is considerably larger, the returned length will be close to the horizontal length of the profile, and the log L log r curve gradually flattens (Figure 6.5). This behavior leads to the difficulty of obtaining a unique slope for the log L- log r relation for the whole range of r as shown in Fig. 6.3. The correct slope of log L - log r and thus the correct D can be obtained by fitting a regression line to the log L - log r data in the nonflattening portion of the curve.

83

log (LNormalized)

Suitable range for regression

log (r)

Figure 6.5: Effect of scale-length on estimation of length with the original divider method.

Figure 6.6: A screenshot of Excel VBA Tool developed for fractal analysis 84

Mandelbrot (1985) proposed modified (1-D) divider method where fractal analysis was intended to be performed at different magnifications. For each magnified profile, the divider technique can be applied to estimate the D value according to Equation 6.1. The D value is expected to increase with the magnification factor and to eventually reach a constant value asymptotically. This constant value is expected to provide the correct fractal dimension value for self-affine profiles. A subroutine was written to implement this algorithm using Excel VBA. A snapshot of the tool is shown in the Figure 6.6.
6.5 6.5.1 FRACTAL ANALYSIS Application of Modified Divider Technique

Figure 6.7 shows the log (LNormalized) vs. log (r) plots obtained for the Canyon sandstone sample (Depth 6071 ft.) for magnification factors of 100 and 1000. The calculated fractal dimension of the crack-path D is 1.0686 and 1.092 for magnification factors of 100 and 1000 respectively. There is a slight increase in fractal dimension with increase in magnification factor (Figure 6.8), which is consistent with the observation made by Mandelbrot (1985) for self-affine fractal behavior. Kulatilake (2006) showed that for fracture roughness, estimated fractal-dimension increases with increase in magnification and flattens to an asymptotic value at a magnification of 1000 (Figure 6.9). Very small increase in fractal dimension with an order of increase in magnification can be taken as an indicative of asymptotic value. This asymptotic value can be considered as the real fractal dimension of the crack-path of the sandstone sample from Canyon. Subsequent fractal analysis on Barnett shale and Yates sample (from earlier experiment by Rijken, 2005) was done with a magnification of 1000x.

85

0.25 log(LNormalized) = - 0.0686 log(r) + 0.1876 | R2 = 0.9995 0.20

log (LNormalized)

0.15

0.10

0.05

0.00

-0.05 -2 -1 0 1 2 3 4 5

log (r)
0.25

log(LNormalized) = - 0.0920 log(r) + 0.1984 | R2 = 0.9986

0.20

log (LNormalized)

0.15

0.10

0.05

0.00 -2 -1 0 1 2 3 4 5

log (r)

Figure 6.7: Variation of log (Normalized length) with log (scale) for crack-roughness data of Canyon samples from 6071 ft. depth for with the magnification factor of (a) 100, and (b) 1000; Sub-critical test carried out in an aqueous environment of pH 13. 86

1.5 1.4 1.3 D 1.2 1.1 1.0 1 10 100 1000 10000 Magnification Factor
Figure 6.8: Variation of fractal dimension with magnification factor for crack-path roughness data (Canyon samples from 6071 ft. depth). Figure 6.9: Variation of fractal dimension with magnification factor for fracture surface roughness data (adapted from Kulatilake, 2006). 87

6.5.2

Fractal Analysis Random Path

The behavior of a random path with segment lengths in a Gaussian distribution was also investigated as a part of this study. To perform this study, digitized crack-path of one of Canyon sand samples (Sample 4) was selected. A random path was generated with almost equal mean and standard deviation for the segment lengths. The skewness of the randomly defined crack-path length was however very close to zero; randomly distributed segment length being Gaussian in nature. Skewness of the two data-sets were measured by excel worksheet function skew (). Skewness is defined as follows
x x n i Skewness = (n 1)(n 2) s 3

A comparison of mean, standard deviation and skewness for the actual crack-path segment lengths and randomly generated segment lengths is presented in Table 6.1. Figure 6.10 presents a comparison of the fractal analysis results for these two crackpaths. The log (normalized length) vs. log (scale-length) graph for randomly generated crack-path is continuously curving, whereas the graph for the actual crack-path length systematically follows the power-law.

Table 6.1: Comparisons of mean, standard deviation and skewness parameters of an actual crack-path and a randomly generated crack-path with segment lengths in a Gaussian distribution.
Mean digitized crack-path randomly distributed path 18.96 19.04 Standard deviation 4.76 4.74 Skewness 0.3892 -0.0097

88

0.20

0.15

log (LNormalized)

0.10

0.05

0.00 -1 0 1 2 3 4

log (r)

Figure 6.10: A comparison of fractal analysis from the actual crack-path and randomly generated crack-path; red triangles are from actual crack-path follow power law systematically; blue diamonds are from randomly generated crack-path continuously curving. The continuously curving shape of a random path with segments lengths in a normal distribution doesnt conform to fractal behavior. This observation prompted a study to investigate whether the fractal nature of the crack-path can be used as a tool to characterize controls on crack growth during an experiment. If so, this could be extended to naturally-fractured rocks and some estimate about the possible chemical or petrographic controls during the formation of these natural-fractures could be made.

89

6.5.3

Fractal Analysis Subcritical Test Samples

Fractal analysis was carried out for five samples of Canyon sandstone tested in air (three samples) and caustic with pH 13 (two samples). Fractal dimension for the samples tested in air were of the range of D = 1.0605 to 1.0640 (magnification 100x.); for the samples tested in caustic were in the range of D = 1.0686 to 1.0763 (magnification 100x.). Fractal analysis was also carried out on a crack-path created on a Barnett shale sample with a magnification factor of 1000x. For Barnett Shale crack-path, the fractal dimension was D = 1.0450, whereas that for Canyon sandstone crack-path was 1.0920 at the same order of magnification. The lower fractal dimension conforms to the visual crack-path roughness of these two crack-paths: crack-path in Canyon sandstone is more tortuous than that in Barnett Shale. However, the subcritical index, as reported by Rijken (2005), for Barnett Shale is 289 50, which is much higher than that for Canyon Sandstone which is 62 8. This again is in conformity to the earlier observation that fractal dimension alone cannot be used to characterize sub-critical index of rock. A close look at the Figure 6.11 reveals that the points for the Canyon sandstone analysis follow the power-law more systematically than the points for Barnett-shale sample. Crack-path roughness analysis was also carried out on a crack-path on Yates. The fracture-dimension is 1.0849. The fractal dimension falls in the same range as that of Canyon sand samples. Visual impression gives the similar degree of roughness for Canyon samples and Yates sample. Fractal analysis result for Yates is shown in Figure 6.14. Table 6.2 presents a summary of subcritical indices, fractal dimension and yintercept for different specimens. Subcritical index value for Yates and Barnett shale are the average value as reported by Rijken (2005). The change in fractal dimension is very

90

small as compared to the change in subcritical indices. There is a weak correlation between subcritical indices and fractal-dimension.

Canyon

Barnett Shale

Figure 6.11: BSE images of crack-path in Canyon Sandstone and Barnett Shale, both tested in air, at magnification 1000x.

91

Figure 6.12: Variation of log (Normalized length) with log (scale) for crack-roughness data of Canyon samples from 6071 ft. depth with the magnification factor of 100x in (a) air, and (b) caustic with pH 13. 92

Figure 6.13: A comparison of fractal analysis for Canyon sand (D = 1.092) and Barnett shale (D = 1.045).

Figure 6.14: Fractal analysis of a Yates sample; fractal dimension is 1.0849, which is in the same range as Canyon sandstone; visual impression also gives the same order of roughness for the samples. 93

Table 6.2: A summary of fractal analysis results and sub-critical indices for given samples.
Sample Name
Canyon (6071 ft.) Canyon (6071 ft.) Canyon (6071 ft.) Canyon (6071 ft.) Canyon (6071 ft.) Canyon (6071 ft.) Barnett Shale Yates

Environment
Air Air Air Caustic Caustic Caustic Air Air

Fractal Analysis Magnification


100x 100x 100x 100x 100x 1000x 1000x 1000x

D
1.0605 1.0634 1.0640 1.0763 1.0686 1.0920 1.0450 1.0849

y-intercept
0.1682 0.1730 0.1677 0.2031 0.1876 0.1984 0.0764 0.1499

R2
0.9908 0.9834 0.9989 0.9881 0.9995 0.9986 0.9822 0.9957

Sub-critical Index
70.614 52.718 66.221 48.634 33.666 33.666 289 64

94

1.1000 1.0800 1.0600 1.0400 1.0200 1.0000 0 50 100 150 200 250 300

Fractal dimension

Subcritical Index

0.2500 0.2000

y-intercept

0.1500 0.1000 0.0500 0.0000 0 50 100 150 200 250 300

Subcritical Index

Figure 6.15: (a) fractal dimension vs. subcritical index, (b) y-intercept vs. subcritical index. 95

Chapter 7: Conclusions
Subcritical crack propagation experiments were performed on seven different sandstone samples and one sample of granite using the double torsion test methodology. Petrographic analysis for the 7 sandstone samples (but not the granite) were done, which consisted of 250 count point-counts on thin-sections. Most of the samples fall on the arkosic side of the Q-F-L diagram (Folk, 1980), but were higher in quartz than feldspar grains. Tests were carried out in control environments of air, de-ionized water, brine, surfactant and alkaline fluids. Cationic surfactant (Quaternary Ammonium Compound M-Quat-32) and Sodium Hydroxide (NaOH) were used as proxies. The expected reduction in fracture toughness and subcritical index going from air to water environments was observed for all the samples, a result of water being a more reactive environment. Caustic solutions reduced strength and subcritical index more than did water, but surfactant solutions performed approximately the same as water, even though they were expected to be more reactive. This variability in the effect of surfactant solution on fracture-toughness and subcritical indices was attributed to local variation in surface charge of the rock-samples, as adsorption of cationic surfactants on the rocksurface shall largely be controlled by the surface charge. The drop in fracture-toughness and subcritical index in presence of high-pH were attributed to increased chemical reaction and increased solubility at the crack-tip. The reduction in subcritical index correlated strongly with clay content of the sandstones, but no other mineralogic parameter seemed to affect fracture results significantly. This behavior has a major implication on ground-stability and alteration of flow-behavior during chemical flooding. A systematic study of variability in subcritical index values in air showed that subcritical index correlated inversely with the load drop observed during crack 96

propagation. Subcritical indices decreased as load drop increased, with stable values usually resulting at load drops of 0.5 lbs or more for the 1.5 mm thick test samples. Tests performed in more reactive environments (i.e., aqueous solutions) typically resulted in higher load drops and less variable subcritical indices. Tests carried out at load levels corresponding to fracture-toughness values gave more consistent values of subcritical indices. However, since load drop is a function of sample size, it is suggested to perform a spread of experiments to establish acceptable load-drops for new rock types and test environments. High rate data acquisition techniques were implemented to try to capture crack propagation results beyond the stage I power-law velocity versus stress intensity factor behavior. Stage II constant velocity regime was observed, but only in one sample tested in air. The spline method (piecewise polynomial fit) provided a good tool to improve noisy data in the numerical computation of crack velocities from load decay data. A central difference scheme was used for numerical differentiation of load-decay data. Activation volume values were estimated from velocity versus stress intensity factor curves using published values of crack-tip radius of curvature, giving results on the order of 15.0 x 10-7 m3/mol for sandstone tested in air, about 5 times higher than previously reported values for quartz. The values dropped about 50% when measured in aqueous environments. A strong correspondence was found between activation-volume, a thermodynamical parameter, and subcritical indices. This observation can be used to integrate thermodynamical and conventional fracture-mechanical approaches to subcritical crack-growth. Fracture roughness of experimentally created crack paths was analyzed using fractal analysis. Fractal dimensions for magnification factors of 100 and 1000 showed a slight increase with magnification, which is consistent with the observation made by 97

Mandelbrot (1985) for self-affine fractal behavior. Very small increases in fractal dimension with an order of magnitude increase in magnification was indicative of reaching asymptotic values and was considered as the representative fractal dimension of the crack-path. Using the modified divider technique, crack-path tortuosity exhibited a clear fractal nature, with results for sandstone, shale and limestone exhibiting fractal dimensions from D1.04 to D1.09. Data for the Canyon sandstone analysis followed the power-law more systematically than the points for Barnett-shale sample. On a log (normalized length) vs. log (scale-length) plot, a randomly generated crack-path was continuously curving, whereas the graph for the actual crack-path length systematically followed the power-law. The fractal dimension indicated a negative correlation with subcritical index, but the results showed a significant scatter.

98

Bibliography
Adefashe, H.A. 2006. Determining the Fracture mechanics Properties of Sedimentary Rocks using Double Torsion Testing. MS Thesis, University of Texas at Austin. Anderson, O.L. and Grew, P.C. 1977. Stress Corrosion Theory of Crack Propagation with Applications to Geophysics. Reviews of Geophysics and Space Physics 15 (1): 77-104. Atkins, A.G., Lee, C.S. and Caddell, R.M. 1975. Time-temperature Dependent Fracture Toughness of PMMA. Journal of Materials Science 10 : 1381-1393. Atkinson, B.K. 1976. Aspects of Materials Science of Interest to Tectonic Studies. Journal of the Geological Society 132: 555-562. Atkinson, B.K. 1979a. A Fracture Mechanics Study of Subcritical Tensile Cracking of Quartz in Wet Environments. Pure and Applied Geophysics 117 (5): 1011-1024. Atkinson, B.K. 1979b. Fracture Toughness of Tennessee Sandstone and Carrara Marble using the Double Torsion Testing Method. International Journal of Rock Mechanics, Mining Science & Geomechanical Abstracts 16: 49-53. Atkinson, B.K. 1980. Stress Corrosion and the Rate-Dependent Tensile Failure of a FineGrained Quartz Rock. Tectonophysics 65: 281-290. Atkinson, B.K. 1981. Subcritical Crack Propagation in Rocks: Theory, Experimental Results and Applications. Journal of Structural Geology 4 (1): 41-56. Atkinson, B.K. 1984. Subcritical Crack Growth in Geological Materials. Journal of Geophysical Research 89 (B6): 4077-4114. Atkinson, B.K. and Meredith, P.G. 1981. Stress Corrosion Cracking of Quartz: A Note on the Influence of Chemical Environment. Tectonophysics 77: T1-T11. Atkinson, B.K. and Meredith, P.G. 1987a. The Theory of Subcritical Crack Growth with Applications to Minerals and Rocks. In Fracture Mechanics of Rock, ed. B.K. Atkinson, Chap. 4, 111-166. Academic Press Geology Series. Atkinson, B.K. and Meredith, P.G. 1987b. Experimental Fracture Mechanics Data for Rocks and Minerals. In Fracture Mechanics of Rock, ed. B.K. Atkinson, Chap. 11, 477-525. Academic Press Geology Series. Bando, Y., Ito, S. and Tomozawa, M. 1984. Direct Observation of Crack Tip Geometry of SiO2 Glass by High-Resolution Electron-Microscopy. Journal of American Ceramic Society 67 (3): C36-7. 99

Barnett, R.L. and Kerrich, R. 1980. Stress Corrosion Cracking of Biotite and Feldspar. Nature 283: 185-187. Barton, C.C. 1995. Fractal Analysis of Scaling and Spatial Clustering of Fractures, In Fractals in the Earth Sciences, ed. C. C. Barton and P. R. La Pointe, Plenum Press, New York: 141 178. Beaudoin, J.J. 1985a. Effect of Humidity on Subcritical Crack Growth in Cement Paste. Cement and Concrete Research 15: 871-878. Beaudoin, J.J. 1985b. Effect of Water and Other Dielectrics on Subcritical Crack Growth in Portland cement Paste. Cement and Concrete Research 15: 988-994. Beaudoin, J.J. 1987. Subcritical Crack Growth in Low-Porosity Cement Systems. Journal of Materials Science Letters 6 (2): 197-199. Beaumont, P.W.R and Young, R.J. 1975. Failure of Brittle Polymers by Slow Crack Growth. Journal of Materials Science 10: 1334-1342. Bhatnagar, A., Hoffman, M.J., and Dauskardt, R.H. 2000. Fracture and Subcritical CrackGrowth Behavior of Y-Si-Al-O-N Glasses and Si3N4 Ceramics. Journal of American Ceramic Society 83 (3): 585-596. Bourdet, D., Ayoub, J.A. and Pirard, Y.M. 1989. Use of Pressure Derivative in Well Test Interpretation. SPE Formation Evaluation(June 1989): 293 302. Charles, R.J. 1958a. Static Fatigue of Glass I. Journal of Applied Physics 29 (11): 15491553. Charles, R.J. 1958b. Static Fatigue of Glass II. Journal of Applied Physics 29 (11): 15541560. Charles, R.J. and Hillig, W.B. 1962. The Kinetics of Glass Failure by Stress Corrosion. In Symposium on Mechanical Strength of Glass and ways of Improving It, Italy, September 25-29. Chevalier J, Saadaoui M, Olagnon C, Fantozzi G. 1996. Double-Torsion Testing a 3YTZP Ceramic. Ceramics International 22(2): 171-177. Clark, D.G. and van Golf-Racht, T.D. 1985. Pressure-derivative Approach to Transient Test Analysis: A high-permeability North Sea Reservoir Example. Journal of Petroleum Technology, November 1985: 2023 2039. Costin, L.S. 1987. Time-dependent Deformation and Failure. In Fracture Mechanics of Rock, ed. B.K. Atkinson, Chap. 5, 111-166. Academic Press Geology Series. 100

Cook, R.E. and Liniger, E.G. 1993. Kinetics of Indentation Cracking in Glass. Journal of American Ceramic Society 76 (5):1096-1105. Danilatos, G.D. (1993) Introduction to the ESEM instrument. Microscopy Research and Technique 25: 354361. Dunning, J.D., Lewis, W.L., and Dunn, D.E. 1980. Chemomechanical Weakening in the Presence of Surfactants. Journal of Geophysical Research 85 (B10): 5344-5354. Dunning, J.D., Petrovski, D., Schuyler, J. and Owens, A. 1984. The Effects of Aqueous Chemical Environments on Crack Propagation in Quartz. Journal of Geophysical Research 89(B6): 4115-4123. Dunning, J. and Miller, M. 1984/85. Effects of Pore Fluid Chemistry on Stable Sliding of Berea Sandstone. PAGEOPH 122: 447-462. Dutton, R. 1974. The Propagation of Cracks by Diffusion. In: Fracture Mechanics of Ceramics, Vol. 2: Microstructure, Materials, and Applications, eds. Bradt, R. C., Hasselman, D. P. H. & Lange, F. F. Plenum Press, New York: 647657. Economides, M.J. and Nolte, K.G. 2000. Fracturing Fluid Chemistry and Proppants. In Reservoir Stimulation, Chap. 7, 7-13. John Wiley & Sons Ltd. Evans, A.G. 1972. A Method for Evaluating the Time-Dependent Failure Characteristics of Brittle Materials-and Its Application to Polycrystalline Alumina. Journal of Materials Science 7: 1137-1146. Evans, A.G. 1974. Slow Crack Growth in Brittle Materials under Dynamic Loading Conditions. International Journal of Fracture 10 (2): 251-259. Evans, A. G., Wiederhorn. M.1974. Crack Propagation ad Failure Prediction in Silicon Nitride at Elevated Temperatures. Journal of Materiasl Science 9: 270 278. Evans, A.G. and Johnson, H. 1975. The Fracture Stress and its Dependence on Slow Crack Growth. Journal of Materials Science 10: 214-222. Feder, J. 1988. Fractals. Plenum Press, New York: 283. Freiman, S.W. 1984. Effects of Chemical Environment on Slow Crack Growth in Glasses and Ceramics. Journal of Geophysical Research 89 (B6): 4072-4076. Folk, R. L., 1980, Petrology of Sedimentary Rocks, Hemphill Publishing Company, Austin, Texas: 182.

101

Fuller, E.R. Jr. 1979. An Evaluation of Double-Torsion Testing-Analysis. In Fracture Mechanics Applied to Brittle Materials, ASTM STP 678, ed. S.W. Freiman, 19-37. American Society for Testing and Materials. Fyfe, W.S., Price, N.J. and Thompson, A.B. 1978. Fluids in the Earths Crust. Elsevier, Amsterdam: 383. Green, D.W. and Wilhite, G.P. 1998. Enhanced Oil Recovery. Society of Petroleum Engineers. 287-288. Grenet, L., 1899, Mechanical Strength of Glass, Bull. Soc. Enc. Industr. Nat. Paris, Series 5, No. 4: 838-848. Griffith, A.A. 1920. The Phenomena of Rupture and Flow in Solids. In Philosophical Transactions of the Royal Society of London, Series A, Papers of a Mathematical or Physical Character, 221: 163-198. Gurney, C. 1947. Delayed Fracture in Glass, Proc Phys Soc London 59: 169185. Gurney, C and Pearson, S. 1949. The Effect of the Surrounding Atmosphere on the Delayed Fracture of Glass. Proc. Phys. Soc. B, 62: 469476. Hamlin et al. 1995. Report of Investigations No. 232, Bureau of Economic Geology, University of Texas at Austin. Henry, J.P., Paquet, J. and Tancrez, J.P. 1977. Experimental study of crack propagation in calcite rocks, Int J Rock Mech Min Sci Geomech Abstr 14: 8591. Hillig, W. S. and Charles, R. J. 1965. Surfaces, Strain-dependent Reactions and Strength. In High Strength Materials, ed. V.F. Zachey, 682-705. John Wiley, New York. Holder, J., Olson, J.E., and Philip, Z. 2001. Experimental Determination of Subcritical Crack Growth Parameters in Sedimentary Rock. Geophysical Research Letters 28 (4): 599-602. Irwin, G.R. 1957. Analysis of Stresses and Strains near the End of a Crack Traversing a Plate. Journal of Applied Mechanics, ASME 24: 361-364. Irwin, G.R. and Kies, J.A. 1954. Critical Energy Rate Analysis of Fracture Strength. In Welding Research Supplement, April: 193-198. Irwin, G.R. and Wells, A.A. 1965. A Fracture Toughness Criterion for Concrete. Engineering Fracture Mechanics, 21 (5): 1055-1069. Ishido, T. and Mizutani, H. 1980. Relationship between Fracture Strength of Rocks and zeta-Potential. Tectonophysics, 67: 13-23. 102

Kies, J.A. and Clark, A.B.J. 1969. Fracture Propagation Rates and Times to Fail Following Proof Stress in Bulk Glass. In Fracture: Proceedings of the Second International Conference on Fracture, Brighton: 483-491. Kranz, Robert L. 1980. The Effects of Confining Pressure and Stress Difference on Static Fatigue of Granite. Journal of Geophysical Research 85 (B4): 1854-1866. Kulatilake, P.H.S.W., Shou, G., Huang, T.M. and Morgan, R.M. 1995. New Peak Shear Strength Criteria for Anisotropic Rock Joints. International Journal of Rock Mechanics and Mining Science & Geomechanics Abstracts 32: 673697. Kulatilake, P.H.S.W., Balasingam, P., Park, J and Morgan, R. 2006. Natural Rock Joint Roughness Quantification through Fractal Techniques. Geotechnical and Geological Engineering (24): 11811202 Lewis, W., and Dunn, D. 1976, Effect of Aqueous Surfactants on Crack Propagation Rates in Orthoquartzite. Geological Society of American Abstro. Programs 8 (7): 978. Lawn, B.R. 1975. An Atomistic Model of Kinetic Crack Growth in Brittle Solids. Journal of Materials Science 10: 469-480. Lawn, B.R. 1983. Physics of Fracture, Journal of the American Ceramic Society 66 (2): 83-91. Lawn, B.R., ed. 1993. Fracture of Brittle Solids. Cambridge University Press. Lewis and Karunaratne. 1981. Determination of High-Temperature Kr-V Data for Si-AlON Ceramics. In Fracture Mechanics for Ceramics, Rocks, and Concrete, ASTM STP- 745. Liang, Y. and Davis, A.M. 2002. Energetics of Multicomponent Diffusion in Molten CaO-Al2O3-SiO2. Geochimica et Cosmochimica Acta 66 (4): 635 - 646. Lockner, D. 1998. A Generalized Law for Brittle Deformation in Westerly Granite. Journal of Geophysical Research 103 (B3): 5107-5123. Lu, C., Vere-Jones, D. and Takayasu, H. 1999. Avalanche Behavior and Statistical Properties in a Microcrack Coalescence Process. Physical Review Letters 82 (2): 347 350. Mandelbrot, B. B. 1967. How Long is the Coast of Britain? Statistical Self-Similarity and Fractional Dimension. Science 156: 636638. Mandelbrot, B. B. 1982. The Fractal Geometry of Nature. W.H. Freeman, New York: 15 103

Mandelbrot, B.B., Passoja D.E. and Paullay, A.J. 1984. Fractal Character of Fracture Surfaces of Metals. Nature 308: 721722. Mandelbrot, B.B. 1985. Self-affine fractals and fractal dimension. Physica Scripta, 32: 257260. Mandelbrot, B. B. 1986. Self-affine fractal sets. In Fractals in Physics, ed. L. Pietronero and E. Tosatti, North Holland, Amsterdam: 3 28. Martin, R. J. III 1972. Time-dependent Crack Growth in Quartz and its Application to the Creep of Rocks. Journal of Geophysical Research 77: 1406-1419. Martin, R. J. III and Durham, W. B. 1975. Mechanisms of Crack Growth in Quartz. Journal of Geophysical Research 80: 4837-4844. Meredith, P.G. and Atkinson, B.K. 1983. Stress Corrosion and Acoustic Emission during Tensile Crack Propagation in Whin Sill Dolerite and Other Basic Rocks. Geophysical Journal of Royal Astronomical Society 75: 1-21. Meredith, P.G. and Atkinson, B.K. 1985. Fracture Toughness and Subcritical Crack Growth during High-Temperature Tensile Deformation of Westerly Granite and Black Gabbro, Physics of the Earth and Planetary Interiors 39: 33-51 Michalske T.A. and Freiman, S.W. 1982. A Molecular Interpretation of Stress Corrosion in Silica. Nature 295: 511-512. Miller, S.M., McWilliams, P.C. and Kerkering, J.C. 1990. Ambiguities In Estimating Fractal Dimensions of Rock Fracture Surfaces, In: Proc. 31st U.S. Symposium on Rock Mechanics, A.A. Balkema, Rotterdam, The Netherlands: 471478. Miller II, W.K., Peterson, R.E., Stevens, J.E. Lackey, C.B. and Harrison, C.W. 1994. Insitu Stress Profiling and Prediction of Hydraulic Fracture Azimuth for the West Texas Canyon Sands Formation. SPE Production and Facilities (Aug. 1994): 204. Mills, J.J. and Westwood, A.R.C. 1980. Influence of Adsorption Kinetics on Chemomechanically Enhanced Hard Rock Drilling. Journals of Materials Science15: 3010-3016. Nara, Y. and Kaneko, K. 2005. Study of Subcritical Crack Growth in Andesite Using the Double Torsion Test. International Journal of Rock Mechanics & Mining Sciences 42: 521530. Novozhilov, V.V., 1961, Theory of Elasticity, Translation of the Russian book Teoriya Uprugooti, Israel Program for Scientific Translations, Jerusalem, pp. 232.

104

Olson, J.E. 1993. Joint Pattern Development: Effects of Subcritical Crack Growth and Mechanical Crack Interaction. Journal of Geophysical Research 98 (B7): 1225112265. Olson, J.E. 2004. Predicting Fracture Swarms-The Influence of Subcritical Crack Growth and the Crack-Tip Process Zone on Joint Spacing in Rock. In The Initiation, Propagation, and Arrest of Joints and Other Fractures, ed. J.W Cosgrove and T. Engelder. Geological Society, London, Special Publications 231: 73-87. Orowan, E. 1944. The Fatigue of Glass under Stress. Nature 154 (3906): 341-343. Orowan, E. 1949. Fracture and Strength of Solids. Reports on Progress in Physics 12 (48): 185-232. Outwater J.O., Murphy, M.C., Kumble, R.G., and Berry, J.T. 1974. Double Torsion Technique as a Universal Fracture Toughness Test Method. In Fracture Toughness and Slow-Stable Cracking, ASTM STP 559, 127-138. American Society for Testing and Materials. Pabst, R.F. and Weick, J. 1981. Double Torsion Measurements With and Without a Guiding Notch. Journal of Material Science 16: 836-838. Paris, R.C. and Sih, G.C. 1965. Fracture Toughness Testing and its Applications, ASTM STP 381: 30-83. American Society for Testing and Materials. Park, N. 2006. Discrete Element Modeling of Rock Fracture Behavior: Fracture Toughness and Time-Dependent Fracture Growth, PhD Dissertation, The University of Texas at Austin. Pletka, B.J., Fuller, E.R. Jr., and Koepke, B.G., 1979. An Evaluation of Double-Torsion Testing Experimental. In Fracture Mechanics Applied to Brittle Materials, ASTM STP 678, ed. S.W. Freiman, 19-37. American Society for Testing and Materials. Pletka, B.J. and Wiederhorn, S. M. 1982. A Comparison of Failure Predictions by Strength and Fracture Mechanics Techniques. Journal of Materials Science 17: 1247-1268. Pollet, J.C. and Burns, S.J. 1977. Thermally Activated Crack Propagation Theory. International Journal of Fracture 13 (5): 667-679. Poncelet, E.F. 1965. Modern Concepts of Fracture and Flow. Technical Report: 2-65. Poulter Research Laboratory, Stanford Research Institute.

105

Poon, C.Y., Sayles, R.S. and Jones, T.A. 1992. Surface Measurement and Fractal Characterization of Naturally Fractured Rocks. J. Phys. D: Appl. Phys. 25: 12691275. Rebinder, P.A. and Shchukin, E.D. 1973. Surface Phenomena in Solids during the Course of their Deformation and Failure, Uspekhi fizicheskikh nauk 15 (5): 533-669. Rice, J.R. 1978. Thermodynamics of the Quasi-static Growth of Griffith Cracks. Journal of the Mechanics and Physics of Solids 26: 61-78. Rijken, P., 2005, Modeling Naturally Fractured Reservoirs: From Experimental Rock Mechanics to Flow Simulation, PhD Dissertation, The University of Texas at Austin. Schmidt, R.A. 1980. A Microcrack Model and Its Significance to Hydraulic Fracturing and Fracture Toughness Testing. Proceedings of 21st US Symposium on Rock Mechanics: 581-590. Scholz, C.H. 1968. Mechanism of Creep in Brittle Rocks. Journal of Geophysical Research. 73: 3295-3302. Scholz, C.H. 1972. Static Fatigue of Quartz. Journal of Geophysical Research 77 (11): 2104-2114. Schramm, L.L., Mannhardt, K. and Novosad, J.J. (1991) Electrokinetic properties of Reservoir Rock Particles. Colloids and Surfaces 55: 309 331. Segall, P. 1984. Rate-dependent Extensional Deformation resulting from Crack Growth in Rock. Journal of Geophysical Research 89 (B6): 4185-4195. Simpson, L.A. 1973. Effect of Microstructure on Measurements of Fracture Energy of Al2O3. Journal of American Ceramic Society 56: 7-11. Stevens R.N. and Dutton, R. 1971. Propagation of Griffith cracks at high temperatures by mass transport processes, Materials Science and Engineering 8 : 220234 Swanson, P.L. 1981. Subcritical Crack Propagation in Westerly Granite: An Investigation into the Double Torsion Method. International Journal of Rock Mechanics, Mining Science & Geomechanical Abstracts 18: 445-449. Swanson, P.L. 1984. Subcritical Crack Growth and Other Time- and EnvironmentDependent Behavior in Crustal Rocks. Journal of Geophysical Research 89 (B6): 4137-4152. Tada, H., Paris, P.C., and Irwin G.R. 3rd edition, 2000. The Stress Analysis of Cracks Handbook, New York: ASME Press, Professional Engineering Publication. 106

Tait, R.B., Fry, P.R. and Garrett, G.G. 1987. Review and Evaluation of the DoubleTorsion Technique for Fracture Toughness and Fatigue Testing of Brittle Materials. Experimental Mechanics: 14-22. Thomson, R.M. 1973. The Fracture Crack as an Imperfection in a Nearly Perfect Solid. Annual Review of Material Science 3: 31-51. Thomson, R., Hsieh, C., and Rana, V. 1971. Lattice Trapping of Fracture Cracks. Journal of Applied Physics 42 (8): 3154-3160. Trantina, G.G. 1977. Stress Analysis of the Double Torsion Specimen. Journal of American Ceramic Society 60: 338-341. Westwood, A.R.C. 1974. Tewksbury Lecture: Control and Application of EnvironmentSensitive Fracture Processes. Journal of Materials Science 9: 1871-1895. Wiederhorn, S. M. 1967. Influence of Water Vapor on Crack Propagation in Soda-lime Glass. Journal of American Ceramic Society 50: 407-414. Wiederhorn, S. M. 1974. Subcritical Crack Growth in Ceramics. In Fracture Mechanics of Ceramics, eds. R.C. Bradt, D.P.H. Hasselman and F.F. Lange, 2: 613-646. New York: Plenum Press. Wiederhorn, S.M. and Bolz, L.H. 1970. Stress Corrosion and Static Fatigue of Glass. Journal of the American Ceramic Society 53 (10): 543-548. Wiederhorn, S.M., Johnson, H., Diness, A.M., and Heuer, A.H. 1974. Fracture of Glass in Vacuum. Journal of American Ceramic Society 57: 336-341. Wiederhorn, S.M., Fuller E.R., and Thomson, R.M. 1980. Micromechanisms of Crack Growth in Ceramics and Glasses in Corrosive Environments. Metal Science 14 (8): 450-458. Weiss, J. 2001. Self-Affinity of Fracture Surfaces and Implications and a Possible Size Effect in Fracture Energy. International Journal of Fracture 109: 365381. Williams, D.P. and Evans, A.G. 1973. A Simple Method for Studying Slow Crack Growth. Journal of Testing and Evaluation 1 (4): 264-270. Wu, C.Cm, Freiman, S.W., Rice, R.W., and Mecholsky, J.J., 1978. Microstructural Aspects of Crack Propagation in Ceramics. Journal of Materials Science 13: 2659-2670. Zhurkov, S.N. 1965. Kinetic Concept of the Strength of Solids. International Journal of Fracture Mechanics 1: 311. 107

Vita

Raushan Kumar was born in Muzaffarpur, India, on February 6th, 1978. His mother is Bibha Sinha and his father is Suman Kumar. Raushan received his B.Tech. from the Department of Petroleum Engineering at Indian School of Mines, Dhanbad, India in 1997. He worked with Oil and Natural Gas Corporation Ltd (ONGCL), India for six years both in downstream and upstream in various positions. In Fall 2007, he started his M.S. at the University of Texas at Austin in the Department of Petroleum and Geosystems Engineering under supervision of Dr. Jon E. Olson. During his time at the University of Texas at Austin, he worked twice as a Petroleum Engineering Intern for ChevronTexaco in Houston, TX. He also is the recipient of the National Talent Search Examination (NTSE) Scholarship, Government of India, 1993.

Permanent address:

s/o Prof. Suman Kumar New Professors Colony Motihari, India 845401.

This thesis was typed by the author.

108

Das könnte Ihnen auch gefallen