The application of surface science to fatigue: The role of surface chemistry and surface modification in fatigue crack initiation in silver single crystals

The application of surface science to fatigue: The role of surface chemistry and surface modification in fatigue crack initiation in silver single crystals

Acta metall, mater. Vol. 40, No. 10, pp. 2769-2780, 1992 Printed in Great Britain 0956-715192 $5.00 + 0.00 Pergamon Pre~s Lid THE APPLICATION OF SUR...

1MB Sizes 0 Downloads 16 Views

Acta metall, mater. Vol. 40, No. 10, pp. 2769-2780, 1992 Printed in Great Britain

0956-715192 $5.00 + 0.00 Pergamon Pre~s Lid

THE APPLICATION OF SURFACE SCIENCE TO FATIGUE: THE ROLE OF SURFACE CHEMISTRY A N D SURFACE MODIFICATION IN FATIGUE CRACK INITIATION IN SILVER SINGLE CRYSTALS T. S. SRIRAM, M. E. FINE and Y. W. C H U N G Department of Materials Science and Engineering, gobert R. McCormick School of Engineering and Applied Science, Northwestern University, Evanston IL 60208, U.S.A. (Received 25 September 1991)

Abstract--The role of surface chemistry in the initiation of fatigue cracks was studied for silver single crystals, Fatigue tests up to crack initiation were carried out on pure silver specimens and specimens surface-alloyed with gold. The ambient used was oxygen. Concurrent investigations of the effect of gold on oxygen adsorption on Ag(111) surfaces were carried out using X-ray photoelectron spectroscopy. A strong correlation was observed between the number of cycles to crack initiation and the reactivity of slip steps exposed by the fatigue process as well as the rate of oxygen adsorption on the exposed slip steps. It was also found that oxygen adsorbed on the slip steps was transported into subsurface regions during fatigue. When the same test was performed in bromine, subsurface transport of bromine was not observed: yet the fatigue life up to crack initiation was found to be the same as that in oxygen. The absence of bromine transport is likely due to its larger atomic size. Based on these studies, we conclude that the dominant mechanism for accelerating fatigue crack initiation in an active environment is the reduction of dislocation reversibility by strong adsorption of ambient species on exposed slip steps. R6sumg--On 6tudie le r61e de la chimie superficielle dans l'initiation des fissures de fatigue darts des monocristaux d'argent. Des essais de fatigue jusqu'fi l'initiation de fissures sont effectu6s sur des ~chantillons d'argent pur et sur des ~chantillons alli6s superficiellement fi l'or. Le milieu est l'oxyg6ne. Des 6tudes simultan6es de l'effet de l'or sur l'adsorption d'oxyg6ne sur des surfaces (111) d'argent sont effectu~es par spectroscopic de photoblectrons X, On observe une forte corrdlation entre le nombre de cycles jusqu'fi l'initiation de fissures et la vitesse d'adsorption de l'oxyg~ne sur les marches de glissement expos6es. On trouve aussi que l'oxyg~ne adsorb6 sur les marches de glissement est transportb dans les rdgions situbes sous la surface pendant la fatigue. Quand le m6me essai est effectud dans le brome, le transport sous la surface n'est pas observ6; cependant la dur6e de vie jusqu'~i l'initiation de fissures est la m~me que dans l'oxyg~ne. L'absence de transport de brome est probablement dfi fi son plus grand diam6tre atomique. En se basant sur ces etudes, on conclut que le m6canisme dominant pour accdl6rer l'initiation des fissures de fatigue dans un environnement actif est la r6duction du mouvement r6versible des dislocations par une forte adsorption des espaces ambiantes sur le marches de glissement expos6es.

Zusammenfassung--Die Rolle der Oberfl/ichenchemie bei der Bildung yon Ermfidungsrissen wird an Silbereinkristallen studiert. Ermiidungversuche werden bis zum Auftreten von Rissen an reinen Silbereinkristallen und an Kristallen mit Gold-legierter Oberfl/ichenschicht ausgefiihrt. Umgebungsatmosph/ire ist Sauerstoff. Gleichzeitige Untersuchungen des Einflusses yon Gold auf die SauerstoffAdsorption auf (lll)-Ag-Oberfl/ichen werden mittels R6ntgen-Photoelektronenspektroskopie durchgef/ihrt. Zwischen der Zahl der Zyklen und dem RiBbeginn, der Reaktivitfit dcr durch die Ermfidung enstandenen Gleitstufen und der Sauerstoff-Adsorptionsrate an diesen Stufen wird eine enge Korrelation beobachtet. Augerdem wird beobachtet, dab der an den Stufen adsorbierte Sauerstoff w/ihrend der Ermfidung in Bereiche unterhalb der Oberfl/iche transportiert wird. Wird derselbe Versuch in Brom durchgeffihrt, dann findet sich kein solcher Transport von Brom; die Standzeit bis zur RiBbildung jedoch ist wie bei Sauerstoff. Wahrscheinlich wird Brom wegen seines gr6Beren Atomdurchmessers nicht transportiert. Aus diesen Untersuchungen folgern wit, dab der wesentliche, die RiBentstehung beschleunigende Mechanismus in einer aktiven Umgebung darin besteht, dab die Reversibilit~t der Versetzungsbewegung durch starke Absorption der Umgebungsatome an den Gleitstufen verringert wird.

INTRODUCTION AND BACKGROUND The presence o f a chemically active e n v i r o n m e n t is k n o w n to have a p r o f o u n d effect o n the fatigue b e h a v i o r o f m o s t metals [1, 2]. Periodic polishing o f the surface d u r i n g a fatigue test is k n o w n to p r o l o n g fatigue life indefinitely [l], indicating t h a t fatigue

d a m a g e generally starts at the surface. Chemical interaction o f the e n v i r o n m e n t with the surface o f the material is therefore primarily responsible for the acceleration of fatigue damage. Several m e c h a n i s m s have been p r o p o s e d to explain the role o f a gaseous e n v i r o n m e n t on fatigue behavior. These are discussed at length in an excellent

2769

2770

SRIRAM et al.: FATIGUE CRACK INITIATION IN SILVER CRYSTALS

review by Sudarshan and Louthan [3]. Of the models proposed, two appear to be the most promising. They are the "rewelding" model proposed by Thompson et al. [1] and the "debris layer" model proposed by Shen et al. [4]. Thompson et aL [1] proposed that gas adsorption on slip steps prevents their rewelding during subsequent cyclic deformation, thereby opening up a crack. A second mechanism suggested by the same authors along with Fujita [2], was that the adsorbed molecules are transported into the slip bands by dislocation activity. This would lead to obstruction of dislocation motion resulting in accelerated crack initiation. Shen et al. [4] dispute the rewelding model based on their investigations of frequency effects on the fatigue of aluminum as a function of oxygen pressure. They found that in intermediate partial pressures of oxygen, fatigue life decreased at higher frequencies, contrary to what would be expected from the rewelding model. According to the rewelding model, where oxidation is the rate controlling step, higher frequencies should enhance fatigue life. Shen et al. [4] proposed that in an active environment, dislocation accumulation in the subsurface regions is enhanced due to the presence of a thicker oxide. This near surface accumulation of dislocations is referred to as the "debris layer". Formation of cavities and voids is accelerated in this debris layer, leading to accelerated crack initiation. Thus the debris layer model explains accelerated crack nucleation on the basis of greater difficulty in the egress of dislocations. However, other experiments on frequency effects by Hordon [5] reveal a frequency dependence quite the opposite of that observed by Shen et al. [4]. The increase in fatigue resistance in vacuum was attributed to the retardation of oxygen and water adsorption at crack surfaces. A similar frequency dependence has also been observed by Snowden [6] in the fatigue of lead under different ambient pressures. Thus the results of Shen et al. [4] appear to be anomalous. Nevertheless, the model proposed by these authors explaining environmental effects on fatigue behavior has gained acceptance in some circles. For instance, it is the model preferred by Sudarshan and Louthan [3] in their review. The aforementioned studies concentrate on environmental effects on the entire fatigue process up to failure. The study of environmental effects on crack initiation alone is complicated by the difficulty in separating the initiation and propagation regimes. Cracks must be distinguished from intrusions which also develop in the early stages of fatigue. It is also not clear whether cracks grow from zero length ("barrierless" initiation) or if there is a crack nucleation stage in the fatigue process. Models such as the random slip model of Cheng and Laird [7], and the "unzippering" model of Neumann [8] predict that cracks result from the natural roughening of the surface which eventually leads to stress concentrations large enough to propagate a crack, In such

a scenario, the definition of initiation would depend on the resolution of the technique used to examine the surface. According to these authors, initiation is defined to occur when cracks of a certain size are detected. Mura and coworkers [9-12] have supported the idea that a crack nucleation stage exists in the fatigue process. Venkataraman et al. [11] used a free-energy criterion to predict crack nucleation in a slip band consisting of a pileup of edge dipoles. This model was later refined to include multiple slip bands and effects of random slip [12]. The criterion for crack initiation was modified such that a critical free energy density was required to initiate cracks. Environmental effects were introduced as variations in slip irreversibility. This model was able to predict S-N curves for initiation, a Coffin-Manson law for initiation and sizes of initiated cracks, all of which are in fairly good agreement with available data on copper. Experimental investigations of environmental effects on crack initiation must contend with first identifying cracks and then analyzing effects of environment. Hunsche and Neumann [13] devised a micromilling technique to examine the surface of a fatigued specimen in cross-section. They differentiated cracks from intrusions based on the included angle between the faces. They concluded that crack initiation in copper single crystals was delayed considerably in vacuum in comparison to laboratory air. The mechanism proposed by Thompson et al. [1] was invoked to explain the environmental effect. The cracks observed by these authors were well-developed stage I cracks. Kwon et al. [14] studied temperature and environmental effects on crack initiation in Cu single crystals. Cracks were identified using a replica technique with the smallest identifiable crack depth being of the order of 1000nm. Specimens were examined for cracks at a much earlier stage than in the work of Hunsche and Neumann [13]. The cracks identified were pit-shaped and spaced regularly along slip bands with a morphology similar to that found in numerous other investigations (e.g. [15]). The cross-sectional technique of Neumann [13] does not permit identification of such cracks. Kwon et al. [14] found that testing in an inert environment at room temperature led to a five-fold increase in the number of cycles to crack initiation (n~). One difficulty with the replica technique used is the possible existence of smaller cracks, thereby leaving open the question of nucleation vs growth from zero length. Neither Hunsche and Neumann [13] nor Kwon et al. [14] attempted to identify specific chemical interactions that led to accelerated crack nucleation in air. The role of chemisorption in accelerating crack growth in gaseous environments has been investigated by Gangloff and Wei [16]. They showed that dissociative chemisorption of hydrogen and hydrocarbons could accelerate embrittlement by gaseous hydrogen. Gangloff [17] showed that coadsorption of

SRIRAM et al.:

FATIGUE CRACK INITIATION IN SILVER CRYSTALS

ethylene mitigated hydrogen embrittlement in a highstrength steel. This was attributed to hydrogenation of ethylene to ethane. These studies clearly demonstrate the importance of characterizing gas-surface interactions and correlating them to fatigue behavior. However, this correlation is difficult to obtain in studies on crack propagation because of the complicating effects of crack geometry. In the present work, the approach is to characterize the interaction of ambient gases with the slip plane using surface science techniques, and to correlate the results with fatigue tests in controlled environments. Identification of just-initiated cracks with the highest possible resolution is accomplished using a scanning tunneling microscope. The surface is examined at a much earlier stage of fatigue than in previous studies of crack initiation. The resolution afforded by the STM enables us to determine ni with much greater accuracy than previously possible. Consequently, environmental effects can be evaluated more accurately. The material chosen for this investigation is silver. Single crystals oriented for single glide were used in order to eliminate the effects of grain boundaries and multiple slip. Oxygen was chosen as the active environment. Oxygen is known to chemisorb dissociatively on A g ( l l l ) at temperatures above 190 K with a sticking probability of 10-5-10 -6 and a heat of adsorption of 167.2kJ/mol [18]. At temperatures below ~ 170 K, oxygen is adsorbed in its molecular form with a much lower heat of adsorption. Over the same temperature range, Venkataraman et al. [19] showed that there is a dramatic increase in the number of cycles to crack initiation (nl) as the temperature is lowered. At temperatures below 170 K, there is no difference in ni between oxygen and inert environments. This is the first direct confirmation of the fact that strong chemisorption of the ambient species on exposed slip steps is essential for an environmental effect. The present work continues the investigation of mechanisms by which gas adsorption on slip steps accelerates crack initiation. Preliminary results from the present investigation have been published earlier [20, 21]. Experiments were designed to address several aspects of the problem. Fatigue tests up to initiation were conducted under a wide range of oxygen partial pressures to determine the extent of chemisorption required to influence crack initiation. Previous studies of ambient pressure effects on fatigue (e.g. [22]) found that environmental effects disappeared at pressures as high as 1 P a . This was attributed to geometric effects. The crack acts as a narrow pipe resulting in a pressure gradient along its length, allowing for the pressure at the crack tip to be several orders of magnitude smaller than the ambient pressure. The effects of crack geometry are irrelevant in the case of crack initiation because the crack does not yet exist. Tests were also conducted at different frequencies to determine the relative importance of mechanical vs

2771

environmental factors influencing crack initiation. The ambient pressure for these tests was chosen based on the results of the previous tests at different oxygen pressures, in order to maximize the variation in n, with frequency. Another question to be addressed is the fate of the adsorbed oxygen. One possibility is the transport of the adsorbate below the surface by the fatigue process. This has been seen for the case of crack propagation by Swanson and Marcus [23]. They obtained oxygen sputter profiles on fracture surfaces of A1, Ti and Monel fatigued in air and vacuum. It was seen that oxygen persisted to a greater depth in the specimens fatigued in air. In the present work, oxygen sputter profiles were obtained on samples fatigued under different oxygen pressures to investigate the extent of subsurface transport of oxygen during the fatigue process. Fatigue tests up to initiation were also conducted in a bromine environment to study the influence of subsurface transport on the crack initiation process. Bromine adsorbs dissociatively on A g ( l l l ) with a very high sticking probability of 0.75 [24]. Since Br has an atomic radius almost twice that of oxygen it is less likely to be transported into subsurface regions. Thus fatigue testing in a Br environment can provide information as to whether subsurface transport plays an important role in crack initiation. The reactivity of exposed slip steps can be varied by surface alloying. Greenfield [25] found that alloying the surface of copper single crystals with gold inhibits fatigue damage. We employed a similar technique to vary the reactivity of the near-surface regions of the fatigue specimens. Gold films of different thicknesses were evaporated onto the surface of silver specimens and diffused in by annealing. Gold does not dissociatively chemisorb oxygen at any temperature [26]. Therefore, the presence of gold on the exposed slip steps reduces the number of sites available for oxygen chemisorption. Thus the maximum oxygen coverage possible on the slip steps is a fraction of that of pure silver. The effect of gold alloying on the reactivity of Ag(111) toward oxygen was examined using XPS. The effect of environment on crack initiation was correlated to the reactivity of the slip plane toward oxygen for different film thicknesses. EXPERIMENTAL Silver single crystals of 99.999% purity were grown using the Bridgman technique. Fatigue specimens were cut from these crystals in the single glide orientation shown in Fig. 1. The specimens were mechanically polished to 0.3/~m and annealed at 700K in vacuum to relieve polishing stresses. Gold films were evaporated onto both sides of some specimens. Film thicknesses used were 0.1, 0.5 and 1.0/~m. The specimens were heated in situ in the evaporator to 700 K to remove the native oxide layer prior to gold

2772

SRIRAM et al.:

FATIGUE CRACK INITIATION IN SILVER CRYSTALS

[~4~]

J I I

i i

I I I I ,/t . . . . /

[321]

45.5°1 I

/

S

Fig. 1. Single glide orientation of fatigue specimens.

evaporation. The coated specimens were annealed in vacuum for 16 h at 725 K. Fatigue tests were performed under fully reversed total strain control with a strain amplitude of 0.1% and a frequency of 0.25 Hz. Tests were performed using a flange-mounted piezoelectric fatigue drive [27] which can be inserted into a vacuum chamber with a gas introduction manifold. Some of the tests were performed using an older servohydraulic machine [28] when the piezoelectric drive was not yet in operation. Pure silver specimens were tested up to crack initiation under oxygen partial pressures of 2.6 x 103, 0.13, 0.013 and 0.0013 Pa of oxygen and vacuum (1.3 x 10-6Pa). Under 0.013 Pa of oxygen, tests were also conducted at 0.5 and 1 Hz. Below 0.13 Pa, the pressure in the test chamber was maintained by continuous flow of oxygen from a leak valve while throttling the main pump to the chamber. The surface-alloyed specimens were tested under 2.6 x 103 and 1.3 x 10 -6 Pa of oxygen. A pure silver specimen was also tested under 0.0013 Pa of bromine. To ensure the cleanliness of bromine, we adopted the following procedure: 99.9% bromine liquid was frozen using a dry ice-methanol slush and the region above it was pumped to < 0.13 Pa and valved off. The bromine was allowed to warm up and equilibrate with its own vapor and was then introduced into the chamber through a variable leak valve. A scanning tunneling microscope was used to detect just-initiated cracks. To minimize effects of tip geometry, we employed tips with apex radii < 50 nm and an included angle of < 20 ° (see Fig. 2). Details of topography which are difficult to image using the SEM due to poor vertical resolution are dramatically resolved with this technique. Thus the STM can resolve just-nucleated cracks at the earliest stages of

fatigue. The specimens were fatigued up to a certain number of cycles and examined for cracks along slip bands. The criteria for identifying cracks are discussed in the following section. The number of cycles to initiation was determined to within a "window" of cycles which is the range of cycles over which cracks appear. Typical windows are 1000 cycles wide. The effect of gold on the reactivity of the Ag(111) surface toward oxygen was evaluated using X-ray photoelectron spectroscopy. An Ag(111) surface was cleaned in UHV following the procedure of Felter et al. [29]. The specimen was then cooled to 120 K and exposed to 4 x 105 Langmuirs of oxygen (1 Langmuir = 1.3 x 10 -4 Pa. s). XPS was performed at sample temperatures of 120, 163, 183, 200 and 300 K. Two monolayers of gold were evaporated onto the surface and the treatment and analysis were repeated after the specimen was annealed at 675 K for periods ranging from 0 to 10 min. The equipment used was a PHI 548 system with a double-pass cylindrical mirror electron energy analyzer and a dual anode X-ray source. Mg K~ X-rays were used for all experiments. The base pressure of the chamber was 4 x 10 -s Pa. The initial exposure to oxygen was done at a low temperature because of the low sticking probability of oxygen on Ag(111) at room temperature. As the specimen is warmed, part of the oxygen desorbs and part of it dissociates on the surface. The fraction of oxygen retained on the surface at room temperature is used as a measure of the reactivity of the surface.

Fig. 2. Pt-Ir tip used for STM imaging.

SRIRAM et al.: FATIGUE CRACK INITIATION IN SILVER CRYSTALS Oxygen Auger sputter profiles were obtained ex situ from the surfaces of pure silver specimens fatigued under various partial pressures of oxygen and under 0.0013 Pa of bromine. All the specimens were oriented in the same manner with respect to the ion gun to minimize shadowing effects. Point Auger analyses were also done on some of the specimens as a function of sputtering time, on and away from slip bands. RESULTS AND DISCUSSION (a) Detection of cracks

The key to evaluating environmental effects on crack initiation is the identification of just-initiated fatigue cracks. Cracks must be distinguislmd from other topographic features. Previous studies have used crack size [14] and shape [8] as determining criteria to assign a value for n,, i.e. in order for a feature to be identified as a crack, it must exceed a certain depth [14] or should have an included angle that is infinitesimally small [8]. In the first case, the choice of defining crack depth is dictated primarily by the resolution of the instrument used to observe them. Thus, features smaller than the resolution limit cannot be found leading to possible inaccuracy in the measurement of n,. The latter criterion for identifying cracks, using the included angle, is applicable only when the crack is well into the propagation regime. This is because the specimens have to be sectioned in order to observe the included angle. Only those cracks that have propagated can be distinguished l¥om other surface features such as intrusions. The present work avoids these difficulties by the use of the STM. The topographic features identified as cracks using STM satisfied the following criteria: 1. They were invariably associated with slip bands. 2. They were pit- or arrowhead-shaped, typically ~ 1/~m in length along the slip bands and ~0.1 ~m deep. They were easily distinguishable from intrusions which were several tens of microns long. 3. Unlike intrusions or extrusions, they did not appear until a certain number of cycles was reached, which depended on the environment. The use of STM makes the observation of very shallow cracks possible. The lateral resolution of the STM is limited only by the tip geometry. It was found that the cracks found were several times larger than the lateral resolution limit indicating that the tips used were quite adequate for these observations. A typical just-initiated crack along a slip band is shown in the STM image of Fig. 3. This is a topographic image displayed as an isometric surface plot, The usual procedure followed to detect cracks is to find a slip band and traverse its length. This AMM

4 0 I(,~



2773

procedure is repeated for several ( ~ 10) bands until a crack is seen. If no cracks are seen, the specimen is cycled further. A typical slip band without cracks is shown in Fig. 4. Slip bands can be intrusions, extrusions or intrusion-extrusion pairs. (b) Effects of oxygen partial pressure initiation

on crack

Figure 5 shows the variation in n, with oxygen partial pressure for pure silver cycled at 0.25 Hz. Also shown are n, for tests at 0.5 and I Hz for oxygen pressure of 0.013 Pa. At 0.25 Hz, there is a transition in n~ between 0.13 and 0.0013 Pa. On either side of this pressure range, n, is not strongly dependent on oxygen partial pressure. This behavior can be explained in terms of the number of cycles to form a monolayer of chemisorbed oxygen on exposed slip steps. Table 1 shows a comparison of the number of cycles for monolayer coverage (nml) o n a slip step under different oxygen pressures and the corresponding nl values, nmj was calculated based on the known value of the sticking probability of oxygen on A g ( l l l ) between 10-5-10 -6 [30]. At pressures below 10 --3 Pat nmj is much larger than n,. At these pressures, dislocation processes lead to crack initiation before environment can influence the process. For pressures greater than 0.1 Pa, a monolayer of chemisorbed oxygen forms very rapidly on the slip step. Therefore, the effect of environment saturates as the coverage of the slip steps saturates and there is no further decrease in n~ with increasing oxygen pressure. At intermediate pressures, nm~ is of the order of n,. Figure 5 also shows the effect of varying test frequency on n, at 0.013 Pa. It is clear that at higher frequencies, n, is larger. The ratio n,/nm~is an indicator of the relative importance of environmental and mechanical factors in crack initiation. The lower the value of n~/nml, the smaller is the effect of environment. This dependence is valid only for those conditions where environmental effects or mechanical effects are not saturated. At 0.013 Pa it is seen that this ratio decreases with increasing frequency, indicating that the effect of environment is attenuated. This is in contrast to the observations of Shen et al. [4], who found that increasing frequencies reduced the fatigue life of aluminum. Our observations indicated that the debris layer model of Shen et al. [4] cannot be used to explain the environmental effects observed on silver. Further evidence of this is the observed dependence of n, on oxygen pressure. Since silver does not form a multilayer oxide below 6.5 kPa of oxygen [30], for the fatigue tests described here, the oxide layer thickness cannot be greater than a monolayer. Thus, the debris layer model cannot account for the environmental effects observed. Figure 6 shows SEM images of the overall morphology of specimens fatigued in 2.0 kPa of oxygen and in vacuum. The slip bands formed in oxygen are seen to be more widely spaced. Both specimens have been fatigued up to initiation. The dependence of

2774

SRIRAM et al.: FATIGUE CRACK INITIATION IN SILVER CRYSTALS

Fig. 3. STM image showing a just-initiated crack. The image is shown in an isometric representation with dark shading indicating low regions and bright shading indicating high regions of the topography. The crack is along the edge of the slip band near the center of the image.

slip band spacing on environment has been shown theoretically to be related to dislocation reversibility. The details of the evolution of slip bands and initiation of fatigue cracks in silver will be discussed in a separate publication [31].

(c) Role of surface reactivity in crack initiation The variation in the reactivity of Ag(111) with gold content on the surface was investigated using XPS. The reactivity of a clean Ag(111) surface was first evaluated. Figure 7 shows the O(ls) photoelectron peak as a function of temperature after a clean A g ( l l l ) surface was exposed to 4 x 105 Langmuirs (1 Langmuir = 1.3 x 10 -4 Pa.s) of oxygen at 120 K. The binding energy is referenced to the Ag 3d5/2 peak at 367.9eV. The O(ls) level shows a shift towards smaller binding energies at 160-180 K. Using additional results from high resolution electron energy loss spectroscopy and thermal desorption

studies, Albers et al. [32] showed that this shift is due to the transformation of weakly bound molecular oxygen to strongly bound atomic oxygen. The XPS results on a clean A g ( l l l ) surface are used as a reference to evaluate the extent of dissociative chemisorption of oxygen on Au/Ag(111). O(ls) photoelectron peaks were obtained over the same range of temperatures for A g ( l l 1) with different concentrations of gold (XAu). XAu was varied by annealing for different periods of time after deposition of two monolayers of gold. The change in position and intensity of the O(ls) peak as a function of temperature for various annealing times, is shown in Fig. 8. For longer annealing times it can be seen that a greater fraction of oxygen is retained at room temperature. Since the O(ls) peak at 300K is due to chemisorbed oxygen, the variation in the room temperature oxygen peak corresponds to the variation in the reactivity of the

SRIRAM et al.:

FATIGUE CRACK INITIATION IN SILVER CRYSTALS

2775

Fig. 4. STM image showing slip bands before crack initiation.

surface. XAu can be c o m p u t e d as a function o f annealing time from available diffusion data [33]. Figure 9 shows a plot o f XAu VS the fraction o f oxygen retained on the surface, normalized by the value for pure silver. F r o m this plot one can estimate the saturation coverage o f oxygen on an A g ( l l l ) surface for different gold concentrations, and determine the gold coverage required to suppress oxygen dissociation. The results o f fatigue tests up to crack initiation on specimens with different average surface concen-

16000 --

12000 0

8000

rl 1.0

Hz.

0

4000

Oxygen pressure (Pal

Table 1 Numberof cycles for monolayer coverage (nml)

Numberof cycles for crack (initiation n,)

2600 0.13 0.013 0.0013 1.3 x 10 6

0.0025-0.025 25-250 25(~2500 2501~25000 2.5 x 106_2.5 x 107

2500 2500 4500 9000 ll000

0 IlJllm IIIH~

10 7

Iflllm IIIll~ IIIIli

10 -5

10 -3

IIIII '~ IIIIIm HIHI|i IIIHli

10 -1

10

I[llJ

103

O x y g e n partial p r e s s u r e (Pa)

Fig. 5. Plot of ni vs oxygen pressure at 0.25 Hz. At 0.013Pa, results from tests at 0.5 and I Hz are also indicated.

2776

SRIRAM et al.:

FATIGUE CRACK INITIATION IN SILVER CRYSTALS

Ag (111) with 2 M.L. Au annealed at 625 K

'

~

120 K 160 K 300 K

2 min

Fig. 6. SEM images of overall topography at crack initiation for specimens fatigued under 2.6 kPa and 1.3 x 10-6 Pa of oxygen. trations of gold are shown in Figure 10. The average surface concentrations for different initial film thicknesses were obtained using energy dispersive X-ray spectroscopy (EDXS). The diffusion distance of gold for the annealing time used is about 1-2 #m, which is of the same order as the depth from which X-rays are emitted. The upper curve shows minimal variation indicating that the variation of n i in vacuum for different values of XAu is not significant. This

.

120 K 160 K 180K

~

200 K ~

300 K I

538.5

I

I

533.5 528.5 523.5 Binding energy (eV)

I

518.5

Fig. 7. O(ls) binding ener~es on Ag(ltl) as a function of temperature.

533.5 528.5 523.5 Binding energy (eV)

518.5

Fig. 8. O(ls) binding energies as a function of temperature on Ag(111) coated with 2 monolayers of gold and annealed for various lengths of time.

Clean Ag(111)

E

538.5

establishes that alloying the surface with gold does not significantly affect near-surface deformation by solid solution strengthening. The lower curve shows that in oxygen, ni varies significantly with XAuThe environmental effect on ni is seen to disappear at the same gold concentrations at which oxygen dissociation is suppressed. Thus, the presence of gold on the slip planes results in suppression of environmental effects. The morphology of gold coated specimens is quite different from that of pure silver specimens. Figure 11 shows the morphology of slip bands and cracks for a specimen with XAu of 0.5. The slip bands are much less pronounced than for pure silver. A polycrystalline crust is present on the surface along whose grain boundaries cracks initiate. The cracks

SRIRAM et al.: FATIGUE CRACK INITIATION IN SILVER CRYSTALS 04

2777

r

Vacuum

~12000

"2 0.3

O

y-

m 8000 -g

Oxygen

o

4000

-x,, °

O O t'o

0'2 I

\ •

~'0.1

/



O

o 0 ,

I 0.4

i

I 0.8

A v e r a g e gold c o n c e n t r a t i o n

Fig. 9. Variation in n, as a function of surface alloying in vacuum and oxygen. The window for each data point is 1000 cycles wide.

do, however, form along the observed slip bands. Despite the differences in morphology observed, it is believed that the action of the gaseous environment is similar in the gold coated specimens. This can be deduced from the observed dependence of n i on environment. The surface grain structure does not

I 0,2

I 0.4

1 " ~ " " ~ ," 0.6

0.8

1.0

Surface concentration of gold Fig. 10. Variation in the ratio of O(ls) peaks at 120 and 300 K as a function of gold coverage. extend very far below the surface as evidenced by the observation of slip bands in the same orientation as pure silver.

(d) Role of subsurface transport The discussion so far has centered on the identification and characterization of gas-slip step interactions and their effects on fatigue behavior. The next step is to examine the mechanisms by which adsorbed

Fig. 11. SEM images of slip bands on surface alloyed specimens fatigued up to crack initiation. Lower image shows cracks forming along grain boundaries in the crust.

2778

SRIRAM et al.:

FATIGUE CRACK INITIATION IN SILVER CRYSTALS

gases influence near-surface slip. The mechanism proposed by Thompson et al. [1] involved the subsurface transport of the adsorbate and its subsequent interaction with near-surface dislocations. The existence of subsurface transport of oxygen was determined using Auger sputter profiling. Figure 12(a) shows the 0(503 eV)/Ag(356 eV) Auger peak ratios as a function of sputtering time for silver specimens fatigued until just prior to crack initiation. It is evident from the profiles that specimens fatigued under higher oxygen pressures show the presence of oxygen well after the native oxide has been sputtered off. For specimens fatigued under low pressures of oxygen the oxygen level drops off to zero after the native oxide is removed. Figure 12(b) shows point analyses on and away from a slip band as a function of sputtering time for the specimen fatigued under 2.6 kPa of oxygen. Clearly, subsurface transport of oxygen is confined to slip bands. The mechanism by which oxygen is transported below the surface is not clear. Oxygen is known to have a very high diffusivity in silver [34]. Short circuit diffusion along dislocation

(a) 0.6

a b c d e

0.4

2.6 x 103 0.13 0.013 1.3 x 10-3 1.3 x 10-6

Pa Pa Pa Pa Pa

0.2

.o

0

2¢ t~ 0.4 (D Q. O3

10

20

30

40

(b) CONCLUSIONS

<

a On slip band b Off slip band

0.3

0.2

~_..~.a

0.1

_~_--

I L~_~ 0

cores could also occur due to the high dislocation densities in the slip bands. Figure 13 schematically illustrates the processes that could occur upon adsorption of ambient species on exposed slip steps. This diagram is an atomistic reinterpretation of the model proposed by Thompson et al. [1]. One possible process leading to increased irreversibility is the blocking of dislocation motion by the adsorbate. Another possibility is that the adsorbate is transported into subsurface regions by the fatigue process. This could occur by short-circuit diffusion through dislocation cores or by formation of Cottrell atmospheres. The O-Ag bond energy (167.2 kJ/mol, which is the heat of adsorption of oxygen on Ag(ll 1) [18]) is significantly smaller than the cohesive energy of Ag (284.6 kJ/mol) [35]. Thus, crack formation could be facilitated along a plane of incorporated oxygen. To determine the relative importance of subsurface transport vs the blocking of returning dislocations by adsorbates, we conducted a fatigue test up to initiation in a bromine environment with the same operating conditions (strain amplitude = 0.1% and frequency = 0-25 Hz.). ni was found to be 1500 cycles, comparable to that in oxygen. This shows that bromine has a strong effect on crack initiation. Figure 14 shows a Br Auger sputter profile for a specimen fatigued in a bromine environment up to crack initiation. Point analyses on and off a slip band show that there is no significant transport of bromine below the surface. The suppression of bromine transport can be explained by its large atomic size. These results imply that the primary mechanism of accelerated crack initiation in bromine is the blocking of returning dislocations resulting from the presence of adsorbates on the exposed portion of slip planes.

5

~ 10

~-i"

I 15

--3 2O

Sputtering depth (nm) Fig. 12, (a) 0(503 eV)/Ag(356eV) Auger electron intensity ratios as a function of sputtering for specimens fatigued under various oxygen pressures. (b) O/Ag Auger eletron intensity ratios as a function of sputtering depths at points on and away from a slip band for specimen fatigued under 2.6 kPa oxygen.

1. Strong chemisorption of ambient species is necessary to accelerate fatigue crack initiation in gaseous environments. 2. Rapid adsorption of a monolayer of ambient gas is necessary to maximize the effect of environment. This is illustrated by the dependence of ni on oxygen pressure and test frequency. 3. The effect of a gaseous environment on fatigue crack initiation is limited by the reactivity of the exposed slip planes. The presence of gold on the slip planes in the near-surface regions strongly suppresses the effect of oxygen on fatigue crack initiation. 4. There is evidence that gases adsorbed on the slip planes can be transported beneath the surface during the fatigue process. This subsurface transport is not critical to the acceleration of crack initiation as shown by testing of silver in a bromine environment. While crack initiation is accelerated in the presence of bromine, there is no evidence that adsorbed bromine is transported below the surface. Thus, the most

SRIRAM et al.:

FATIGUE CRACK INITIATION IN SILVER CRYSTALS

2779

(a) 000000000000000 000000000000000 oo000000000000000 0000000±0000000~0 0000000 000000000 0000000 000000000 0000000000000000 0000000000000000

f

(b)

0000000000000000 0000000000000000 0000000000000000

o

o*oooooooooooo

0000000000000000

O~O00000TOOOOOTO00

---0000000000000000 0000000000000000 0000000000000000

(c)

0000000000000000 0000000000000000 0000000000000000

o'o°ooo'oo o o 0000

0000

ooooo 00000

ooo 000

OOOOOTO00000000000 0 00000000000000000 0000000000000000 0000000000000000 (d)

0000000000000000 0000000000000000

ooO

OOOOOOOOOOOOOO

oooo±ooo ooooo ooo O000OO000 0 0 0 0 0 OOO %o2oo ooo ooooo o 5 oooooooooooooooo OAg

o0

0000000000000000 0000000000000000

Fig. 13. Schematic illustration of the effect of adsorbates on dislocation reversibility.

i m p o r t a n t m e c h a n i s m t h a t leads to accelerated crack initiation is the blockage of r e t u r n i n g dislocations by adsorbates present o n the exposed p o r t i o n s o f slip planes.

0.05 I 0.04 ~-

o Acknowledgements--The authors would like to acknowledge the invaluable help of Ken Lehmann, Ken Sawusch and Mark Seniw in the preparation of specimens. Financial support from NSF in the form of grant no. DMR 9001662 is gratefully acknowledged.

0.03 13. 0')

<

0.02

09

0.01 •

0

REFERENCES

~ i

20

"

40

60

~ 1 1

I

I

80

1oo

Sputtering time (min) Fig. 14. Br/Ag Auger peak intensity ratio as a function of sputtering for specimen fatigued under 1.3 x 10 -4 Pa bromine.

1. N. Thompson, N. J. Wadsworth and N. Louat, Phil. Mag. 1, 113 (1956). 2. F. E. Fujita, Fract. Solids 20, 657 (1963). 3. T. S. Sudarshan and M. R. Louthan, Int. Mater. Rev. 32, 121 (1987). 4. H. Shen, S. E. Podlaseck and I. R. Kramer, Acta rnetall. 14, 341 (1966). 5. M. J. Hordon, Acta metall. 14, 1173 (1966). 6. K. U. Snowden, Phil. Mag. 10, 435 (1964).

2780

SRIRAM et al.:

FATIGUE CRACK INITIATION IN SILVER CRYSTALS

7. A. S. Cheng and C. Laird, Fatigue Engng Mater. Struct. 4, 331 (1981). 8. P. Neumann, in Physical Metallurgy (edited by R. W. Cahn and P. Haasen), Vol. 2, p. 1553 (1983). 9. K. Tanaka and T. Mura, J. appl. Mech. Trans. A.S.M.E. 48, 97 (1981). 10. M. R. Lin, M. E. Fine and T. Mura, Acta metall. 34, 619 (1986). 11. G. Venkataraman, Y. W. Chung, Y. Nakasone and T. Mura, Acta metall. 38, 31 (1990). 12. G. Venkataraman, Y. W. Chung and T. Mura, Acta metall. 39, 2621 (1991). 13. A. Hunsche and P. Neumann, Acta metall. 34, 207 (1986). 14. I. B. Kwon, M. E. Fine and J. Weertman, Acta metall. 37, 2937 (1989). 15. C. V. Cooper and M. E. Fine, Metall. Trans. 16A, 641 (1985). 16. R. P. Gangloff and R. P. Wei, Metall. Trans. 8A, 1043 (1977). 17. R. P. Gangloff, Basic Questions in Fatigue, Vol. II, ASTM STP 924 (edited by R. P. Wei and R. P. Gangloff). Am. Soc. for Testing and Materials, Pa (1988). 18. C. T. Campbell, Surf Sci. 157, 43 (1985). 19. G. Venkataraman, T. S. Sriram, M. E. Fine and Y. W. Chung, Scripta metall, mater. 24, 273 (1990). 20. T. S. Sriram, M. E. Fine and Y. W. Chung, Scripta metall, mater. 24, 279 (1990).

21. T. S. Sriram and Y. W. Chung, Proc. Morris E. Fine Symp. (edited by P. K. Liaw, H. L. Marcus, J. R. Weertman and J. S. Santner), p. 495. TMS (1991). 22. K. U. Snowden, Acta metall. 12, 295 (1964). 23. J, W. Swanson and H. L. Marcus, Metall Trans. 9A, 291 (1978). 24. P. J. Goddard, P. Schwaha and R. M. Lambert, Surf Sci. 71, 351 (1978). 25. I. G. Greenfield, Acta metall. 19, 857 (1971). 26. C. Backx, C. P. M. de Groot and P. Biloen, Surf Sci. 104, 300 (1981). 27. T. S. Sriram, M. E. Fine and Y. W. Chung, Rev. scient. lnstrum. 62, 2008 (1991). 28. W. J. Lee, M. E. Fine, Y. W. Chung and J. Baker, Rev. scient. Instrum. 57, 2854 (1986). 29. T. E. Felter, W. H. Weinberg, P. A. Zhdan and G. K. Boreskov, Surf Sci. 97, L313 (1980). 30. C. T. Campbell and M. T. Paffett, Surf Sci. 143, 517 (1984). 31. T. S. Sriram, M. E. Fine and Y.-W. Chung, to be published. 32. H. Albers, W. J. J. van der Wal and G. A. Bootsma, Surf Sci. 68, 47 (1977). 33. Smithells Metals Reference Book, 6th edn, pp. 13-32 (1983). 34. R. A. Outlaw and S. N. Sankaran, J. Mater. Res. 3, 1378 (1988). 35. CRC Handbook of Physics and Chemistry, 72nd edn, pp. 9 114 (1991-92).