The ductile fracture of precipitation-hardened Al-1.79wt.%Cu alloys containing θ′ and θ precipitates

The ductile fracture of precipitation-hardened Al-1.79wt.%Cu alloys containing θ′ and θ precipitates

Materials Science and Engineering, 49 (1981) 229 - 240 229 The Ductile Fracture of Precipitation-hardened Al-1.79wt.%Cu Alloys Containing 0' and 0 P...

1MB Sizes 0 Downloads 37 Views

Materials Science and Engineering, 49 (1981) 229 - 240

229

The Ductile Fracture of Precipitation-hardened Al-1.79wt.%Cu Alloys Containing 0' and 0 Precipitates I. Y. T. CHAN* and H. G. F. WILSDORF

Department of Materials Science, University of Virginia, Charlottesville, VA 22901 (U.S.A.) (Received January 24, 1981)

SUMMARY

The ductile fracture of Al-1.79wt.%Cu alloys aged to contain either O' or 0 precipitates was investigated using electron microscopy. In situ fracture experiments in a high voltage electron microscope involving thin specimens revealed intense dislocation activities ahead o f the crack tip. Crack propagation was by hole coalescence, leaving behind serrated crack edges as shown in the transmission electron microscope, while scanning electron microscopy showed lamellae on fracture surfaces. In bulk specimens, dimpled fracture surfaces indicated that crack propagation occurred by void coalescence. Microcrack initiation sites were most commonly found at dislocation boundaries rather than at precipitate-matrix interfaces. Consistent with the experimental results, a comparison between the estimated available energy for decohesion from dislocation boundaries and from interracial energies predicted the dislocation boundaries to be the favorable decohesion sites. The precipitates can be expected to have influenced the development o f dislocation patterns during the work-hardening stage, but they played only a minor role in the final process o f fracture initiation.

1. INTRODUCTION

The process of ductile fracture has been known to proceed in the sequence of fracture initiation by microcracks or microvoids, their growth and their subsequent coalescence. One *Present address: Department of Physics, Arizona State University, Tempe, AZ 85281, U.S.A. 0025-5416/81/0000-0000/$02.50

of the foremost goals of fracture research is to establish criteria for fracture initiation. Generally, fracture initiation sites are inhomogeneous regions inside the material, either pre-existing or deformation induced, i.e. second-phase inclusions, and groups of point, line and planar defects respectively. Although much attention has been focused on the study of interphase-interface decohesion [1 - 7], recent results from high purity materials have demonstrated that dislocation boundaries were also active fracture initiation sites [8 - 13]. From the types of metals and alloys investigated so far, it appears that dislocation boundaries were the primary decohesion sites in high ductility materials while second-phase inclusions might play a more important role in materials with a lower ductility. It is the aim in this investigation to assess the relative importance of these two kinds of inhomogeneity in the fracture process of a precipitationhardened alloy. The A1-Cu system was chosen for this study because it is made up of a highly ductile matrix (aluminum) containing harder inclusions (0' and 0 precipitates), i.e. it has both kinds of inhomogeneity in abundance. This investigation is based on in situ fracture experiments inside a high voltage electron microscope and post-fracture studies using both scanning electron microscopy (SEM) and transmission electron microscopy (TEM). In situ experiments allowed the direct observation of microcrack initiation and crack propagation at high magnifications. Information on the development of microstructural processes thus obtained is otherwise inaccessible using conventional experimental techniques such as fracture toughness testing and fractography. SEM and TEM post-fracture studies were complementary to the in situ experiments in characterizing the microstructures. In addi© Elsevier Sequoia/Printed in The Netherlands

230 tion, fracture studies of bulk specimens aged to contain the same precipitate phases were also performed. These bulk specimens were fractured in the Instron tensile-testing machine and then examined by SEM. This served to provide a comparison between the results obtained from thin specimens and bulk specimens. 2. SPECIMEN PREPARATION The bulk specimens were machined from sheets 0.17 cm thick. Each specimen was solid solution treated for at least 12 h before the aging treatment. The aging conditions and results are given in Table 1. The specimens were then electropolished and gauge marks were drawn b y a Rapidograph marker at a b o u t 0.5 cm apart along the entire gauge length. The fracture experiments were performed in the Instron tensile machine at a constant elongation rate of 0.05 cm min -1. Portions containing the fracture surfaces were cut o f f a t a b o u t 0.7 cm away from the fracture surfaces and then m o u n t e d on a carbon specimen m o u n t for SEM examination. The general preparation of thin specimens for in situ experiments has been described

previously together with details of the video recording technique used with high voltage electron microscopy (HVEM) studies [ 1 4 ] . Tensile specimens 1.2 cm long and 0.2 cm wide were punched o u t from sheet 0.02 cm thick. The grain size varied between 0.01 and 0.03 cm. Electron-transparent areas with a diameter of a b o u t 0.1 cm were polished electrolytically in the center of the specimens [ 15]. The best specimens contained no holes and showed slip lines on the surface of all grains after deformation preceding crack initiation. 3. RESULTS 3.1. Mechanical testing

The numerical results obtained from mechanical testing are listed in Table 2. In each case, fracture occurred inside the gauge length which had the largest elongation. Each crack appeared first at the center portion of the specimen width and then propagated outward. The time which elapsed from the appearance of the first hole to complete fracture was approximately I min for every specimen, independent of the precipitate phases present. Thus the average crack speed

TABLE 1 Aging c o n d i t i o n s and results

for bulk s p e c i m e n s

Annealing time

Aging temperature

Aging time

Averagegrain size a

(h)

(°C)

(h)

(cm)

12 12 12

-160 300

68 22

~ 0.1 ~ 0.07 ~ 0.07

Precipitate obtained SSS

0' 0

aThe range o f grain sizes for individual s p e c i m e n s was b e t w e e n 0 . 0 2 and 0.2 cm. TABLE 2 Results o f tensile fracture e x p e r i m e n t s o n bulk s p e c i m e n s

oy (MPa)

of C(MPa)

160

65

140

42

116

32

Type ofprec~i~te

UTSa(MPa)

SSS

0' 0

RA d(%)

eF

490

87

470

90

2.1 2.3

220

58

0.8

a U l t i m a t e tensile stress U T S = m a x i m u m l o a d / A 0. b y i e l d stress Oy = load at 0.2% strain/A 0. CFracture stress o t = load at the time w h e n first h o l e a p p e a r e d / A t. d R e d u c t i o n in area R A = ~(A 0 --Af)/Ao~ × 100%. eTrue fracture strain ef = ln(Ao/Af) where A 0 is the initial area and A f is the fracture surface area measured from SEM micrographs.

231

appeared to be independent of fracture stress of and fracture strain el. Long straight slip bands were present on every grain inside the gauge region and multiple slip was observed in grains close to the final fracture surface.

3.2. Fractography The fracture surfaces of the supersaturated solid solution (SSS), 0' and 0 specimens are shown in Fig. 1. The reduction in area and the appearances of the fracture surfaces were remarkably similar in the SSS and O' specimens. Each contained large deep shear dimples arranged in a narrow band. At higher magnifications the SSS specimens were seen to have a slightly higher density of large dimples than 0' specimens had. This suggests that there were more void initiation sites in the SSS specimens. Although the average density of O' precipitates measured from TEM micrographs was 2.8 X 108 c m - 2 , the density of the large dimples in the 0' specimens was 2.5 × 10 5 c m - 2 , a difference of three orders of magnitude. Furthermore, since the SSS specimens were essentially inclusion free within each grain, the formation of this dimple type was evidently unrelated to the presence of 0' precipitates. X-ray diffraction maps of copper on the fracture surface showed an essentially random distribution. However, this result is thought to be due to the large penetration depth of the incident electrons in aluminum. The dimples found on the fracture surfaces of the 0 specimens were mainly of the equiaxed type. The average dimple density was 5 × 107 c m - 2 , twice the density of the 0 precipitates. Their sizes were much smaller and the dimple density was two orders of magnitude higher than for the SSS and 0' specimens. 3.3. In situ fracture observations All cracks were transgranular and initiated at the thinnest part of the polished area. Microcracks appearing as holes were seen to open up ahead of the advancing crack. The number of holes varied from 3 to 5 with separations ranging from 2 to 4 ~m. The video recordings showed that the crack propagation speeds in the 0' specimens were considerably more uniform than those in the 0 specimens. Crack propagation speeds were measured and are listed in Tables 3 and 4. The cracks in the 0 specimens often advanced

TABLE 3 Crack velocities in specimens containing 0' precipitates determined from video recordings of HVEM in situ experiments

Time period of measurement

Magnification of video picture

(s)

(x lO 3)

50 45 200 15 80 90 90

80 40 30 30 30 30 30

Crack velocity ( ~ s-1 )

200 300 150 600 150 200 250

TABLE 4 Crack velocities in specimens containing 0 precipitates determined from video recordings of HVEM in situ experiments

Time period of measurement

Magnification o f video picture


(x 10 3)

40 0.8 5 75 40 0.5 1

40 40 20 30 30 30 30

Crack velocity (• s-1 )

500 70000 15000 800 600 130000 50000

a a

a a

aveloeity measured for sudden crack advances.

suddenly over distances of microns {Fig. 2). At the end of sudden advances, the crack tip was blunted and thinning ahead of the tip resumed in a slightly different direction. As a consequence the contour of the crack edges, if we ignore "dragon's teeth", were more jagged than for the 0' specimens. This is interpreted as resulting from differences in precipitate sizes and in their distribution. Because the 0 precipitates were irregularly distributed and larger, they provided inhomogeneity on a larger scale, which resulted in a corresponding variation in the local stress state. When a crack propagated into an originally non-transparent region, thinning of the material ahead of the crack tip was observed, often to the extent of 10 pm; the intense plastic zone at the crack flanks was a b o u t 1 pm

232

(a)

(b)

(c) Fig. 1. Fracture surfaces of bulk specimens as seen by SEM: (a) SSS surface; (b) precipitation-hardened (8') surface; (c) precipitation-hardened (0) surface. wide. This indicates t h a t t h e stress field o f t h e c r a c k was v e r y m u c h d i r e c t e d t o w a r d s t h e f o r w a r d direction. Cell-like p a t t e r n s were

o b s e r v e d a h e a d o f t h e c r a c k tip as s h o w n in Fig. 3. As a c r a c k a d v a n c e d , individual cell shapes c h a n g e d rapidly as a result o f intense

233

Fig. 2. Time sequence micrographs taken from videotape recording at 1/15 s intervals, showing sudden crack advance in a specimen containing 8 precipitates fractured in the high voltage electron microscope. The arrows indicate the boundary at which the event was initiated. Dragon's teeth as well as ligaments and further hole initiation are discernible in (d).

Fig. 3. A dislocation cell pattern ahead of crack tip in a specimen containing 0' precipitates (videotape recording). The arrow indicates a microcrack. Dragon's teeth can also be seen.

234

dislocation activity. Although it was impossible to identify individual dislocation sources, approximate "active regions" where the cell pattern changed most rapidly could be discerned. Along the propagation direction ahead of the crack tip there were at least three of these "active regions" per micron within the transparent area. Compared with the hole spacing, each ligament between successive holes contained a few of these regions. It was found that the ligaments further deformed until rupture occurred, resulting in the formation of d.ragon's teeth. The spacing of the dragon's teeth ranged from 0.3 to 0.6 pm, which is significantly less than the original hole spacing but comparable with the spacing of the active regions. Evidently, additional microcracks were created during the deformation of ligaments which were closely related to active regions. The microcracks were seen to be initiated at the cell boundaries which appeared as dark lines in the micrographs. They then increased in size by propagating along those lines. The overall process gave the impression that the cracks propagated by splitting along those boundaries. In general, few precipitates were observed in the thinned regions. At first it was believed that they were masked by complex image contrasts but subsequent post-fracture studies by TEM revealed that they were rarely present immediately adjacent to the crack. This indicates that precipitate-interface decohesion was n o t the major mechanism for microcrack initiation.

3.4. Fractography of thin specimens Observations from optical microscopy revealed that at least two sets of slip lines appeared in the grains before a crack would eventually propagate through the crystal. As deformation continued, the slip lines became finer and eventually unresolvable by the optical microscope, i.e. the crack always passed through the most heavily deformed region where the slip lines were most dense. As seen by SEM, the fracture surfaces of each thin specimen consisted of lamellae. From the thick part of the crack flank to the crack the topography changed from parallel slip steps spaced about 0.7 ~m apart, to very fine and wavy slip steps and then finally to thin, triangular-shaped fractured ligaments. The

number of 0 precipitates found at the crack edge was much smaller than the number of dragon's teeth. In 0' specimens, stereo microscopy (Fig. 4) allowed us to see that the wavy lamellae were arranged to form shallow dimples. From all observations it was concluded that the sequence of events in the fracturing process of 0' and 0 specimens was similar: (i) microcracks were initiated after substantial multiple slip; (ii) ligaments continued to deform until the final separation took place at heavily drawn out ligaments, leaving behind dragon's teeth.

Fig. 4. Stereo pair SEM micrographs showing a striated ledge fracture surface of a thin specimen containing 0t precipitates. The white dots are believed to be atmospheric etching of the 0 t precipitates; their density distinctly decreases toward the final fracture surface where dragon's teeth can be seen.

235

3.5. Dislocation patterns In situ HVEM experiments revealed that microcracks were initiated at dark lines which were arranged in cell patterns ahead of the crack tip (Fig. 3). When examined closer by TEM, these lines were seen as bands and did n o t represent well-defined sharp boundaries. By tilting the specimen they were confirmed to be due to Bragg diffraction contrast and n o t to bend contours. Convergent beam microdiffraction (CBMD) patterns taken from both sides of such a boundary revealed rigid rotations of the same pattern. CBMD patterns from the dark bands exhibited substantially stronger scattering than scattering from areas away from the bands, suggesting the existence of a high dislocation density. The rotation across these boundaries was found to be between 2 ° and 5 ° [ 1 5 ] . All selected area diffraction (SAD) patterns obtained from the crack flank almost resembled polycrystalline patterns with all the crystallites close to the same orientation, i.e. the SAD patterns can be reproduced by rigidly rotating the basic pattern over a small angle and then superimposing it on t o p of the original pattern. Similar SAD patterns were

(a)

obtained by Humphreys [16] from deformed A1-Cu-Si alloys. The a m o u n t of misorientation present within the area that yielded the SAD pattern can be estimated by measuring the angles subtended by the multiple spots from the same diffracting plane. It was found that 85% of the angles thus measured were less than 7 ° and 50% were between 2.5 ° and 4.5 ° . Although these measurements did n o t deal with the misoriented areas individually, the results agreed well with those from the CBMD patterns. Figure 5 is a typical dark field image of a crack flank. The misorientations present among the different areas are shown by the differences in their diffracting contrast. The shapes of the large bright areas were irregular and generally elongated along the crack edge, with typical dimensions of a b o u t 0.5 pm. Within these areas, smaller subdivisions can be discerned. These are interpreted as subcells and their sizes ranged from less than 0.1 p m to a b o u t 0.2 ~m. In all cases their boundaries were also irregular and n o t well defined. At the thinner region, tangled individual dislocations along a boundary were resolved. At even smaller thicknesses, dislocation boundaries

(b)

Fig. 5. Dark field micrographs of a crack flank using an Al(200) reflection. The arrows point along dislocation boundaries in which individual dislocations are resolved. The specimen contained 0' precipitates.

236 formed b y short dislocations lying approximately normal to the surfaces were often seen. These boundaries occurred when dislocations under the influence o f surface image forces rearranged into their lowest energy configuration. Measurements of dislocation densities p were performed whenever possible and the results were consistently of the order of I0 n cm -2.

Many fine fringe patterns existed in dark field micrographs of thicker areas (Fig. 6). From their relationship with the diffraction vectors these were identified, as rotational moir~ fringes. Statistics from 86 sets of fringes showed that the majority had spacings between 2 and 4 nm. These corresponded to rotations of 3 ° - 5 ° between volume elements. This is in excellent agreement with the results from SAD and CBMD patterns. However, although the SAD and CBMD patterns revealed misorientations between areas connected two dimensionally in the image, the moir~ fringes originated from misorientations between overlapping areas. Thus far the results demonstrated the existence at the crack edge of misoriented volume elements which are three dimensionally connected and separated by high dislocation density boundaries.

Fig. 7. Dislocation cells between large 0 precipitates at the crack flank. 1.7 pm) of the 0 precipitates. Apart from a few 0 precipitates protruding from the crack edge (no more than five precipitates in any specimen), no consistent relationships between precipitates and decohesion sites were found. In 0' specimens, precipitates were mostly seen at the dislocation cell boundaries. Some o f the 0' precipitates were deformed and some appeared to have been rigidly rotated. Dislocations wrapped around the precipitates were also noted. Although the initial distribution of the 0' precipitates was uniform (average separation, approximately 0.5 pm), often no precipitates were found over large areas at the fracture flank. This is understandable, since the strain in this area is estimated to exceed a few hundred per cent.

Fig. 6. A dark field micrograph of a crack flank using

an Al(311) reflection, showing moir6 fringes.

3.6. Precipitates Some 0 precipitates near the crack were plastically deformed, and dislocation cells were seen in the region of the matrix between the 0 precipitates (Fig. 7). The cell size was approximately 0.2 pm, significantly less than the original separation (which amounted to

3.7. Dragon's teeth Dragon's teeth are the remains of ruptured ligaments. Their thickness was estimated to be less than 50 rim. Their presence seems to be a general p h e n o m e n o n in the fracture o f ductile materials as they have also been found in silver [ 8 ] , stainless steel [ 9 ] , iron [ 1 0 ] , beryllium [ l i ] and copper [ 1 2 ] . They were generally triangular in shape and dark

237

Fig. 8. Dragon's teeth of a specimen containing 0

precipitates. Dislocation boundaries formed by short dislocations lying approximately normal to the surface (arrows) are common in dragon's teeth.

contrasts were usually found at their bases where they were blending in with th.e thicker region of the crack flank {Fig. 8). In SEM they appeared as very thin lamellae protruding from the crack edge, as shown in Fig. 4. Dislocation boundaries formed by end-on dislocations were frequently found on these fracture teeth. Occasionally, they contain 0' precipitates, b u t no precipitate was observed to have any specific relationship to hole formation in general. From in situ observations it is clear that the holes at which fracture was initiated locally must be at some point along the edge of the dragon's teeth. If the hole was caused by interface decohesion, then that initiation site should have been marked by a precipitate. The rarity of such an occurrence in both the 0' and the 0 specimens indicated that true interface decohesion rarely occurred. 4. DISCUSSION

The mechanism by which microcracks can be initiated at the interfaces between secondphase inclusions and the ductile matrix has c o m m o n l y been explained b y means of a dislocation pile-up model [17 - 1 9 ] . In this model, when the length of a dislocation pileup increases to a critical value, the dislocations will coalesce at the head of the pile-up, thereby creating a void at the interface. This model requires that the dislocations are primarily confined to one set of slip planes. Therefore its application should be limited to relatively brittle materials. In A1-Cu alloys, with the

ease of dislocation cross-slip the dislocations in the aluminum matrix were seen to be arranged into dislocation cells instead of forming pile-ups. Hence a pile-up mode of interface decohesion cannot be significant in this alloy. From an energy viewpoint the interfacial energy between the matrix material and the inclusion is believed to aid in the interface decohesion process. The value of the interfacial energy has c o m m o n l y been obtained from calorimetry experiments. Boyd and Nicholson [20] obtained a value of 1580 erg cm -2 for the interfacial enthalpy of the peripheral interface of 0' precipitates. Aaron and Aaronson [21] measured the angle of contact of 0 precipitates at grain boundaries and determined the interfacial energy of the 0 precipitates to be 295 erg c m -2. However, it must be kept in mind that these values are only estimates. Furthermore, calorimetry experiments measure the total energy of the interface, which includes the entropic, internal and mechanical (strain) energies, while only the mechanical c o m p o n e n t contributes to the decohesion process. It is known that interface decohesion is much less at incoherent interfaces which often have a lower interfacial energy than semicoherent interfaces. This implies that the mechanical energy available for decohesion from the interface of an incoherent 0 precipitate is larger than that from the interface of the semicoherent 0' precipitate, even though its total interfacial energy is lower. If it is assumed that the entire interfacial energy of the 0 precipitate is mechanical in nature, the upper limit for the available energy from the interfaces would be a b o u t 300 erg cm -2. Since the frequency of microcracks initiating at dislocation boundaries was much higher than that of interface decohesion in the A1-Cu alloy investigated, it appears that the available energy from the dislocation boundaries for decohesion was higher than the mechanical part of the interfacial precipitate energy. An assessment of the energy contributed by the dislocation boundaries appeared to be desirable. According to the present results, the misorientation between adjacent dislocation cells was usually between 2 ° and 5 ° . If the boundaries of the cells were of the simple tilt or twist type, this corresponds to an energy of

238

about 200 - 260 erg cm -2 [22]. However, there are three important facts that must be taken into consideration. (i) Numerical values of the misorientations are based on projections; therefore this represents the lower limit. (ii) The boundaries actually consisted of high density dislocation tangles and therefore their energies will be higher than that of a simple tilt or twist boundary with the same misorientation. (iii) The boundaries analyzed in the postfracture studies were those that were left behind after fracture, while those that had initiated the microcracks had mainly dissociated (Figs. 2 and 8). Evidently, the original boundaries must have possessed higher energies and higher dislocation densities, rendering them as energetically more favorable decohesion sites than precipitate interfaces. The dislocation density at the microcrack initiation site can be estimated through the flow-fracture (r-of) stress expression. I" = o~Gbp 112

where G is the shear modulus and ~ is approximately 0.4. By setting r equal to of, for the 0' specimens where of = 470 MPa, p was computed to be 2.4 X 1012 c m -2. The stored energy given by the expression [23]

vst= P

G b 2 l n t P 1/2 )

--4-

T

yields Ust = 1.3 × l 0 s erg cm -a. From in s i t u experiments the boundaries where microcracks had been initiated were approximately 0.1 p m wide. Thus the twodimensional projected energy density at the microcrack initiation site will be 1.3 × 103 erg cm -2. For 0 precipitates the corresponding values are of = 220 MPa, p = 5.4 × 1011 c m - 2 and Ust = 3.5 × 10 v erg cm -a, and the projected energy density is 3.5 × 102 erg cm -2. Although these numbers are only empirical estimates, t h e y seemed to be representative for the following reasons. (i) The calculated dislocation density agreed well both with dislocation density data published elsewhere on deformed metals (e.g. ref. 24) and with the dislocation densities (1011 cm -2) measured at the crack flank in post-fracture studies, after allowing for dislocations lost during the dissociation of the original boundary.

(ii) If microcrack initiation can be described by an empirical probability function of the form exp(--CU) where C is some constant and U is the total energy needed for decohesion, U becomes U - U' where U' is the energy available from either the interface or the dislocation boundaries, depending on where the decohesion process takes place. Thus the relative frequency of occurrence of each kind of event will depend on the magnitude of the energy available from each process. The results of this investigation that microcrack initiation sites were located primarily at dislocation boundaries agreed with the relative magnitudes of available energy densities estimated for dislocation boundaries and precipitate-matrix interfaces. Rotations between adjacent dislocation cells were found to be continuous. This is probably due to the large strain gradient present at the crack flank, as seen by the sharp change in thickness in both TEM and SEM observations. Models of possible dislocation-precipitate interactions, in which the dislocation cross slipped to avoid the 0' precipitates or left behind loops around the precipitates, have been proposed by various investigators [ 25 27]. However, the dislocation density at the crack flank was too high to allow a clear identification of these interactions. Nevertheless, it is apparent that almost the entire strain was a c c o m m o d a t e d in the aluminum matrix. It is known that for a test specimen of standard diameter the stress state gradually changes after necking from plane strain in the interior to plane stress at the surface (for a detailed discussion see ref. 28). For the thin specimens that were used in the in s i t u experiments it must be assumed that they were under conditions of plane stress, while the bulk specimens that were fractured in the Instron tensile machine were under approximately plane strain as well as plane stress conditions. However, these statements apply only to the i n i t i a l loading conditions and, furthermore, the material was not homogeneous from a microstructural viewpoint so that the local stresses would inevitably become m u c h more complex with increasing strain. After microcracks and/or microvoids had been formed, their stress-raising effect led to further complexities. In view of this, only a phenomenological comparison between the

239

fracture process in a bulk specimen and that in a thin specimen can be made. The final fracture surface of a bulk specimen was covered with dimples while that of a thin specimen consisted of striated ledges containing dragon's teeth. The present results showed that the dimples in the bulk specimens were not correlated with the precipitates in any simple manner. For the thin specimens, in situ observations revealed that the microcracks were initiated at the dislocation boundaries. This is further supported by the fact that the separation between the dragon's teeth (approximately 0.4 pm) was independent of the kind of precipitates present and was considerably less than the separation between 0 precipitates (greater than 1.5 um). It is tempting to interpret the fracture of the thin specimens to be representative of the separation of the dimples in the bulk material, but no definite correlation has been accomplished. However, it is logical to suggest that the striated fracture surfaces of the thin specimens were composed of dimples with severely drawn out ligaments between them, as indicated in Fig. 9. Clearly, both microvoids and microcracks were primarily initiated at dislocation boundaries. The difference in their fracture surface topographies is a result of the different constraints (effect of surfaces) in the microLEDGES

(a)

(h)

SEVERELY DISTORTED DIMPLES

DIMPLES

Fig. 9. A schematic drawing comparing (a) fracture surfaces of thin specimens with (b) those of the bulk specimens. The space between the ledges in (a) can he viewed as severely distorted dimples.

void and microcrack growth processes, i.e. a microvoid is completely enclosed inside the material after nucleation, while a microcrack is essentially an elongated hole through the plane of the thin specimen. Thus the shape of dimples in fracture surfaces will depend on the geometry of the fracture surface. In the fibrous part of a cup-and-cone fracture, dimples will tend to be nearly round or elliptical while, after complete rupture of a foil, dimples can be expected to be markedly elongated on account of the chisel edge geometry. The question whether ruptured crystals of larger original cross section will exhibit zigzag flanks comparable with those obtained from thin foils was answered by experiments with copper crystals (purity, 99.999%) of square cross section with an edge length 0.4 cm. These crystals were ruptured in an Instron tensile-testing machine and prepared for direct examination by TEM. The crack flanks showed dragon's teeth of the same type found in in situ tensile specimens ruptured in the high voltage electron microscope. This is not surprising since, after all, ductile fracture involves the rupture of ligaments in the final separation process, which follows the same sequence of microcrack initiation and growth to coalescence [29]. Even though it has been well established that the precipitates are responsible for the hardening behavior, their role in the final fracture initiation in this low copper concentration aluminum alloy appears to be secondary. We concur with the observations of Lloyd and Kenny [30] that precipitates act as primary dislocation obstacles at the initial stages of deformation; however, with increasing strain the dislocation cell diameters became smaller than the interparticle spacings and thus the dislocations themselves became the major obstacles for their own movement. The situation might be different if the alloy contains a higher copper concentration, in which case the initial separation of the precipitates would be smaller and might reduce or even suppress the formation of distinct dislocation boundaries separating volume elements of appreciable misorientations with respect to each other. Macroseopieally, the hardening rate and yield stress would increase at the expense of ductility, and fracture initiation would be in accordance with mechanisms described in the literature.

240 5. CONCLUSIONS

(1) It was observed by in situ HVEM that the fracture process proceeded in the sequence microcrack and/or microvoid initiation, growth and coalescence. (2) The crack flanks of the thin, i.e. electrontransparent, specimens were populated with low angle dislocation boundaries consisting of dislocation tangles of high dislocation densities. (3) The fracture surfaces of the bulk specimens were covered typically with dimples while those of the thin specimens were covered with ledges with dragon's teeth protruding from them. These differences were due to the different constraints in the microvoid or microcrack growth process. (4) Dislocation boundaries were the primary fracture initiation sites for both bulk and thin specimens in Al-l.79wt.%Cu alloys irrespective of whether they contained 0' or 0 precipitates. This was supported by (i) video tape observations of in situ fracture of thin specimens in a high voltage electron microscope, (ii) dimple counts versus precipitate densities in bulk specimens and (iii) a comparison of relative magnitudes between matrix-precipitate interfacial energies and energies of dislocation boundaries. ACKNOWLEDGMENTS

The interest and support of this research by the Office of Naval Research is gratefully

acknowledged. We thank Mr. B. Ward, Reynolds Metals Co, Richmond, VA, for providing the alloy. REFERENCES 1 K. E. Puttick, Philos. Mag., 4 (1959) 964. 2 M. F. Ashby, Philos. Mag., 14 (1966) 1157. 3 A. R. Rosenfield, Metall. Rev., 13 (1968) 29.

4 K. Tanaka, T. Mori and T. Nakamura, Trans. Iron Steel Inst. Jpn., 11 (1971)383. 5 D. Broek, Eng. Fract. Mech., 5 {1973) 55. 6 A. S. Argon, J. Im and R. Safogler, Metall. Trans. A, 6 (1975) 825. 7 L.M. Brown and W. M. Stobbs, Philos. Mag., 34 (1976) 351. 8 R. L. Lyles and H. G. F. Wilsdorf, Acta Metall., 23 (1975) 269. 9 R. W. Bauer and H. G. F. Wilsdorf, in G. C. Sih (ed.), Proc. Int. Conf. on Dynamic Crack Propagation, Noordhoff, Leyden, 1973. 10 R.N. Gardner, Dissertation, University of Virginia, 1977. 11 T. C. Pollock, Dissertation, University of Virginia, 1977. 12 J. E. French and P. F. Weinrich, Metall. Trans. A, 6 (1975) 785. 13 R.N. Gardner, T. C. Pollock and H. G. F. Wilsdorf, Mater. Sci. Eng., 29 (1977) 169. 14 R.W. Bauer, R. L. Lyles and H. G. F. Wilsdorf, Z. Metallkd., 63 (1972) 525. 15 I. Y. T. Chan, Dissertation, University of Virginia, 1980. 16 F. J. Humphreys, Acta Metall., 27 (1979) 1801. 17 C. Zener, Fracture o f Metals, American Society for Metals, Metals Park, OH, 1948. 18 N. J. Petch, in H. Liebowitz (ed.), Fracture I, Academic Press, New York, 1968. 19 A. N. Stroh, Proc. R. Soc. London, Ser. A, 223 (1954) 404. 20 J. D. Boyd and R. B. Nicholson, Acta Metall., 19 (1971) 1101. 21 H. B. Aaron and H. I. Aaronson, Acta Metall., 18 (1970) 699. 22 L. E. Murr, Interracial Phenomena in Metals and Alloys, Addison-Wesley, Reading, MA, 1975. 23 D. Kuhlmann-Wilsdorf, in J. P. Hirth and J. Weertman (eds.), Work Hardening, Gordon and Breach, New York, 1968. 24 D. A. Rigney, Scr. Metall., 13 (1979) 353. 25 G. Thomas and J. Nutting, J. Inst. Met., 86 (1957) 7. 26 P. B. Hirsch, J. Inst. Met., 86 (1957 - 1958) 13. 27 S. Koda and T. Takeyama, J. Inst. Met., 86 (1957 - 1958) 277. 28 D. Broek, Elementary Engineering Fracture Mechanics, Noordhoff, Leyden, 1974. 29 H. G. F. Wilsdorf, Krist. Tech., 14 (1979) 1265. 30 D. J. Lloyd and D. Kenny, Acta Metall., 28 (1980) 639.