Available online at www.sciencedirect.comwww.elsevier.com/locate/actamat

Acta Materialia xxx (2013) xxxxxxModeling fatigue crack growth resistance of nanocrystalline alloys

Piyas B. Chowdhury a, Huseyin Sehitoglu a,, Richard G. Rateick b, Hans J. Maier c

aDepartment of Mechanical Science and Engineering, University of Illinois at Urbana-Champaign, 1206 West Green Street, Urbana, IL 61801, USAbHoneywell Aerospace, 3520 Westmoor Street, South Bend, IN 46628, USA

c Institut fur Werkstoffkunde (Materials Science), Leibniz Universitat Hannover, D-30823 Garbsen, Germany

Received 18 November 2012; received in revised form 16 January 2013; accepted 17 January 2013Abstract

The description of fatigue crack growth in metals has remained an empirical field. To address the physical processes contributing tocrack advance a model for fatigue crack growth (FCG) has been developed utilizing a combined atomisticcontinuum approach. In par-ticular, the model addresses the important topic of the role of nanoscale coherent twin boundaries (CTB) on FCG. We make the centralobservation that FCG is governed by the dislocation glide resistance and the irreversibility of crack tip displacement, both influenced bythe presence of CTBs. The energy barriers for dislocation slip under cyclical conditions are calculated as the glide dislocation approachesa twin boundary and reacts with the CTB. The atomistically calculated energy barriers provide input to a mechanics model for disloca-tions gliding in a forward and reverse manner. This approach allows the irreversibility of displacement at the crack tip, defined as thedifference between forward and reverse flow, to be determined. The simulation results demonstrate that for both refinement of twin thick-ness and a decrease in crack tip to twin spacing FCG resistance improves, in agreement with recent experimental findings reported in theliterature. 2013 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Fatigue; Nanocrystalline; Nickel; Damage tolerance; Coherent twin1. Introduction

Current assessment of materials for damage tolerance isbased on methodologies that were developed more than40 years ago. These methodologies are empirical and rulebased, such as the well known ASME Design Code [1]that treats combined fatigue and creep damage. Today itremains a challenge to predict material degradation underfatigue loading conditions utilizing scientific principles.Compared with unidirectional deformation, fatigue intro-duces irreversibilities that are characteristic of cyclicaldeformation. These irreversibilities are a strong functionof the crystal structure, the alloy composition, and theinterface interactions. Nanocrystalline materials with twinboundaries [210] have attracted considerable attentionrecently, and possess combined strengthening attributes1359-6454/$36.00 2013 Acta Materialia Inc. Published by Elsevier Ltd. Allhttp://dx.doi.org/10.1016/j.actamat.2013.01.030

Corresponding author.E-mail address: huseyin@illinois.edu (H. Sehitoglu).

Please cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030with higher ductility. On the other hand, their fatigue dam-age tolerance characteristics have received less consider-ation, and the present paper is geared towards building aframework for the modeling of fatigue crack growth innanomaterials.

A number of studies have elucidated the strengtheningmechanisms in nanocrystalline materials under monotonicloading conditions [24,1115]. Fatigue studies of nano-crystalline metals displaying higher endurance limits [1620] compared with their coarse grained counterparts havealso been undertaken. Recent works have also demonstratedsuperior damage tolerance [5,21] in the presence of nanoscaletwins, hence the prospect of enhanced overall fatigue resis-tance with combined monotonic strength holds considerablepromise. In particular, Singh et al. [5] demonstrated thatintroducing nanotwins with a gradually diminishing lamel-lar spacing in ultrafine grained (UFG) Cu substantiallyimproved damage tolerance metrics such as the thresholdstress intensity range DKth and, most significantly, therights reserved.

gue crack growth resistance of nanocrystalline alloys. Acta Mater

2 P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxxnear-threshold crack growth rate da/dN. Moreover, studiesby Sangid et al. [21] on electro-deposited nanocrystallinenickelcobalt alloys with a high volume fraction of anneal-ing twins in the grains further corroborated the existence ofsuperior fatigue crack growth (FCG) impedance. Whilethese studies point to the significance of coherent bound-aries on FCG behavior, understanding the mechanistic ori-gin of such microstructure-driven phenomena necessitatesa detailed study informed by the underlying atomistics,capturing the operative cyclical crack tip plasticity at theappropriate length scale. The current paper has developeda model for subcritical FCG behavior combining atomisticand continuum considerations in the presence of twinlamellae of nanoscale thickness and spacing. The advan-tage of the model is that there are no adjustable parameters(fitting constants) and crack propagation occurs due to theirreversibility of plastic flow at crack tips.

A fatigue crack advances because of the irreversibleglide of dislocations emitted by the crack-tip, the degreeof which dictates the net plastic displacement per cycle[2230]. Pippan et al. [2729] showed that crack tip dis-placement under forward and reverse loading does not can-cel out because of dislocation annihilation, resulting infatigue crack advance. We note that microstructural fac-tors that would influence the degree of glide irreversibilitymust also alter the FCG rates. Specifically, microstructuralobstacles, such as coherent twin boundaries (CTBs) andgrain boundaries (GBs), in the neighborhood of an advanc-ing crack mean that the slip reversibility is difficult to ascer-tain. The extent of irreversibility imposed by theseobstacles is a function of the nature of the slipinterfaceFig. 1. Schematics representing the focus of the investigation in this paper. In funder mode III loading) which interact with a nanoscale twin. Slip-coherent twwork studied the isolated role of twin lamellae width (t) and the crack to twcoherency of the twin boundaries allows glissile motion of dislocations on thebehavior are summarized. In addition to t and d, the glide strength, s0 and the i

Please cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030interactions. At the same time, the presence of such inter-faces influences the resistance to slip propagation so (i.e.the difficulty of plastic flow advancing past the obstacle,manifested as an elevation of the unstable fault energycus). cus is the maximum fault energy during slip establishedfrom the block-like motion of an upper surface relative to alower one. Inevitably, its extrinsic (modified) level willchange due to the intersection of slip with interfaces. Theresulting crack growth rate da/dN is related to the slippaths, residual dislocations, and conservation of theBurgers vectors as influenced by the twin width andspacing.

Fig. 1 depicts the forward slip emission from an advanc-ing fatigue crack and its interaction with a CTB. The nat-ure of the slipCTB interaction is a function of the type ofincident dislocation (pure edge, pure screw or mixed).Residual dislocations with a total Burgers vector br arean outcome of these reactions, which depend on the inter-face orientation and the resolved shear stresses of theincoming and outgoing slip systems [31,32]. Variations insuch sliptwin reactions would ultimately modify the glidepath irreversibility. The fatigue crack growth resistance isexpected to change with the four factors shown in Fig. 1,the irreversibility (denoted p), the intrinsic stress so relatedto the gamma surface (Generalized Stacking Fault Energy),and the twin thickness t and twin spacing d. If the irrevers-ibility p is 0 no crack growth can occur. We show that theirreversibility is dictated by the gamma surface differentialupon forward and reverse flow at the crack tip.

The prevalence of twins, as in the case of the NiCoalloy seen in the transmission electron microscopy (TEM)orward load an advancing fatigue crack emits dislocations (pure screw typein boundary (CTB) interactions dictate the FCG mechanism. The currentin spacing (d) on the FCG behavior in a single nanotwinned grain. TheCTB, unlike incoherent GB. Factors that influence fatigue crack growth

rreversibility, p, under cyclical loading influence fatigue crack growth rates.

gue crack growth resistance of nanocrystalline alloys. Acta Mater

Fig. 2. (a) TEM images of Ni1.62 wt.% Co alloy before the fatigueexperiment. Notice the prevalence of nanotwins. (b) Post-fatigue TEMimages of Ni1.62 wt.% Co alloy. A high degree of sliptwin/GBinteraction is noticeable.

P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxx 3image of a pre-FCG experimental specimen (Fig. 2a),improves the FCG resistance to a considerable degree.Fig. 2b demonstrates an enhanced degree of dislocationpile-up at twin boundaries and GBs, indicating slip-medi-ated crack tip plasticity as the primary deformation mech-anism. Intuitively, the implied improvement in FCG [5,21]points to a special mechanism(s) involved in cyclical slipCTB interactions. Hence, one needs to establish the under-lying governing physics that decide dislocation glide, whichis susceptible to local stress sources (e.g. GB, CTB, residualsessile dislocations).

The dislocation gliding mechanism depends on thedislocation core properties. Glissile motion occurs byPlease cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030alternately rearranging atomic distortion that proceedsvia successive tearing and forming of atomic bonds sur-rounding the core structure [33]. The driving shear stressfor such motion scales with the activation energy barriersfor the translational motion onto a close-packed slip plane.The energetics of dislocation translation lie in the relativemotions of the core atoms. Alteration of the dislocationgliding condition, as influenced by nano-obstacles and therespective energetics, necessitates a non-continuum model-ing framework. In that regard, molecular dynamics (MD)allows the capture and quantification of the physics of slip-ping at the atomistic length scale. MD simulates the timeevolution of atomic nuclei (considered as classical Newto-nian particles) by integrating their equations of motion[34]. The metallic bonding is modeled through a homoge-neously distributed electron cloud functional and a pair-wise interaction potential. A semi-empirical embeddedatom method (EAM) formalism, curve fitted with experi-mental and/or ab initio material properties, employs suchmodeling to accurately describe the bonding energy land-scape [35]. Utilizing MD simulations with an EAM poten-tial Ezaz et al. [31] quantified the energetics of dislocationglide upon interaction with twins under monotonic condi-tions. In the literature some researchers [3638] haveemployed MD-EAMmethods to study massive cyclical slipemissions leading to nanovoid coalescence as the crackadvancing mechanism in the presence of GBs. However,the physics of slip irreversibility accumulation, as theunderlying incentive for crack tip plasticity, has not yetbeen explored.

In our approach we employ atomistic simulations toreveal the nature of sliptwin interaction under cyclicalconditions, and the underlying fault energy barriers. Sucha perspective reveals the exact role of CTBs as irreversibil-ity-inducing microstructural elements as well as effectivebarriers to cyclic slip. Quantification of the cyclical sliptwin reaction energetics allows the calculation of idealshear stresses for to and fro glide, as modified by the pres-ence of CTBs and/or residual dislocations. We incorporatethese atomistically extracted material properties in fracturemechanics-based formulations to simulate FCG undergo-ing large scale slip activities. The mechanics simulationsemploy cyclical irreversibility as the principal driving forceof crack advancement in the presence of nanotwins. Thecombination of two different length scale methodologiesis important, in that the continuum descriptions of FCGutilize input from the governing atomistic physics. Hencewe obtain an in-depth understanding of FCG as influencedby nanotwins. Such an insight highlights the role of somecritical characteristic dimensions associated with thesenano-obstacles (e.g. twin lamellar thickness and twin tocrack tip spacing) on the FCG metrics.

2. Methods

To develop a FCG methodology we employed bothatomistic slip-twin and fracture mechanics-based crack-tipgue crack growth resistance of nanocrystalline alloys. Acta Mater

4 P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxxslip simulations. An open source software LAMMPS(large-scale atomic/molecular massively parallel simulator)[39] was used to perform the MD simulations. A semi-infinite discrete dislocation simulation set-up was thenestablished using input from the MD results. The combina-tion of these simulations provides a convenient conduit toexplore some fundamental aspects of the mechanism ofFCG.

For the MD simulations a nickel single crystal grain wasconstructed with the crystallographic orientation shown inthe inset in Fig. 3. This grain contains a coherent twin offinite thickness. A stress concentrator (atom size void)placed in the matrix simulates a dislocation source. Thewhole system was energetically minimized, using the conju-gate gradient (CG) algorithm [34]. This resulted in an ener-getically stable single nanotwinned grain. The CGalgorithm iteratively solves atomic coordinates to reachthe minimum energy of the system within a predefined con-vergence limit. An acceptance criterion adjusts the newatomic positions, conjugate to the previous ones that fol-low the direction of steepest descent on the potential energycurve. Moreover, enforcement of three-dimensional peri-odic boundary conditions on the supercell eliminates theeffects of free surface energy, thereby simulating a systemof bulk material. The supercell dimensions were configuredsuch that the physical observables (e.g. temperature, pres-sure, kinetics and potential energy of the system) convergedto system size independence. In view of the goals of thepresent work we conducted a number of MD simulationswith varying source to twin distances as well as twin lamel-lar widths. Consequently, the supercell size was variedaccordingly to give the optimally converged dimensionsfor each simulation, avoiding any artifacts of periodicity.

Cyclical shear was applied to the supercell under straincontrol conditions. A strain range of emin = 4.46% toemax = 9.22% (i.e. Re = 0.48) was selected to facilitate slipnucleation from the void (slip source), and sufficient plasticFig. 3. Cyclical stressstrain response of a nanotwinned grain with adislocation source (not shown) in the matrix in the vicinity of the coherenttwin boundary as obtained by MD simulations. The configuration aboveproduces pure screw dislocations.

Please cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030flow to provide to and fro glissile motion across the twin.The MD simulations were run for a duration of severalhundred picoseconds. Such a timescale is inherent in MDsimulations, limited by the computational capability. Ourinvestigation required the calculation of parameters suchas the local plastic shear strain due to slip, the Burgers vec-tors thereof, and the energetics of sliptwin reactions.These parameters are unaffected by the high deformationrates arising from such a timescale. In order to conductnon-equilibrium MD simulations (i.e. evolution of the sys-tem under the imposed conditions) we employed an iso-baricisothermal (NPT) ensemble along with a NoseHoover thermostat algorithm. Hence, the total number ofatoms N, the external pressure P, and the temperature T(at 10 K) of the system were held constant. The dynamicsof deformation proceeded utilizing the velocity Verlet algo-rithm as the time integrator. Atomistic snapshots at differ-ent time points were carefully analyzed using visualmolecular dynamics (VMD) [40] and AtomEye configura-tion viewer [41]. These visualization tools, combined within-house MATLAB programs, helped capture the detailsof sliptwin interactions (e.g. the conservation of Burgersvectors) and calculate fault energies, glide distance of slip,etc. Volume-averaged virial stress formulation, neglectingthe kinetic energy contribution [42], was employed in orderto quantitatively assess the stressstrain response of thesystem.

One essential part of our investigation was to calculatethe energetics of complex cyclical sliptwin reactions,necessitating accurate descriptions of the atomic levelenergy landscape through EAM formulations [35]. A com-parative study of EAM potentials available in the literaturedemonstrated that the Foiles and Hoyt potential [43] pro-vides good agreement between the unstable fault energycus, the density functional theory (DFT) calculations(254 mJ m2), and the intrinsic stacking fault energy (cisf)with the experimental finding (127 mJ m2). The interpla-nar potential energy profiling incorporating all of theseparameters is termed the generalized stacking fault energy(GSFE) curve [44], as shown in Fig. 6. The GSFE repre-sents the energy pathway to create the lattice distortionof a dislocation along the Burgers vector direction. A typ-ical GSFE is calculated by sliding one crystalline half-spaceon top of another on the slip plane along the slip direction.We utilized the Foiles and Hoyt EAM potential to com-pute the modified GSFE, as influenced by local stressesduring back and forth dislocation glide traversing the twin(to be discussed in detail later). For a more thoroughdescription of the MD simulation procedures employedin the present work readers are referred to Ezaz et al. [31].

3. Results

3.1. Molecular dynamics simulations

Fig. 3 shows a typical MD-based cyclical shear stressstrain response of a nanotwinned grain with a dislocationgue crack growth resistance of nanocrystalline alloys. Acta Mater

Fig. 4. (a) Steady-state cyclical sliptwin interaction for the forward part of a MD fatigue cycle (for visualization convenience perfect lattice atoms arerendered invisible and only defect atoms are shown). The dissociated leading and trailing partials emitted from the source (not shown) are approaching theclosest CTB. All vectors (in the matrix and/or twin) are represented in their respective coordinate frames. (b) Two partials recombining upon interactingwith a CTB. The emerging leading Shockley partial (pink) and incident trailing Shockley partial (brown) are shown. The situation shown depicts themetastable phase of the sliptwin reactions. More dislocations subsequently nucleate. (c) Nucleation of multiple dislocations as a result of the incidentdislocationCTB interaction. (d) A simplified double Thompson tetrahedron depiction of the dislocation reactions during forward flow at a CTB depictingconservation of the Burgers vectors. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of thisarticle.)

P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxx 5source located in the matrix. Cyclical deformation isapplied to the extent that it facilitates dislocation nucle-ation, and with a sufficient degree of slip to intersect thetwin, located at a distance d from the source. The strainrange for subsequent cyclical loading is set up such thatto and fro dislocation motions occur across the width tof the twin. Separate MD simulations were carried out withvarying finite twin lamellar widths and crack to twin spac-ings to study the influences of these dimensions. With agradual increase in t or d the external loading needs to beincreased in order for the slip to reach and traverse theentire width of the nanotwin. Consequently, a graduallygreater number of dislocations nucleate with the increasein the applied load. The multitudes of dislocations undergorelatively more complex forms of interactions with theCTBs at larger t and/or d. Nevertheless, we observed a gen-eralized pattern of cyclical sliptwin interaction with simul-taneous incorporation and transmission of slip for all casesPlease cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030of varying t and/or d. Hence the fundamental similaritiesreside in the type of interaction and the introduction ofirreversible slip activity in each cycle, irrespective of thenumber of incident dislocations. In the following sectionthe cyclical sliptwin reaction involving only one incidentdislocation (pure screw type) is described in detail for thecase of smaller t and d (requiring a lower applied load).

As can be seen in Fig. 3, the stressstrain approaches asaturated response as the cyclical sliptwin interactionmechanism also achieves a recurrent steady-state. Sincethe MD simulations were performed on pristine crystals,and at high deformation rates, the stresses from the MDare high compared with the experimental stressstrainresponse. However, the dislocation reactions associatedwith the sliptwin interactions are unaffected, as verifiedby running simulations at different strain rates. Therefore,the cyclical stressstrain plots in Fig. 3 are interpreted onlyto obtain a quantitative estimation of the quasi-steady-stategue crack growth resistance of nanocrystalline alloys. Acta Mater

Fig. 5. (a) Schematic showing calculation of the slip irreversibility p. c represents shear strains due to dislocation glide. (b) Slip irreversibility calculatedfrom MD simulations as a function of crack tip to twin spacing t with constant d = 80 nm. (c) Slip irreversibility calculated from MD simulations as afunction of the crack tip to twin spacing d with constant t = 80 nm.

Fig. 6. Generalized stacking fault energy (GSFE) utilized in our study forpure nickel, calculated using the EAM potential developed by Foyles andHoyt [43].

6 P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxxresponse of the system under investigation. The next sec-tion describes a detailed study of the CTBdislocationinteractions to clarify the reaction type.Please cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.0303.1.1. Cyclical sliptwin interactions

MD simulations revealed the exact nature of the steady-state cyclical sliptwin reactions. During forward loading,after elastic straining, a perfect screw dislocation of Bur-gers vector a

21 1 0 nucleates from the source on the most

favorable slip system with the maximum resolved shearstress. Immediately after emission the perfect dislocationdissociates into two Shockley partials (leading and trailing)separated by a ribbon of stacking fault (Eq. (1)).

a21 1 0Full

! a62 1 1Leading

a61 2 1Trailing

Stacking Fault 1

Under continued application of forward shear loadingthe extended dislocation (with a leading and a trailingShockley partial) glides towards the CTB (Fig. 4a). Beingobstructed by the CTB, the two partials recombine andgenerate new screw dislocations at the site of incidence(Fig. 4b). The new dislocations similarly dissociate intoShockley partials. One of the new dislocations is incorpo-rated into the CTB as twinning partials, and another(extended) is transmitted inside the twin (Fig. 4c). Eq. (2)summarizes this reaction. Transmitted partials a

62 1 1T

and a61 1 2T become a6 2 1 1T and a6 1 1 2T, respectively,gue crack growth resistance of nanocrystalline alloys. Acta Mater

P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxx 7in the matrix frame. The total Burgers vector is then con-served on both sides of Eq. (2). Fig. 4d provides a doubleThompsons tetrahedron depiction of the forward sliptwin reaction.

2

Until the end of the forward loading cycle the twinningpartials on the CTB carry on gliding in opposite directions,gradually increasing their separation distance in order tominimize the elastic strain energy. Because of the glidingof these twinning partials the twin boundary migratesone atomic layer. The transmitted partials inside the twincontinue to glide. As the load is reversed (unloading) thepreviously transmitted leading and trailing partials (insidethe twin) now reverse their directions of motion (upon elas-tic relaxation). The returning extended dislocation interactswith the CTB and undergoes similar multiplication,thereby creating two new twinning partials on the CTBand two Shockley partials in the matrix. Eq. (3) describesthe reverse reaction. The total Burgers vector (upon con-version of twin dislocations to the matrix frame) on bothsides is conserved.3Table 1Summary of cyclical steady-state sliptwin interaction.

Schmid factors (sRSS=s) Sliptwin interaction(s) Inc

Incident CTB Outgoing bs

0.778 1.0 0.778 Transmission, incorporation a2 1

Please cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030With further unloading the twinning partials continue toglide in opposite directions, eventually causing twin migra-tion by one atomic layer. The twin migration process mayinvolve growth or shrinkage of the twin depending on thedirection of motion of the participant twinning partials.The matrix Shockley partials continue gliding towardsthe source until a full dislocation (which also dissociatesto become extended) of opposite sign nucleates from thesource, meets with the returning one, and they annihilateeach other. At the end of unloading another new negativedislocation (considering the original nucleated slip, at thebeginning of forward loading, to be of positive type) nucle-ates and glides towards the twin, repeating the mechanismover subsequent cycles.

Table 1 summarizes the sliptwin reactions. Here bs, beand br refer to the screw, edge component, and residual dis-location on the CTB, respectively. In summary, the reac-tion process involves transmission of unobstructed slippast the CTB (designated outgoing), and incorporation ofslip with br in the CTB. The full dislocations are of purescrew type (which dissociate into partials) for both the inci-dent and outgoing systems. For all the active slip systemsthe resolved shear stress sRSS under global applied stress sis calculated using the formulation:

sRSS rijminj 4In Eq. (4) rij is the remote stress tensor, mi the slip plane

normal vector, and ni the vector representing the slip direc-tion. For our case the applied rij is reduced to s13 (s inFig. 3). As can be seen in Table 1, the magnitudes of theratio of sRSS to s, defined as the Schmid factor (SF), forthe active slip systems are fairly high. The maximum SFis operative on the CTB, which facilitates the incorporationof glissile twinning partials. The next largest SF acting onthe outgoing slip system inside the twin assists in the trans-mission of slip past the CTB. The analyses, as summarizedin Table 1 and Fig. 4, concern forward flow of slip past theCTB. However, the reverse reaction is modified only in theform of enhanced resistance (due to the presence of twin-ning partials in close vicinity) at the CTB. The reverse inci-dence of slip upon the CTB results in similar interactionproducts from dislocations at the reaction site. For morecomplex cases involving a multitude of incident disloca-tions (at larger t and/or d with a higher applied load) theinteraction type remains fundamentally identical, undergo-ing simultaneous incorporation and transmission of slip.The subsequent sections address quantification of the slipirreversibility and the origin of the discrepancy of forwardvs. reverse slip resistance at the CTB.ident slip (matrix) Outgoing slip (twin) Residual slip (CTB)

be bs be br

1 0 0 a2 1 1 0T 0 a2 0 1 1

gue crack growth resistance of nanocrystalline alloys. Acta Mater

8 P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxx3.1.2. Cyclical slip irreversibilities

Fig. 5a shows a simple example of how irreversible dis-location glide during cyclical loading can be quantified,considering the case of a single incident dislocation (atsmaller t or d). Steady-state cyclical sliptwin interactioninvolves the incidence of a dissociated full dislocation(screw) on the CTB. The reaction results in the simulta-neous incorporation and transmission of extended disloca-tions. The final locations of these dislocations are atpositions c and e (partials) and d (extended full disloca-tion, shown as full for simplicity) at the end of forwardloading. The partials at c and e contribute to migrationof the twin by one atomic layer. During reverse flow theextended dislocation at position d is transmitted back intothe matrix, again leaving new partials at c0 and e0, whichrepulse the partials at c and e. The returning crack-boundextended dislocation is annihilated by another incomingdislocation of opposite sign (negative) at location f. Wecalculated the ratio p between the irreversible plastic shearstrain cirr and the total plastic shear strain ctotal generatedby dislocation glide over a cycle using Eq. (5). These c val-ues represent shear strains due to dislocation glide as theyappear in Eq. (5). Even though Fig. 5a shows the cyclicalprocess involving only one incident dislocation, on gradu-ally increasing t and/or d a number of dislocations willcontribute to the overall irreversible phenomena in anidentical manner. The parameter p is then computed asa varying function of t and d. Fig. 5b and c demonstratesthe results.

p cirrctotal

cce cc0e0 cafcad cdb cce cc0e0 cbf

5

The trend for cyclical crack tip slip irreversibilities p, ascalculated in Fig. 5b and c, tends to become independentof t and d at sufficiently large magnitudes. At lower valuesof t and d a relatively small number of dislocations areemitted from the source and traverse the entire thicknessof the twin. Consequently, a shorter spacing between thesource and the twin as well as thinner twins will expeditethe return of source-bound positive slip, and at the sametime preclude gliding of negative slip sufficiently fartheraway from the source. Thus the annihilation process(location f in Fig. 5a) occurs in very close proximity tothe source (as in the insets in Fig. 5b and c). As a result,the magnitude of p is low in the small t and/or d regime,as calculated with Eq. (5). However, an increase in t or dnecessitates a larger applied load in order for slip to occurand traverse the twin. This results in a gradually greaternumber of dislocation emissions. The involvement of mul-titudes of dislocations leads to an increase in the pile-upstress during both forward and reverse flow. This resultsin even greater blockage of returning positive dislocations,thereby permitting unobstructed negative slip to travelfurther away from the source. As a result, the annihilationprocess occurs at a greater distance from the source (asindicated in the insets). Due to the involvement ofPlease cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030large-scale slip activity for even higher t or d the annihi-lation point eventually settles at a stable unvarying posi-tion, corresponding to the plateau region in Fig. 5b and c.

The underlying origin of glide irreversibilities as influ-enced by CTBs can be traced back to the discrepancies inthe energy pathways for slip for forward and reverse trans-mission across CTBs. Such a phenomenon modifies theresistance of slip penetrating the CTBs under cyclical con-ditions, as further discussed below.

3.1.3. Energy barrier and ideal glide strength

The potential energy variationdisplacement relation-ship of a pair of partial dislocations (an extended full dis-location separated by a stacking fault) in an otherwiseperfect fcc crystal is described by the GSFE, as in Fig. 6.During glissile slip motion sliding of atomic planes occursby overcoming the unstable fault energy cus (inset inFig. 6). The unmodified GSFE curve points to the resis-tance of dislocation glide scaling with cus, as imposed bythe crystal. A Shockley partial dislocation glides on the(1 1 1) plane with the Burgers vector along the h1 1 2i direc-tion. The modified GSFE takes the form of an increase incus due to the presence of the CTB (or any other local stresssources), as shown in Fig. 7a. In order to compare the rel-ative resistance encountered during forward and reverseflow the modified cus in the vicinity of a CTB is calculated(Fig. 7c).

With a view to estimating the resistance stress providedby a CTB to the approaching slip for back and forth trans-mission we calculated the modified GSFE in the vicinity ofthe CTB using a dynamic approach. Considering thedynamic nature of dislocation glide over time, computingthe variation in the potential energy difference in somepreselected atoms allows quantification of the modifiedGSFE (discussed in detail in Appendix A). The cus valuesderived from these modified energy curves are plotted asa function of distance normal to the CTB in Fig. 7c. InFig. 7b the dislocation at A is approaching the CTB butis still unaffected by the stresses resulting from thematrixtwin interfacial atomic mismatch. Thus cus at Adenotes the energy barrier that a dislocation has to over-come when it is gliding freely inside the crystal. The magni-tude of cus at A matches the peak in Fig. 6, whichrepresents the energy barrier to unobstructed gliding,amounting to 254 mJ m2. The local stress generated dueto atomic mismatch at and around the CTB elevates cusonce the approaching dislocation is in closer proximity.The maximum energy barrier that the incident dislocationneeds to overcome is achieved when the slip interacts withthe CTB, an intermediate step in formation of the finalreaction products (Fig. 7b and c point B). The elevatedenergy at point B corresponds to a cus value as high as340 mJ m2. Therefore the energy path A! B ! C (redcurve) describes the variation in cus for transmitted disloca-tions during forward flow. In the course of reverse flow thereturning dislocation encounters an even greater energybarrier due to the presence of the dissociated twinninggue crack growth resistance of nanocrystalline alloys. Acta Mater

Fig. 7. (a) Schematic demonstrating the expected increase in energy barrier (cus) due to the presence of a local stress source. The dark line depicts theplanar fault energy for dislocation glide through a perfect crystal, while the green line represents the enhanced energy encountered in the presence of aCTB. The maximum slope of the (un)modified GSFE equals the ideal shear stress of the crystal smax. Ezaz et al. [31] extensively explored the contributionof local stresses to the fault energetics of sliptwin interactions. (b) As a pair of Shockley partials approaches the CTB the energy barrier (cus) is elevated inthe neighborhood of the twinmatrix interface. cus is maximum at the CTB (position B). At C the cus is the same as at A. Upon interaction with a CTB apair of Shockley partials is left on the CTB, while another pair transmits into the twin. As the transmitted pair glides away from the interface cus decreasesto the level of a perfect lattice barrier. During the reverse transmission upon flipping of loading the returning dislocation encounters enhanced cus due tothe presence of the Shockley partials on the CTB (at point D). As the returning extended dislocation is transmitted back into the matrix it undergoes asimilar multiplication, leaving another pair of twinning partials on the CTB. Schematic of load cycles in MD simulations (strain control) showing where A,B, C, D and A0 occur. (c) Change in unstable energy cus as a pair of partial dislocations approach a CTB during forward/reverse loading in the MD fatiguecycle. Path A! B! C provides an energy barrier against forward transmission (red curve), and C !D ! A0 against reverse transmission. The energy atD is greater than at B because of the presence of dissociated Shockley partials on the CTB during reverse loading. (For interpretation of the references tocolor in this figure legend, the reader is referred to the web version of this article.)

P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxx 9

Please cite this article in press as: Chowdhury PB et al. Modeling fatigue crack growth resistance of nanocrystalline alloys. Acta Mater(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030

Fig. 8. The set-up for dislocation dynamics simulations. A mode III crackemits a series of screw dislocations that glide away to interact with ananotwin of finite lamellar width placed at a finite distance. Positivedislocations assume equilibrium positions xfn and x

rn during forward and

reverse loading, respectively. Negative dislocations nucleate during thereverse half cycle, and eventually annihilate returning positivedislocations.

10 P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxxpartials on the CTB. Therefore the reverse energy barrierfollows the path C! D ! A0 (blue curve). The cus maxi-mum reaches a magnitude of 452 mJ m2 at point D. Theelevation of the reverse transmission energy barrier com-pared with the forward barrier can be attributed to theincrease in local stress around the CTB due to the residualtwinning partials. The GSFE for slip glide, as modified inthe above mentioned manner, facilitates calculation ofthe ideal shear strength of the crystal smax. smax is a func-tion of cus, and is calculated from the maximum slope ofthe modified or unmodified GSFE curve (Eq. (6)). Apply-ing corrections for thermal activation and strain rate tothe plastic flow the ideal critical glide strength so at roomtemperature and a typical experimental strain rate can becalculated. Eq. (7) implies that so is also a function of strainrate _e, temperature (T), and activation volume (V). Theprocedure for obtaining so is detailed in Appendix B.

smax smaxcus @c@x

max

6

so sosmax; _e; T ; V 7Below we investigate continuum slip emissions from a

fatigue crack whose glide paths become irreversible uponcyclical loading. Atomistically computed so values are thenutilized to characterize the continuum level dislocationglide, and subsequent fracture mechanics simulations.

3.2. Continuum dislocation simulations

We modeled a pre-existing mode III fatigue crack in thepresence of a nanotwin. The crack emits a series of screwdislocations. These dislocations intersect the twin and,eventually, their cyclical glide paths become irreversiblevia annihilation. The slip glide resistance (due to latticefriction and penetrating twins) influences the equilibriumpositions and the total glide path irreversibility. Cracksadvance by accumulating plastic displacement at the tip,originating from the irreversibility of cyclical slip. For agiven crack length (a), twin thickness (t) and twin positionfrom the crack tip (d) and stress intensity levels (DKIII) wecan predict the corresponding values of da/dN. da/dN isexpressed as a function of the equilibrium positions of dis-crete dislocations at a certain DKIII.

3.2.1. Fracture mechanics calculationsIn the continuum model we selected a mode III fatigue

crack and the associated emission of pure screw disloca-tions (Fig. 8). The MD derived glide strengths for a screwdislocation were utilized. The continuum dislocations canovercome the glide resistances under the applied externalload, and eventually assume equilibrium positions. At theequilibrium position a dislocation emitted from a cracktip experiences three forces: (1) resolved applied shearstress sApplied; (2) image stress sImage; (3) pile-up stresssPile-up. With the nucleation of new dislocations the localstress at the crack tip decreases due to enhanced imageand pile-up stress. In order to compensate for this decreasePlease cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030the applied load has to be increased to facilitate furthernucleation. Thus Eq. (8) summarizes the net shear stresssn acting on the nth dislocation.

sn sApplied sImage sPile-up 8sn is formulated in Eq. (9).

sn KIIIffiffiffiffiffiffiffiffiffi2pxn

pApplied

lb4pxnImage

lb2pxn

Xin

ffiffiffiffiffiffiffiffiffiffiffixixn

s1

xi xnPile-up

9

Eq. (9) gives sn as a function of KIII, the applied globalstress intensity factor for mode III loading, l, the shearmodulus, b, the Burgers vector, and xn, the location ofthe nth dislocation along its glide path from the source(crack tip). sn ought to be of sufficiently large magnitudein order for slip to overcome the unstable energy barrier(cus). Therefore with increasing global applied loading snneeds to surpass and/or equal so to initiate glide. ThusEq. (10) provides the conditions for gliding, which is fur-ther rearranged to formulate Eq. (11).

sn P so 10

KIIIffiffiffiffiffiffiffiffiffi2pxn

p lb4pxn

lb2pxn

Xin

ffiffiffiffiffiffiffiffiffiffiffixixn

s1

xi xn so 0 11

Eq. (11) provides the equilibrium conditions for disloca-tions. The final equilibrium positions (xi) of all dislocationsduring both forward and reverse loading (at maximumKIII) were solved from Eq. (11). These xi values were uti-lized as the input for a FCG rate formulation (discussedlater in Eq. (15)).

In order to clarify the procedure for the mechanics-based simulations let us consider a simple case consistingof a very low applied DKIII such that cyclical crack tip plas-ticity involves only one discrete dislocation (designated 1 ingue crack growth resistance of nanocrystalline alloys. Acta Mater

Fig. 9. A single dislocation demonstrating the slip trajectory during da/dN calculations. A mode III crack emits a screw dislocation (designated 1) duringforward loading which assumes an equilibrium position at maximum forward load (trajectory shown in red). During reverse loading (blue trajectory), afterelastic relaxation the dislocation starts to return, and is annihilated by a newly nucleated dislocation of opposite sign (designated 2). (For interpretation ofthe references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 10. Irreversibility of crack tip emitted dislocation activities for threecases of finite nanotwin lamellar spacings.

P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxx 11Fig. 9) with no obstacle (twin) in the glide path. At a cer-tain time point in the loading cycle the forward xf1 andreverse xr1 equilibrium positions of the disloction aresolved by setting the lattice friction stress equal to theapplied resolved shear stress. In the forward half-cycle asthe applied load is increased the dislocation continues toglide away until the maximum KIII is reached. Fig. 9 dem-onstrates the trajectory for the case of a single dislocationduring forward/reverse loading. In forward loading a dislo-cation nucleates at a critical KIII value and then glides awayto assume its final position (red curve). During reverseloading the dislocation does not immediately start to returntowards the crack tip because of elastic strain recovery. Asthe shear stress in the reverse direction exceeds the latticefriction resistance it starts to glide towards the crack tipand eventually is annihilated by a newly nucleated negativedislocation. Continued reverse loading triggers the nucle-ation of another negative dislocation which repeats themechanism over another cycle. This simplistic demonstra-tion of the irreversiblity of a discrete dislocation glide pathover a fatigue cycle elucidates the fundamental procedureof the continuum-based simulations.

The introduction of nanoscale twins on the glide path ofslip modifies the total irreversibility as well as slip obstruc-tion by the crack. Eq. (12) provides an evaluation of theslip irreversibility parameter p (previously defined as theration between cirr and ctotal) as a function of the disloca-tion positions at equilibrium during forward and reverseflow (denoted by the superscripts f and r, respectively).

p cirrctotal

Xni1

xri2xfi xri

12

Fig. 10 shows the evolution of p as evaluated with thespecified values of t at constant d, with a change in theapplied stress intensity factor range. p increases non-Please cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030linearly (square root trend) with DKIII, and eventuallyachieves a plateau. The computed tendency of p highlightsthe functional dependence of irreversible glide phenomena,and hence the crack tip plasticity, on the variation in twinthickness t on the nanoscale. This is consistent with theMD calculations elucidated earlier in Fig. 5b and c. Inresponse to changes in the twin lamellar width the equilib-rium positions of dislocations change accordingly. Thiswould lead to different degrees of irreversibility in the over-all dislocation glide paths, as implied through Eq. (12).These results points to a change in FCG rate as a functionof t or d. In order to further explore the t and/or d depen-dence of FCG a comparison of da/dN under such condi-tions was evaluated.gue crack growth resistance of nanocrystalline alloys. Acta Mater

Fig. 11. Determination of the critical matrixtwin interface zone size (2q).

Fig. 12. (b) da/dN vs. DKIII plots demonstrating the influence of (a) twinwidth and (b) crack to twin distance on FCG and the threshold properties.

12 P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxx3.2.2. FCG simulations

A cyclical crack accumulates plastic displacement on anincremental basis. If there are n dislocations emitted fromthe crack tip the plastic displacement at the tip caused byeach emission contributes to the overall crack extension.Based on the formalisms introduced earlier the rate ofcrack tip advancement per cycle can be formulated as:

dadN

Z xfmax0

du 13

dadN

xfmax

2l

Xni1

sfn Dsn 14

In Eq. (13) xfmax is the maximum distance away from thecrack during forward loading traveled by the farthest dislo-cation, u is the crack tip displacement as a function of xiand l is the shear modulus in the slip direction, sfn is theshear stress at the end of the forward half-cycle (a functionof xfi ), and Dsn is associated with the distance (xi) traveledby the returning crack bound dislocations (a function ofxfi xri ). Combination of Eqs. (9) and (14) leads to theda/dN formulation given in Eq. (15). In this formalismda/dN is essentially a function of the equilibrium disloca-tion positions. The solutions for these slip locations (xi)at the maximum KIII, as obtained from the equilibriumcondition (Eq. (11)) during forward/reverse flow, provideinput to the da/dN evaluation. Calculation of da/dN in thismanner inherently incorporates the atomistically computedideal glide stress as well as the twin penetration strength.

dadN

xfmaxDKIII2l

ffiffiffiffiffiffi2p

pXni1

1ffiffiffiffixfi

p 1ffiffiffiffiffiffiffiffiffiffiffiffiffiffixfi xri

p !

xfmaxb8p

Xni1

1

xfi 1xfi xri

xfmaxb4p

Xni1

Xji

ffiffiffiffiffiffiffiffiffiffiffixfjxfi

s1

xfj xfi

0@

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffixfj xrjxfi xri

s1

xfj xrj

xfi xri

1A 15

We employed the da/dN formulations from Eq. (15) toquantitatively investigate the sensitivity of the slip blockingstrength of nanotwins against FCG. Fig. 11 demonstratesthat for a twin placed at a large distance d from the cracktip (of the order of 1000 the Burgers vector) FCGbecomes totally independent of the resistance of the inter-face to the interacting slip. Thus such a calculation pro-vides information regarding the critical zone of influenceof the CTB. As shown in Fig. 11, the critical interface influ-ence size resides in the plateaux regions for which da/dN isindependent of the slip-blocking strength at small d values.The CTB influence zone ranges across a few Burgers vectorvalues, consistent with the MD findings (Fig. 7c). Determi-nation of the neutral CTB influence size assists in the calcu-lation of FCG as a function of t and d.Please cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030Fig. 12a and b shows how the FCG metrics change withvariations in t and d, as predicted earlier in the evaluationof cyclical slip irreversibilities in Fig. 10. These plots of da/dN vs. the applied DKIII reveal a t and d dependence of therate of FCG as well as the threshold behavior. For smallert and/or d FCG is significantly influenced by the nanotwingue crack growth resistance of nanocrystalline alloys. Acta Mater

P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxx 13dimensions. Thinner twins as well as twins placed veryclose to the crack tip lead to enhanced FCG metrics.4. Discussion

4.1. Cyclical sliptwin interactions and irreversibilities

Earlier works have pointed out the role of the localstress state and the type of dislocation (screw, edge ormixed) in dictating the nature of dislocationCTB interac-tions [31,45]. In the present work we have examined thecyclical sliptwin interaction mechanism for fatigue cracktip-nucleated dislocations. Table 1 summarizes theobserved dislocationCTB reactions during forward flow.Our case consists of pure screw incident dislocations thatglide and intersect with the CTB. In the literature the inci-dence of a screw dislocation upon a coherent twin report-edly activates one of two mechanisms [7,46]: (i)incorporation (absorption) of the incident dislocation fol-lowed by dissociation into Shockley partials on the CTB;(ii) direct transmission of the incident dislocation throughthe twin. We observed that the incident dislocation causesnucleation of dislocations at the reaction site, as describedin Fig. 4c. As listed in Table 1, the SFs on the CTB andoutgoing system are 1.0 and 0.778, respectively, which aresufficiently high to facilitate the reaction observed. Thisreaction consists of simultaneous incorporation and trans-mission of glissile dislocations on and across the CTB, asshown in Fig. 4c. As clarified in the recent literature, block-ing of slip by CTBs (incorporation) promotes to a consid-erable extent macroscopic ductility, by allowing dislocationglide along the interface, unlike incoherent GBs [7]. Thisspecial feature of CTBs, along with permitting transmis-sion, results in an enhancement of the macroscopicmechanical properties. Sangid et al. [21] ascribed the bal-ance of macroscopic strength and ductility of nanotwinnedmaterials to the superior FCG metrics. As per our observa-tions, the types of sliptwin interactions which promoteboth strength and ductility on the micro-scale are operativeduring to and fro glissile motions of slip across nanotwins.

Farkas et al. [36], Nishimura and Miyazaki [37] and Pot-irniche et al. [38] studied the mechanism of fatigue crackadvancement through the nucleation and coalescence ofnanovoids formed due to crack tip slip activity by MD.While these simulation studies addressed the FCG mecha-nism experimentally observed in especially the smaller sized(

14 P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxxCTB. The enhanced energy barrier during reverse flowowing to this residual slip on the CTB would pose a greaterdegree of obstruction to reversing dislocations. The presentwork quantifies the degree of glissile irreversibility asrelated to the characteristic microstructure dimensions, aswell as captures the governing physics in terms of theunderlying energetics. Even though the present study islimited to discussions of cases concerning only screw dislo-cations, similar physics are expected for edge and mixeddislocations. Pure edge or mixed dislocations leave residualdislocations on the CTB upon interacting with coherenttwins, as examined earlier in the literature [6,7,31,49].Despite the varied nature of sliptwin reactions from caseto case, the enhanced resistance encountered by reversingslip and their subsequent annihilation by dislocations ofopposite sign would occur in a similar generalized patternfor mixed or pure edge cases. Hence the overall irreversibleglide path pattern would essentially be identical with simi-lar trends in FCG characteristics.

4.2. Role of microstructural dimensions on FCG

In the fracture mechanics simulations of a mode III fati-gue crack emitting screw dislocations the cyclical slip irre-versibility p is again evaluated as a function of theapplied stress intensity factor. In Fig. 10 the functionaldependence of p on DKIII shows a square root trend. Thiscould be attributed to the involvement of multitudes of dis-locations that glide away to interact with the nanoscaleobstacle (twin) at larger KIII. The forces barring glidingof dislocations away from the source (originating fromincreasing pile-up and image stress) restrict dislocationmovement to a greater extent at larger KIII. Mughrabi[50] summarized the estimated cyclical slip irreversibilities,which were experimentally found to be almost negligible atlow loading amplitudes (leading to a long fatigue life), andclose to unity at larger loading amplitudes (resulting in ashort fatigue life). For diminishing twin lamellae spacingsthe irreversibility also decreased, as depicted in Fig. 10.The mechanism lies in the lowered capability of thinnertwins to hold back dislocations at larger applied KIII,resulting in a low slip irreversibility.

Another important question regarding the transmissionstress required for dislocations to penetrate these nanoscaleobstacles is illustrated in Fig. 11. From the MD simula-tions we can estimate the critical twinmatrix interfacezone size as normalized by the Burgers vector within whichthe CTB stress field would effectively elevate the energybarrier to transmission (Fig. 7c). The effective range ofthe interface influence zone size is in the range of a few mul-tiples of the Burgers vector. To determine the role of thepenetration strength of the nanotwins we looked at theevolution of da/dN at constant KIII with a varying degreeof CTB influence zone (i.e. transmission strength of thetwins). The results, shown in Fig. 11, indicate a neutral pla-teau in da/dN at low 2q/b for varying twin to crack spac-ings, where q is the distance from the twinmatrixPlease cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030interface. For twins located far enough from the crackthe rate of FCG is independent of the influence of disloca-tion transmission stress. We chose a value of 2q in this neu-tral plateau for subsequent da/dN vs. DKIII simulations inorder to compare the role of t and/or d on the FCGproperties.

The continuum simulation framework involving discretedislocations provides quantitative evidence of a role of themicrostructural dimensions t and d in FCG. Fig. 12a and bdemonstrates how the change in any one of these micro-structural characteristic lengths (with the other being keptconstant) affects da/dN. As can be seen, both the Parisand threshold regimes are influenced by such a variationin t or d. For a decrease in either the width of the nanotwinlamella or the twin to crack tip spacing da/dN alsodecreases. Crack advancement in the threshold regime isminuscule by nature, and is characterized by cycle by cyclediscrete plasticity. Our approach of modeling the crack tipplasticity accounts for individual dislocation contributions,thereby faithfully capturing the incremental crack growthfor both massive and minute slip activities.

The results have important implications in understand-ing mode II (shear mode) fatigue crack growth, as wellwhere edge dislocations are emitted from the crack tip.Although at the continuum level the description of edgeand screw dislocations are similar, as noted by Pippan,the reactions at the boundaries and residual dislocationswill differ, necessitating a complete analysis with corre-sponding MD simulations. This may explain the funda-mental differences between different threshold levelsobserved in the literature for mixed mode loading cases.

In summary, the critical microstructural characteristiclengths (t and d) associated with these obstacles play a pro-nounced role in FCG simulations. The correlation of FCGmetrics with the change in these characteristic microstruc-tural dimensions is governed by variations in the irrevers-ibility and blockage of slip emitted from the crack tipinteracting with an annealing nanotwin in the vicinity.FCG progresses via a combination of these two phenom-ena as influenced by nanotwins. If the CTB permits areduced degree of cyclical slip irreversibility the cyclicalcrack extension will act likewise. Therefore, FCG isexpected to occur at varying rates corresponding to thechanges in these characteristic microstructural dimensions(t and d). The insight obtained from such observations fur-ther clarifies the mechanism of cycle by cycle crack propa-gation. Slip irreversibility increases non-linearly as thesecharacteristic lengths become greater, eventually reachingsaturated levels. Similar trends in da/dN with respect tochanges in spacing of the twin to the crack tip or the twinlamella thickness are observed in the continuum FCG sim-ulations. Fig. 12a and b summarizes the variations in da/dN with these changes in the microstructural characteristicnanodimensions, thereby mapping the crack growthregimes. The trends in FCG in these calculations are con-sistent with earlier experimental findings in the literatureas reported by Sangid et al. [21] and Singh et al. [5].gue crack growth resistance of nanocrystalline alloys. Acta Mater

Fig. A1. The sensitivity of the calculation of cus (bulk and CTB) on thelength scale of the chosen tracing atoms. The desired ranges of l and w(normalized by the theoretical core width of a screw dislocation 2n [51])for tracing the potential energy variation due to the oncoming dislocationis highlighted. cus (bulk and/or CTB) converges with increasing l anddecreasing w.

P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxx 155. Conclusions

Utilizing MD and fracture mechanics simulations of dis-crete dislocation formulations we have studied the mecha-nism of FCG at the appropriate length scale by quantifyingcyclical slip irreversibilities. Our goal was to isolate the roleof nanotwins in the mechanism of FCG, as brought to ourattention by recent experimental findings. The major con-tributions of this work can be summarized as follows.

1. The analysis presented in this study underscores the roleof twin spacing and twin lamellar width on FCG. Wenote that the influence of these nanodimensions becomesmore prominent when the twin spacing and twin widthare typically less than 20 nm. The increase in FCG resis-tance is governed by the modified cyclical slip irrevers-ibility and dislocation annihilation behavior upon sliptwin interaction. The FCG metrics, such as da/dN andthe threshold stress intensity range, are increased to asubstantial extent on refinement of the nanotwin spacingand thickness. However, as these characteristic dimen-sions increase their role in FCG becomes less.

2. The investigation unfolded enhanced energy barriers forslip to glide across nanoscale CTBs owing to the pres-ence of residual dislocations during reverse flow undercyclical conditions. Quantification of the mismatch inenergy barriers to to and fro glide across twins providesa physical explanation for the irreversible glissile motionof slip. Such an insight extends our mechanistic under-standing of previously observed experimental findingson FCG as influenced by nanoscale twins.

3. Considerable attention has been devoted to ensuringconvergence of the unstable energy values when choos-ing fault dimensions of the order of the dislocation core.The sensitivity of the selected area of atoms to the unsta-ble energy values was determined for dislocationadvance in the matrix and near CTBs. In addition, thefriction stress levels were scaled to account for the strainrate and temperature effects. As a result, the differentialin friction stress upon forward and reverse loading wasshown to play the most significant role in FCG.

Acknowledgements

Support for this work was provided primarily by Hon-eywell Aerospace Corporation. We acknowledge the useof the parallel computing resource, the Taub cluster, atthe University of Illinois.

Appendix A

Since dislocation glide occurs via motion of the corethrough consecutive tearing and forming of atomic bondsaround the core the sequential rows of atoms on the slipplane ahead of the oncoming dislocation alternately comewithin the influence of the mobile core. Therefore, by calcu-Please cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030lating the variation between the enhanced potential energy(E) of such atoms and the bulk perfect lattice energy(Eperfect) one would be able to compute the c surface (usingEq. (A1) after Vitek et al. [44]). We devised a novel tech-nique of calculating the fault energies during dynamic slipmotion.

c E EperfectA

A1

Ahead of an oncoming dislocation a group of atoms,designated tracing atoms (with an area on the slip planeA = wl in Fig. A1), were carefully selected, where l and ware the distances parallel and normal to the dislocationline, respectively. The curves in Fig. A1 substantiate thelength scale independence of modified/unmodified cus onthe range of the selected tracing area dimensions (normal-ized by the theoretical core width of a screw dislocation,2n d where d is the interplanar distance (d 2.03 A fornickel) between (1 1 1) slip planes [51]).

A large w value spreading beyond the influence of thecore distortion encompasses lower energy bulk atoms inthe calculation, therefore the normalized values are suffi-ciently small for convergence, as shown in Fig. A1. Thevalue of l was also selected to ensure convergence, as againshown in Fig. A1. The energy of the tracing atoms was con-firmed to have a cus value consistent with density functionaltheory. For the case of an extended dislocation, as the lead-ing Shockley partial approaches the tracing atoms thevalue of c starts to increase, and achieves a maximum valuecus when the leading partial passes it. The departing leadingpartial leaves a stacking fault behind in its wake. At thispoint c assumes the value cisf. As the trailing partial trans-lates the value of c in the tracing area again starts togue crack growth resistance of nanocrystalline alloys. Acta Mater

Fig. A2. (a) Schematic demonstrating the atomic configuration of the fcc structure to illustrate the physics of dislocation dissociation. (Left) fcc stackingon consecutive (1 1 1) planes. Silver, green and red atoms represent planes A, B and C, respectively. (Right) Top projection of the (1 1 1) plane shows thedissociation of a full dislocation into two partials. Such dissociation is energetically favored over slipping along b1 due to the lowered cus in the partialdirections (b2 and b3). (b) Comparison of the standard (unmodified) GSFE curves for pure nickel using both molecular statics (MS) and current dynamiccalculation methods for the case of a full dislocation (b1) dissociating into two Shockley partials (b2 and b3). (For interpretation of the references to colourin this figure legend, the reader is referred to the web version of this article.)

16 P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxxincrease, eventually decreasing to the perfect lattice value.Then the value of c returns to zero.

c as a function of the reaction path coordinate computedin this study (away from the twin boundaries) is based onthe geometry shown in Fig. A2a and produces the well-known baseline GSFE of the sliding half-block approach(Fig. A2b).

In order to calculate the modified GSFE due to localstress sources the tracing atoms are selected at varyingproximities from the local stress source. Therefore, thestress source would contribute accordingly to the potentialenergy of the tracing atoms. Thus one can calculate themodified cus or the whole c displacement plot as influencedby the stress source in that particular position. Such a tech-nique is applied to obtain the variation of cus near the CTBduring forward/reverse flow (Fig. 7c). The maximum slopecalculated from such a modified whole c displacement (asin Fig. 7a) provides smax under the influence of local stress,which is then appropriately scaled to room temperatureand the lower strain rate, as explained in Appendix B.

Appendix B

Plastic flow (i.e. dislocation glide) as a function of tem-perature is modeled by an Arrhenius-type equation as fol-lows [52].

_c _co exp EakT

B1

where Ea is the activation energy barrier, T is the absolutetemperature (Kelvin), k is the Boltzmann constant, and _cois a constant associated with the rate of deformation. Thederivative of Ea with respect to the glide resistant stressso (a function of temperature and strain rate) providesthe kinetic signature of plastic deformation, which isPlease cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030defined as the activation volume V. V scales with thephysical area swept by the dislocations. For nanocrystal-line materials or in confined volumes where there are lim-ited dislocations sources it is of the order of several b3,while for bulk materials its magnitude can be 1000b3. Itsmagnitude can be determined experimentally or fromMD simulations. From the MD simulations the differencein flow stresses (Ds) for the nucleation of a single disloca-tion loop is calculated at two different strain rates ( _c2 and_c1) and a constant temperature, using Eq. (B2) [52]:

V @Ea@s

kT @ ln _c@s

kTDs

ln_c2_c1

B2

We note that Ea is a decreasing function of so. Consid-ering the linear so dependence of Ea [52] one can write

Ea E V so B3Using a temperature normalization introduced by Zhu

et al. [53],

so E

V kTV

lnkTNtol_cV

B4

where E is the athermal activation energy barrier, N is thenumber of nucleation sites, to is the attempt frequency, andl is the shear modulus. The value of E is established froma knowledge of the critical stress at 0 K and V. Bothquantities are determined from the MD simulations con-ducted in this study. Then the above equation allows thedetermination of so at different strain rates and tempera-tures. The constants utilized in our work are as follows:Boltzmann constant k = 1.3806503 1023 m2 kg s2 K1;athermal activation energy barrier E = 1.3 eV; activationvolume V = 2.25b3; Burgers vector~b a

61 1 2, where lat-

tice constant a = 3.52 A; number of nucleation sitesN = 100; attempt frequency to = 3.14 1011 Hz; sheargue crack growth resistance of nanocrystalline alloys. Acta Mater

Table B1The unstable stacking fault energy, the maximum (critical) stress for slip,and the critical stress so at room temperature and typical experimentalstrain rates (1104 s1).GSFE (pure nickel) cus (mJ m

2) smax (GPa) so (GPa)

Intrinsic (unmodified) 254 6 1.8Forward transmission 340 12.6 3.8Reverse transmission 452 21.3 6.4

P.B. Chowdhury et al. / Acta Materialia xxx (2013) xxxxxx 17modulus l = 76 GPa; temperature T = 300 K (room tem-perature); shear strain rate _c 1 104 s1. Because theMD calculations are conducted at 10 K and at high strainrates it was deemed necessary to scale the results for lowerstrain rates and room temperature. We note that this is stilla topic of current research; the modifications made aboveand illustrated in Table B1 represent the current state ofknowledge. Future refinements in this area will not, how-ever, change the conclusions reached in this work.

The intrinsic (unmodified) cus in Table B1 correspondsto the unstable stacking fault energy of pure Ni. The mod-ified cus levels in Table B1 are for a dislocation within thetwin boundary zone, as explained in the text. The corre-sponding critical stress levels are encountered during for-ward/reverse transmission, as shown in Fig. 7c. The smaxvalues are obtained from the slope of the c displacementcurves from the simulations. The third column is obtainedfrom Eq. (B4).

References

[1] ASME. Criteria of the ASME boiler and pressure vessel code fordesign by analysis in Sections III and VIII, Division 2. NewYork: ASME; 2010.

[2] Asaro RJ, Suresh S. Acta Mater 2005;53:336982.[3] Deng C, Sansoz F. Acta Mater 2009;57:6090101.[4] Lu L, Schwaiger R, Shan ZW, Dao M, Lu K, Suresh S. Acta Mater

2005;53:2169.[5] Singh A, Tang L, Dao M, Lu L, Suresh S. Acta Mater 2011;59:2437.[6] Wu ZX, Zhang YW, Srolovitz DJ. Acta Mater 2009;57:4508.[7] Zhu T, Li J, Samanta A, Kim HG, Suresh S. Proc Natl Acad Sci USA

2007;104:30316.[8] Zhao Y, Zhu Y, Lavernia EJ. Adv Eng Mater 2010;12:76978.[9] Zhong S, Koch T, Wang M, Scherer T, Walheim S, Hahn H, et al.

Small 2009;5:226570.[10] Zhu YT, Narayan J, Hirth JP, Mahajan S, Wu XL, Liao XZ. Acta

Mater 2009;57:376370.[11] Kumar KS, Van Swygenhoven H, Suresh S. Acta Mater

2003;51:5743.[12] Lu K, Lu L, Suresh S. Science 2009;324:34952.[13] Lu L, Shen Y, Chen X, Qian L, Lu K. Science 2004;304:4226.Please cite this article in press as: Chowdhury PB et al. Modeling fati(2013), http://dx.doi.org/10.1016/j.actamat.2013.01.030[14] Weertman JR. Mater Sci Eng A 1993;166:1617.[15] Conrad H, Narayan J. Scripta Mater 2000;42:102530.[16] Boyce B, Padilla H. Metal Mater Trans A 2011;42:1793.[17] Hanlon T, Kwon YN, Suresh S. Scripta Mater 2003;49:675.[18] Hanlon T, Tabachnikova ED, Suresh S. Int J Fatigue 2005;27:1147.[19] Padilla H, Boyce B. Exp Mech 2009;50:5.[20] Xie J, Wu X, Hong Y. Scripta Mater 2007;57:5.[21] Sangid MD, Pataky GJ, Sehitoglu H, Rateick RG, Niendorf T, Maier

HJ. Acta Mater 2011;59:7340.[22] Boettner, McEvily, Liu. Philos Mag 1964;10:95.[23] Forsyth PJE. Acta Metal 1963;11:703.[24] McClintock FA. In: Drucker DC, Gilman JJ, editors. Fracture of

Solids. London: Gordon & Breach Science Publishers, 1963, p. 65.[25] Neumann P. Acta Metal 1969;17:1219.[26] Neumann P. Acta Metal 1974;22:1167.[27] Pippan R. The dislocation-free zone and the threshold of the effective

stress intensity range. Elmsford (NY): Pergamon Press; 1989.[28] Pippan R. Acta Metal Mater 1991;39:255.[29] Pippan R. Int J Fracture 1992;58:305.[30] Riemelmoser F, Pippan R, Stuwe H. Int J Fracture 1997;85:157.[31] Ezaz T, Sangid MD, Sehitoglu H. Philos Mag 2011;91:1464.[32] Hartley CS, Blachon DLA. J Appl Phys 1978;49:4788.[33] Cai W, Bulatov VV, Chang J, Li J, Yip S. In: Nabarro FRN,

Hirth JP, editors. Dislocations in solids. Amsterdam: Elsevier; 2005.p. 1.

[34] Frenkel D, Smit B. Understanding molecular simulation: fromalgorithms to applications. New York: Academic Press; 2001.

[35] Tadmor EB, Miller RE. Modeling materials continuum, atomisticand multiscale techniques. Cambridge: Cambridge University Press;2011.

[36] Farkas D, Willemann M, Hyde B. Phys Rev Lett 2005;94:165502.[37] Nishimura K, Miyazaki N. Comput Mater Sci 2004;31:269.[38] Potirniche GP, Horstemeyer MF, Gullett PM, Jelinek B. Proc Roy

Soc Lond A Math Phys Eng Sci 2006;462:370731.[39] Plimpton S. J Comput Phys 1995;117:119.[40] Humphrey W, Dalke A, Schulten K. J Molec Graphics 1996;14:338.[41] Li J. Modeling Simul Mater Sci Eng 2003;11:173.[42] Zhou M. Proc Roy Soc Lond A Math Phys Eng Sci

2003;459:234792.[43] Foiles SM, Hoyt JJ. Acta Mater 2006;54:3351.[44] Vitek V. Philos Mag 1968;18:773.[45] Mahajan S, Chin GY. Acta Metal 1973;21:173.[46] Jin ZH, Gumbsch P, Ma E, Albe K, Lu K, Hahn H, et al. Scripta

Mater 2006;54:1163.[47] Kumar KS, Suresh S, Chisholm MF, Horton JA, Wang P. Acta

Mater 2003;51:387.[48] Yang Y, Imasogie B, Fan GJ, Liaw PK, Soboyejo WO. Metal Mater

Trans A 2008;39:1145.[49] Christian JW, Mahajan S. Prog Mater Sci 1995;39:1.[50] Mughrabi H. Metal Mater Trans A 2009;40:1257.[51] Hirth JP, Lothe J. Theory of dislocations. New York: McGraw-Hill;

1967. p. 780.[52] Conrad H, Frederick S. Acta Metal 1962;10:1013.[53] Zhu T, Li J, Samanta A, Leach A, Gall K. Phys Rev Lett

2008;100:025502.gue crack growth resistance of nanocrystalline alloys. Acta Mater