Professional Documents
Culture Documents
bodies fail mechanically by subcritical crack growth.1,2 When subjected to an applied stress, cracks located in the glass surface grow until they reach a critical size for failure. The crack velocity then accelerates and failure occurs almost instantaneously. Water or water vapor is the main chemical agent for subcritical crack growth in glass. Without water, cracks either do not grow or require higher stresses than in the presence of water.3 Because of their importance to lifetime prediction in glass, fracture mechanics techniques have been used extensively to study the dependence of crack growth on stress, chemical composition of the glass, and chemical content of the environment.2,4 One observation that has been particularly intriguing is that of an apparent stress threshold below which cracks will not grow in glass. In the glass
LASS
C. R. Kurkjiancontributing editor
Manuscript No. 187228. Received January 2, 2002; approved June 17, 2002. Financial support for S. M. Wiederhorn was provided by the Humboldt Foundation. *Member, American Ceramic Society.
*The use of commercial names is only for purposes of identification and does not imply endorsement by the National Institute of Standards and Technology.
2287
2288
Fig. 1. Crack growth in silicate glasses (taken from Ref. 7). The technique used to collect these data was the double cantilever technique. All specimens were immersed in distilled water. The sodalimesilicate specimens used in Ref. 7 are from the same batch of microscope slides as used here.
Fig. 2. Schematic diagram of the method developed by Michalske9 to prove the existence of crack blunting. The same experimental procedure was followed in the present study.
apparent crack growth threshold and the time delay to restart the crack was a function of the glass composition and the test environment. The time to restart an arrested crack depended on the applied stress intensity factor, the aging time, and the leaching behavior of the glass. The closer the applied stress intensity factor was to the threshold and the longer the hold time, the longer it took to restart the crack. In addition to the time delay for restarting the crack, Michalske9 showed that the appearance of the crack surface at the point of crack arrest depended on the length of the hold time. After carrying out a set of crack growth experiments and breaking the specimen open along the fracture plane, Michalske found that cracks that were momentarily stopped and then restarted left a fine curved line at the point of crack arrest. This line was a consequence of a slight change in the crack growth direction when the crack was stopped and restarted. For cracks that were held for 16 h, new features in the form of light and dark bands were observed at the crack tip. These were thought to represent the nucleation of multiple sharp cracks from the crack that had been blunted. Gehrke et al.11 also showed that marks were left on the surface where the crack had been arrested. These marks seemed to differ from those observed by Michalske; however, in both studies the features were just above the limit of optical resolution, so that an accurate comparison between the two studies is difficult. In this paper we use the atomic force microscope (AFM) to explore the nature of the features that form along an arrested crack front. As will be seen, this technique has the height and lateral resolution to quantify the nature of the fracture features that are left by the arrested crack. We shall show that the features are not a consequence of crack blunting. We then speculate as to their cause and the possible applicability of a crack-growth theory presented by Chuang and Fuller12 as an explanation of the results. II. Experimental Procedure
The experimental techniques used in the present study were identical to those used by Wiederhorn and Bolz7 and very similar
to that used by Michalske.9 Crack growth was measured on sodalimesilicate glass microscope slides, with dimensions of 75 mm 25 mm 1.5 mm. We used a batch of glass slides bought earlier for the experiments of Wiederhorn and Bolz.7 The specimens had double side-grooves to maintain the crack along the midline. Crack growth was observed by reflected light from the fracture surface using a 40 telescope. The specimens were dead-weight loaded using a small laboratory pan balance to transmit the load. Our experimental technique differed from that used by Michalske in that he used transmitted polarized light to observe crack motion and did not use side grooves for his specimens. We adapted Michalskes procedure for our experiments. Cracks were propagated at a relatively high value of the stress intensity factor, held at a lower value to promote blunting, and then repropagated at the higher value of KI. The time to repropagate the crack was measured by fitting a straight line to the crack length versus time data collected on the reloaded crack. The intercept of the fitted line with the time axis gave the time to restart the crack. Studies were all conducted in water. The high stress intensity factor for crack growth measurements ranged from about 0.35 to about 0.55 MPa m1/2. The lower stress intensity factors used to induce blunting were 0.22, 0.24, and 0.27 MPa m1/2; the hold time was 16 h. Once a crack started propagating, it was permitted to propagate from 50 to 1000 m and then again held at a lower stress intensity factor for 16 h. The restart stress intensity factors ranged from 0.35 to 0.55 MPa m1/2. All measurements were conducted on a single specimen immersed in water. A total of 23 crack arrest experiments were conducted on the slide. On completing the crack growth experiments, we broke our specimen in half along the crack plane to expose both fracture surfaces. We then examined and compared both surfaces with an optical microscope and an atomic force microscope (AFM). Despite the fact that we kept a record of the crack length and all changes in applied stress intensity factor, a one-to-one correlation between the features on the crack surface and the record of stress intensity factor changes was not always possible. Therefore, in
September 2002
2289
Fig. 3. Measurement of the time to restart a crack held below the apparent crack growth threshold. Plotted on this graph is the time to restart the crack versus the stress intensity factor that was applied to the crack. Data points with the arrows showed no detectable motion during the test.
Fig. 4. Optical photomicrographs of cracks held below the apparent crack growth threshold and then repropagated. This figure shows a crack front where light and dark bands were observed along the arrested crack front. It is worth noting that the bands appear to be irregularly distributed along the crack front.
examining the crack surface, we concentrated on those features that were similar to the ones reported by Michalske,9 and describe them in detail below. The instrument used for this study was a Digital III atomic force microscope, which accommodated our specimens: 13 mm high, 75 mm long, 1.5 mm wide. Before placing a specimen on the AFM stage, it was first examined in reflected light using a standard optical microscope (Leica Model DMRM, Materials Research Microscope). In so doing we were able to align the specimen normal to the optical beam so that the fracture surface was also normal to the AFM probe. Furthermore, a set of low-magnification micrographs was made that could be used as a guide in searching for interesting areas to be studied by AFM. Before carrying out the study, the fracture surfaces were cleaned with acetone and then air-dried. The specimen was scanned using the contact mode, which gave clear pictures of the fracture surfaces.
decorated by a few dark bands. In Fig. 6, an arrested crack consists only of a dark band (whether the band was light or dark depended on the setting of the Nomarski lens used to study them). Investigation of the arrested crack fronts by AFM shows that all of these features fit a consistent pattern.
(3) Atomic Force Microscopy Micrographs taken with the AFM indicate a variation in the height of the fracture surface along the front of the arrested crack. Figure 7, for example, shows a micrograph taken from a region that was similar to the one shown in Fig. 4. Crack propagation on the figure was from left to right. At the extreme left of the figure, crack growth was on a single plane. At the point where the load on the crack was reduced, the crack changed its pattern of motion. As indicated by the light and dark patches that formed, portions of the crack now lay 25 nm above the initial fracture plane, whereas other portions lay 25 nm below the original fracture plane. The
III.
Results
(1) Results of Crack Growth Measurements Experimental results were similar to those obtained earlier by Wiederhorn and Bolz7 and by Michalske.9 In Fig. 3 we plot the time to restart crack growth as a function of applied stress intensity factor. When the stress intensity factor to restart the crack was less than about 0.4 MPa m1/2, the times to restart the crack ranged from approximately 1000 s to greater than 6000 s. At higher restart stress intensity factors, 0.45 MPa m1/2 cracks restarted immediately. The negative restart times are a consequence of the uncertainty in the intercept of the line fitted data representing measurements of crack length versus time data. We believe that these negative values represent a zero restart time.
(2) Optical Micrographs of Crack Arrest Marks After breaking the double cantilever specimen in two and examining the fracture surface optically, we obtained a complete confirmation of the features seen by Michalske (Fig. 4). In both studies, light and dark bands were observed at points of crack arrest. We also saw a considerable variation in the shapes and frequency of the light and dark bands. Sometimes, only a small portion of the arrested crack front was decorated with these bands; sometimes the bands appeared to be missing. Thus, in Fig. 5, an arrested crack front consists of a light band that is sparsely
Fig. 5. This figure shows a crack front containing light bands separated by a few dark bands. The dark bands show no regularity along the arrested crack front and also vary in size. The portion marked by the square was examined by AFM in Fig. 9.
2290
Fig. 6. This figure shows a crack front containing only a single dark band. Aside from showing more contrast, it has much the same appearance as the front of a crack that was unloaded and immediately reloaded.
height changes appear to have occurred in alternative, but irregularly spaced, patches along the crack front. As crack growth continued from left to right, the height difference between the light and dark patches decreased until the crack again propagated on a single plane. A further analysis of the AFM data, by electronic sectioning, Fig. 8 shows that the crack changed its direction of propagation, presumably when the load on the crack was reduced. The crack propagated in either an upward or downward direction, locally changing the height of the crack surface compared with what it was before the load was reduced. The angle of propagation varied from about 3 to 5 to the original direction of propagation. The crack again changed its direction of propagation with a reapplication of the higher value of KI. The crack height went through a maximum and then decreased as the two portions of the branched crack again formed a single crack growth plane. Examination of the other crack fronts indicates a similar behavior when only a small portion of the crack is decorated with dark features (Fig. 5) or when only a light or a dark band is observed on the crack surface (Fig. 6). Figure 9 shows an AFM micrograph of a small portion of the crack front shown in Fig. 5. Again, crack propagation was from left to right. The surface in the vicinity of the point of crack arrest was curved. The same features
Fig. 8. Electronic sectioning of the image shown in Fig. 7. The section lines are shown on the small copy of the image located beneath the sectioning image.
that were observed in Fig. 5 are superimposed on the curvature. The few dark bands that were observed in Fig. 5 were associated with local increases in the height of the original fracture surface. The larger light band was associated with a local decrease in the
Fig. 7. An AFM scan of the area shown in Fig. 4. The bright areas indicate surfaces with heights greater than the average height, while the dark areas indicate surface heights that lie below the average height. The crack propagation direction is from left to right.
Fig. 9. An AFM scan and sectioning of a portion of the area shown in Fig. 5. The crack propagation direction is from left to right. The section curves have been displaced for greater clarity. The AFM scan with the section lines is shown on a small copy of the image below the section curves. The features examined are marked off by a white square in Fig. 5. As before, the bright areas indicate surfaces with heights greater than the average, while the dark areas indicate surface heights that lie below the average.
September 2002
2291
Fig. 11. Superposition of the two surfaces of a crack arrest feature on top of one another. The crack propagation direction is from left to right. Because the AFM views each surface from the opposite direction, one of the surfaces had to be flipped by 180 along the direction of crack propagation in order to make this comparison. The estimated vertical scatter of the measurement is 2 nm.
IV.
Discussion
Fig. 10. An AFM scan and sectioning of the area shown in Fig. 6. The crack propagation direction is from left to right. The AFM scan with the section lines is below the sectioned image. In this figure, the change in direction of crack growth is the same along the entire crack front.
height of the surface. Sectioning of the surface, also shown in Fig. 9, supports this interpretation. Again, the direction of crack propagation changed along the crack front. In the locality of the dark bands of Fig. 5, the crack growth is directed upward at an angle of 3 to 5, while in the vicinity of the light band, the crack is directed downward at about the same angle. The angle of crack growth again changed when the stress intensity factor was increased, and the inclined cracks rejoined each other to form a single crack surface again. When only one light band or one dark band was observed on the crack arrest front (Fig. 6), only one of the features reported above was seen by AFM sectioning (Fig. 10). In this figure, the surface was also curved in the direction of crack propagation. As before, the features at the point of crack arrest are superimposed on the general curvature. At the point of crack arrest, the crack changed its direction of propagation by an angle of 3 to 5 and then propagated at a slightly different angle when the load is again raised. Eventually, the crack resumed its original direction of growth. So, this figure also shows the change in direction that is characteristic of crack arrest; however, the change is the same along the entire front of crack arrest. In discussing the meaning of these features, it is important to know whether the top and bottom surfaces match or whether the stress corrosion process removed a portion of the crack surface during the 16 h hold period. To investigate this possibility, the identical feature on the upper and lower surface was sectioned. The section from the upper surface was then laid on the section from the lower surface after flipping it vertically so that up and down were the same for both sections. This procedure was repeated on a second feature with identical results. The result of this procedure is shown in Fig. 11. As can be seen, the upper and lower surfaces appear to match one another, to within an experimental scatter of 2 nm.
The shades in Figs. 5 and 9 were reversed. The shade in Fig. 5 depends on the adjustment of the Nomarski lens. In the AFM picture in Fig. 9, the shade depends on the height over the surface. Figures 5 and 9 were taken from the identical area of the fracture surface. Up and down are determined from the AFM picture.
This study shows that cracks in sodalimesilicate glass held below the apparent threshold limit change their direction of growth by an angle ranging from about 3 to 5 to the original crack growth direction. This change in crack propagation angle does not occur uniformly along the crack front. For some crack fronts, up and down portions of the crack alternate irregularly all along the crack front. For others, most of the crack is up or down with only a few small sections in the other direction. And finally, there are some cracks that change propagation direction in only one way, up or down. We suggest that this change in crack growth direction is a consequence of the reduction of applied stress intensity factor when the load on the cracked microscope slide is reduced, and occurs after the reduction in load. The AFM micrographs can be used to estimate the crack growth rate during the hold period. Assuming that the deviation of the crack during the hold period was about 5 and that the out of plane propagation distance of the crack was about 40 nm, typical in our studies, a crack growth rate of 40 109/[sin (5) 16 3600] 8 10 12 m/s is obtained. These velocities are consistent with those reported recently by Kocer and Collins.8 for a sodalime silica glass in 100% rh air at similar stress intensity factors. Changes in crack growth direction are often explained in terms of changes in the far-field load applied to the crack geometry. Thus, when the applied load on the double-cantilever beam geometry used in the present study is changed, the angle of the specimen to the applied forces changes slightly, imposing a small edge-sliding mode to the crack geometry, which in turn changes the crack growth direction. This is the reason why faint marks are left on the specimen surface each time the applied stress intensity is changed. Similarly, applying a torsional load (Mode III) to the crack causes the crack to twist out of plane.11,13 These modifications to the applied loads cannot explain the kinds of observations in the present study. The changes in crack tip geometry observed here are of the tilt kind, yet cannot be due to a far-field edge-sliding mode, because a remotely applied edge-sliding mode deformation would cause the entire crack to change its direction of growth (either up or down). Such a change in growth direction could not explain our observation that some portions of the crack go up while others go down, or the randomness of the change in crack growth direction. A modification of the near-field stress arising from fluctuations in the glass chemistry might also cause the crack to deviate in its path if residual stresses are associated with the chemical fluctuations. Such a variation in the glass structure might cause the crack to go up and down in some random manner along the crack front as shown in Fig. 4. However, the variation in structure could not easily explain why in some cases most of the crack goes up or
2292
down (Figs. 5 and 6). Because of these objections, we reject the ideas that either an applied stress or near-field stress modification due to fluctuations in the glass caused the changes in crack growth direction. Instead, we turn our attention to the stress corrosion process that is responsible for crack growth as the possible source of our observations. A possible explanation for the experimental observations of the present paper can be obtained from the ChuangFuller12 modification of the CharlesHillig5 model of crack growth. In both models, crack growth results from a stress-enhanced corrosion process at the crack tip. Both models treat the crack as an elliptical hole in the solid. Both assume that the corrosion rate, r, normal to the surface of the ellipse is determined by the following equation: r r 0 exp V
s
force and hence strain will be the ones to rupture during the crack growth process. Unlike crystals, there will be no favored direction for crack growth vis-a-vis the structure of the specimen. Second, ` glass has an open structure consisting of cross-linked SiO2 tetrahedral chains joined at their corners. Distances that are important in glass are the distance between the chains, which are the order of 0.5 nm. The breakage of bonds to form a crack will distort the structure of the crack tip to resemble that of a rounded surface similar to an ellipse. V.
10
Summary
/ /RT
(1)
where is the stress at the surface, V is an activation volume for crack growth, s is the surface tension of the water glass interface, is the radius of curvature of the interface, is a numerical constant, R is the universal gas constant, and T is the temperature. The coefficient to the equation, r0, is temperature dependent. Both Hillig and Charles6 and Chuang and Fuller12 considered other terms in the corrosion equation, but this is the simplest equation that explains our data. Furthermore, Eq. (1) can be derived directly from absolute rate theory and so has a basis in chemical reaction kinetics.14 Charles and Hillig5 developed their equation for the stress corrosion limit by considering the corrosion rate at the tip of a cracklike elliptical notch. Above a critical stress the crack tip sharpened, as the crack got longer, whereas below the critical stress, the crack tip became blunted. At a critical stress the crack maintained a constant radius while propagating. This stress was defined as the static fatigue limit, because below this stress, the crack became more blunt and eventually stopped growing. Chuang and Fuller12 modified the theory by calculating the change in crack shape all along the ellipse, so that the crack shape was determined entirely by the corrosion process. Chuang and Fuller only considered the initial stage in the corrosion process, since they assumed the original crack to be an elliptical shape, but did not iterate a change in crack shape as the corrosion proceeded. Nevertheless, they were able to see tendencies in the crack growth process that were not in the CharlesHillig treatment. In particular, they showed that when the stress is reduced below the blunting limit determined by the CharlesHillig analysis, corrosion occurs most quickly not at the major axis of the ellipse, but along its flanks. The maximum corrosion rate decreases and moves further from the tip of the notch as the load is reduced further. With regard to the present experiment, these calculations suggest that the crack does not stop growing below the estimated fatigue limit, but changes its direction and decreases its rate of growth. Positive and negative changes in angle should occur, thus accounting for the bifurcation in crack growth direction and the irregularity of the bifurcation. The principal drawback to the model is its continuum nature. The model does not take into account the molecular structure of the glass, especially near the crack tip where the changes in growth direction are predicted. Two features of glass should obviate this problem. First, glass has a random structure, so that at the crack tip, the bond subjected to the maximum force will vary along the crack perimeter. Those bonds suffering the largest deformation
This study duplicates the findings of Michalske9 and Gehrke et al. that cracks in sodalimesilicate glass held below the static fatigue limit experience a time delay in restarting their growth when the stress intensity factor is again raised above the fatigue limit. AFM results show that below the apparent fatigue limit, cracks change their direction of growth by an angle of about 3 to 5 to the original crack growth direction, resulting in a waviness to the crack front. Raising the stress intensity factor to a value that is greater than the fatigue limit again changes direction of crack growth and causes the crack to grow on a flat plane. The theoretical findings of Chuang and Fuller that the maximum direction of corrosion below the fatigue limit shifts to the flanks of the crack are used to rationalize the observed change in direction of crack growth. Acknowledgment
Discussions with Drs. B. R. Lawn, E. R. Fuller Jr., and T.-J. Chuang are appreciated.
References
R. J. Charles, A Review of Glass Strength; pp. 138 in Progress in Ceramic Science, Vol. 1. Pergamon Press, Oxford, U.K., 1961. 2 S. M. Wiederhorn, Subcritical Crack Growth in Ceramics; pp. 613 46 in Fracture Mechanics of Ceramics, Vol. 2. Edited by R. C. Bradt, D. P. H. Hasselman, and F. F. Lange. Plenum Publishing Co., New York, 1974. 3 S. M. Wiederhorn, H. Johnson, A. M. Diness, and A. H. Heuer, Fracture of Glass in Vacuum, J. Am. Ceram. Soc., 57 [8] 336 41 (1974). 4 A. C. Fischer-Cripps and R. E. Collins, Architectural Glazings: Design Standards and Failure Models, Build. Environ., 30 [1] 29 40 (1995). 5 R. J. Charles and W. B. Hillig; pp. 51127 in Proceedings of a Symposium on Mechanical Strength of Glass and Ways of Improving It (Florence, Italy, September 2529 1961). Union Scientifique Continentale du Verre, Charleroi, Belgium, 1962. 6 W. B. Hillig and R. J. Charles; pp. 682705 in High-Strength Materials. Edited by V. F. Zackey. Wiley, New York, 1965. 7 S. M. Wiederhorn and L. H. Bolz, Stress Corrosion and Static Fatigue of Glass, J. Am. Ceram. Soc., 53 [10] 543 48 (1970). 8 C. Kocer and R. E. Collins, Measurement of Very Slow Growth of Cracks in Glass, J. Am. Ceram. Soc., 84 [11] 258593 (2001). 9 T. A. Michalske, The Stress Corrosion Limit: Its Measurement and Implications; pp. 277 89 in Fracture Mechanics of Ceramics, Vol. 5, Surface Flaws, Statistics, and Microcracking. Edited by R.C. Bradt, A. G. Evans, D. P. H. Hasselman, and F. F. Lange. Plenum Press, New York, 1977. 10 E. Gehrke, Ch. Ullner, and M. Hahnert, Fatigue Limit and Crack Arrest in Alkali-Containing Silicate Glasses, J. Mater. Sci., 26, 544555 (1991). 11 E. Gehrke and Ch. Ullner, Application of Fractography of Glass to the Study of Environmental Effects on Crack Growth; pp. 77 84 in Advances in Ceramics, Vol. 22, Fractography of Glasses and Ceramics. Edited by V. D. Frechette and J. R. Varner. American Ceramic Society, Westerville, OH, 1988. 12 T.-J. Chuang and E. R. Fuller Jr., Extended CharlesHillig Theory for Stress Corrosion Cracking of Glass, J. Am. Ceram. Soc., 75 [3] 540 45 (1992). 13 E. Sommer, Formation of Fracture Lances in Glass, Eng. Fract. Mech, 1, 539 46 (1969). 14 S. M. Wiederhorn, E. R. Fuller Jr., and R. Thomson, Micromechanisms of Crack Growth in Ceramics and Glasses in Corrosive Environments, Met. Sci, 14, 450 58 (1980).
1
The crack velocity, v, and the corrosion rate, r, are equal at the major axis of the ellipse.