Fatigue behaviour of CMSX 2 superalloy [001] single crystals at high temperature II: Fatigue crack growth

Fatigue behaviour of CMSX 2 superalloy [001] single crystals at high temperature II: Fatigue crack growth

Materials Science and Engineering, A129 (1990) 55-64 55 Fatigue Behaviour of CMSX 2 Superalioy [001] Single Crystals at High Temperature II: Fatigue...

1MB Sizes 0 Downloads 6 Views

Materials Science and Engineering, A129 (1990) 55-64

55

Fatigue Behaviour of CMSX 2 Superalioy [001] Single Crystals at High Temperature II: Fatigue Crack Growth A. DEFRESNE* and L. REMY Centre des Matdriaux P. M. Fourt, Ecole des Mines de Paris, URA CNRS 866, BP 87, F-91003Evry Cedex (France)

(Received January 8, 1990; in revised form February 27, 1990)

Abstract

The fatigue crack growth behaviour of twodimensional cracks was studied in [001] single crystals of CMSX 2 nickel-based superalloy at 650 °C. A n influence of the crystallographic orientation of the crack growth direction on the crack growth kinetics was evident for short cracks tested in air. This influence was mainly attributed to differences in crack closure. Comparison between short and long cracks indicated that no short-crack effect exists in these single crystals. Tests in vacuum gave significant differences in crack growth rate and in fatigue fracture mode, especially at low rates. These results are discussed and used to rationalize previous observations. 1. Introduction

Turbine blades of advanced jet engines are now made of nickel-based superaUoy single crystals. These blades are submitted to severe loading along their main dimension which corresponds approximately to the [001] dendritic solidification direction. The blade root where the blade is attached to the disc is primarily a cyclically loaded notched member. A recent study on CMSX2 [1] investigated the influence of the secondary orientation of the notch (i.e. its crystallographic orientation in the (001) plane) on its low cycle fatigue (LCF) behaviour. A n orientation dependence of the LCF life of notched specimens at 650 °C was not clearly observed within the very large scatter of experimental data. The life of the specimens was considered to be mainly spent in the propagation of a microcrack from casting micropores, and the scatter in life was *Present address: CEA, SRMA, Saclay, France.

0921-5093/90/$3.50

attributed to variations in size and location of these initiation defects. For design purposes it is therefore necessary to determine the growth kinetics of such cracks and the eventual influence of the crystallographic secondary orientation on these kinetics. A number of studies have investigated the fatigue crack growth behaviour of superalloy single crystals at room temperature or at high temperature [2-9]. Little attention was paid, however, to the influence of secondary orientation on the fatigue crack growth rate (FCGR) of single crystals. Howland and Brown [2] used specimens made of SRR 99 single crystals with either a two-dimensional edge crack or a threedimensional corner crack at room temperature. The FCGR curves showed large scatter and the authors concluded the absence of any significant orientation dependence of the crack growth. Diboine et al. [5] used single-edge-notched specimens of PWA 1480 at 20 and 870 °C and observed a slight orientation dependence of the FCGR. Therefore the influence of the secondary orientation of the crack growth direction was studied in CMSX 2 [001] single crystals at 650 °C in air. The fracture mechanics specimens were used with a two-dimensional edge crack to avoid the ambiguity obtained with three-dimensional cracks. Since cracks may nucleate at micropores located either at the surface (involving growth in air) or internally (involving growth in vacuum), these experiments included comparative tests in air and vacuum. FCGR results are reported together with crack closure measurements since crack closure [9] has been shown to be a major parameter in the growth kinetics of long and short cracks in polycrystals [10-12]. Associated metallographic observations of crack paths are described. © Elsevier Sequoia/Printed in The Netherlands

56 2. E x p e r i m e n t a l

procedure

The CMSX 2 alloy composition is the same as in Part I [1]. Single crystals were supplied in the form of cast cylinders 20 mm in diameter and plates 12 mm in thickness with a [001] direction within 10 ° of the major axis. The propagation of long fatigue cracks was studied using compact tension (CT) specimens 40 mm in width and 8 mm in thickness. The propagation of short fatigue cracks was examined using precracked single-edge-notched (SEN) specimens by a test procedure explained below. The CMSX 2 heat treatment was performed before machining the specimens. The crystallographic orientation of the cast bars and plates was determined by a standard Lane back-reflection technique in order to machine the specimens with a specific crystallographic orientation. SEN specimens were machined from the bars to have a crack growth direction along the [010] or [110] direction. CT specimens were machined from the plates but the secondary orientation was imposed by solidification. These CT specimens had a crack growth direction along [010] or [210]. The orientation of every specimen was verified after final machining. For the sake of convenience the orientation of specimens will be labelled as [001]T[UV0]p where the subscripts T and P refer to tensile axis and crack propagation directions respectively. All specimen directions were within 10 ° of the specified orientation. The fatigue tests were conducted on a servohydraulic fatigue-testing machine at 650 °C using sine wave loading at a frequency between 2 and 50 Hz. Tests on other superaUoys showed that in this frequency range the F C G R curves are almost independent of the test frequency (see e.g. refs. 12 and 13). Most tests were run in air for two load ratios, R = 0.1 and 0.7. The secondary orientation dependence of crack growth rate was studied using artificial twodimensional short through-cracks by a procedure reported previously [11]. SEN specimens were fatigue precracked at growth rates down to about 1 0 - 9 m cycle -1 at 650°C. SEN specimens were 18 mm in width and 4 mm in thickness (Fig. 1). Precracks were grown about 2 mm ahead of the notch using the same load-shedding procedure as in CT specimens. The precracked SEN specimens were then machined down to about 11 mm in width and 2 mm in thickness. These machined

! --4-2

I .z

[

• !

Fig. 1. Single-edge-notched specimen geometry (dimensions in millimetres): (a) initially; (b) after machining to leave only a short crack.

SEN specimens contained a straight front through-crack 2 mm in width and about 0.2-0.4 mm in depth. The plasticity was kept to a minimum by the threshold procedure adopted and the gentle machining of specimens. These machined specimens were tested under constant load range (test frequency 2 Hz) during crack growth. A preliminary increasing load test procedure at 50 Hz was used starting from an initial value of AK = 1 MPa m 1/2 with steps of about 1 MPa m 1/2 until crack g r o w t h occurred. Some tests on long cracks in air were carried out using CT specimens. The conventional ASTM procedure was used for this type of specimen. A load-shedding procedure was adopted for the threshold measurement on the specimens using a 1%-2% decrease in load after crack growth of about 0.1 mm. The range of stress intensity factor AK was defined using Strawley's polynomial expression [14]. Several additional tests on CT specimens were carried out at 650 °C under vacuum. These tests were carried out in a chamber in which a pressure of less than 10 - 3 Pa was maintained throughout the test. Load ratios R = 0.1 and 0.5 were investigated under vacuum. The fatigue crack growth results were analysed using an isotropic stress intensity factor. Recent computations by Chan and Cruse have shown that the stress intensity factor depends on the crack angle in a CT specimen but not on the elastic anisotropy [15]. The ASTM stress intensity factor expression for isotropic CT specimens with a mode I crack was found to be applicable for M A R - M 2 0 0 single-crystal CT specimens with an inclined crack, provided the crack angle

57

was less than 30 ° and the projected crack length was used. Crack growth was monitored using a potential drop technique which was calibrated using optical measurements on both sides of the specimen. Crack closure was determined from crackmouth-opening displacement measurements with a clip gauge extensometer located on the front of the specimen. Load displacement (P-6) curves were recorded at low frequency (0.5 Hz) periodically. The crack-opening load Pop was determined by the upper break of the P-6 curve which indicated the crack was fully open [16]. This load Pop was more accurately defined using a corrected displacement 6 ' = aP-6 where a is an adjustable constant which was given by an electronic processor. The stress intensity factor at crack opening, Kop, was deduced from the opening load using the calibration formula for the relevant geometry. After completion of the experiments, the crack path on the specimen surface (SEN or CT) was observed in a scanning electron microscope (SEM). Some specimens were fractured at room temperature to enable SEM fractography. 3. Results

3.1. FCGR experiments using SEN specimens in air The F C G R at 650 °C in air is shown as a function of the stress intensity range AK for an SEN specimen with a crack along the [110] direction in Fig. 2. The initial crack length was 400 # m and i

b

i

iiii

I

i

r

CMSX2 SEN [OOl]T[llO]p T = 6 5 0 "C AIR o

E i0 -a

f

/

10 -7

the test frequency was 2 Hz. The same figure shows the F C G R as a function of the effective stress intensity range Ageff=Kmax-gop. The nominal curve da/dN-AK and the effective curve da/dN-AKef f display threshold behaviour for crack growth rates below 10 -8 m cycle -1. Such behaviour was previously reported in another cast superalloy, M A R - M 0 0 4 [17, 18], and in PM Astroloy [12] and was attributed to oxide blocking of crack growth [17]. Crack closure is rather important for this orientation of crack growth. The crack growth rate for a load ratio R = 0.7 is also plotted vs. A K in Fig. 2. No crack closure was experimentally detected here and this curve is in good agreement with the intrinsic curve da/dN-AKeff for R = 0.1. The F C G R at 650 °C in air is shown vs. A K for the [010] direction of crack growth in Fig. 3. The initial crack length was 450 and 280 # m for specimens B2B and B16 respectively. Good agreement is observed between the two specimens. No consistent measurement of crack closure could be obtained from the crack-mouthopening displacement. An experiment on a specimen with an initial crack length of 215 # m under a load ratio R = 0.7 (Fig. 3) was used to estimate the effective level of AK under R = 0.1, assuming a unique intrinsic curve da/dN-AK~f as was actually observed for the [110] crack orientation. Crack closure appears to be much less important for this orientation of the crack than for [110]e. Figures 4 and 5 compare the nominal and effective F C G R curves respectively of the two orientations for a load ratio R = 0.1. The crack

,"

-~ i0 -7

.L./

e~

o

10 -8

o

oo

\

,_

a



i0-9

Z~ i0 -9

CMSX2 SEN

lo-iO

10-10 0 1 0.7

i0-]i

i

]0

cl





i

i

[OOl]rCOlO]p

T=550'C

AIR

R O. 1 0.7

til~e¢ ~ z=v

t=K ~,v •

n

AK ( M P o ~ )

Fig. 2. Fatigue crack growth rate of short cracks in air at two load ratios R = 0.1 and 0.7 using [001IT[110]p SEN specimens vs. A K or AKeff.

2

AK ( M P , ~ )

Fig. 3. Fatigue crack growth rate of short cracks in air at two load ratios R = 0.1 and 0.7 using [001]x[010]p SEN specimens vs. A K or AKeef.

58 i

i

, ~1~,

T:650

,

I

c.sx~

,Jr i

'

t"

10 -8

'

~,3~

o

'# I

oB o

lO-e

E

10-9

~ "

~

.o/

10 -7

o%

1

'--7

,#-" |g

o

z \.

o

I

CMSX2

2 {

10-7

' , ....

T=650 "C R=O. I AIR o

"C R=O. i

AIR

#

, i

#,.¢

"

q3 l O-g

:t4

i0-io

i0-Io

a ££N 828 [O01]r[[310]F, • SEN 816 [ O O l ] r [ O l O ] m

• SEN 820 [OOl]r[llO]e

10 -11

~

,

,,,,I

,

2

i

10

i

i

CT02 [210]e SEN B20 [ l l 0 ] p i

kK ( M P o ~ )

Fig. 4. Fatigue crack growth rate of short cracks in air vs. A K for two different crack growth directions [010]p and [110]p for a load ratio R = 0.1.

I

,

, ,i,[

CMSX2 T=650 "C AIR

] 0 -7

o

\

E

/~

,~ ss

lO -e

o;

Z 72 \.

It

10 -9

io-io

10-11

i

R C001]r[ll0]p

0.1 •

[001]r[010lp



i

i

i

IIIL

0,7 o 4

i

i0

AKe~£

i

i

i

(MPc iXFm)

Fig. 5. Fatigue crack growth rate of short cracks in air vs. AK,ff for two different crack growth directions [010]p and [110]p and two load ratios R =0.1 and 0.7.

growth rate is almost 10 times higher in the [010] direction than in the [110] direction for the same value of AK. However, when the intrinsic curves da/dN-AKef f are compared, the FCGRs of the two orientations are within a factor of 2. Therefore the influence of the secondary orientation on the FCGR curves is mainly associated with differences in closure. A short-crack effect has been widely reported in polycrystals in the literature [10-12, 19-22]. In numerous cases short cracks seem to grow at faster rates than longer cracks such as occur in CT specimens. In the present case experiments using cracks 0.2-0.5 mm deep in SEN specimens

i0 -]i

, ,, t,,,l 2

10

,



, ~ , AK (MPo~)

Fig. 6. Comparison of fatigue crack growth rate in air of short (SEN specimen) and long (CT specimen) cracks vs. A K or AKeff for a load ratio R = 0.1.

show no noticeable influence of crack depth on FCGR in conditions where such effects were reported previously [11, 12, 22]. Therefore CT specimens were used to study the behaviour of long cracks in these crystals. Tests were carried out only for the [210] crack direction intermediate between [010] and [110]. The FCGR is plotted against AK and experimental AKeff in Fig. 6 for a load ratio R = 0.1. For a crack growth rate smaller than 10 -~ m cycle-~ a threshold behaviour is observed. Significant closure occurs and the effective curve da/dN-AKeff for R = 0.1 was found to be in good agreement with the nominal curve da/dN-AK for R = 0.7. This intermediate [210] crack orientation of the CT specimens gives rise to an FCGR curve which is in good agreement with the curve for short cracks in SEN specimens for the [110] orientation (Fig. 6). The effective curve for the CT specimen agrees with that for the SEN specimens within experimental accuracy. These results support the view that there is no significant crack length dependence of the FCGR in single crystals, at least for cracks longer than 0.2 mm. A similar conclusion was drawn by Hicks and Brown from experiments on S R R 9 9 single crystals [23].

3.2. FCGR tests under vacuum Three CT specimens with a crack along the [010] direction were tested in vacuum for two load ratios, 0.1 and 0.5. The FCGR curves da/dN-AK almost fit a single curve whatever the load ratio

59 i

r

i

i

ii,

I

i

,

i

,

,

CMSX2 T - 6 5 0 "C VACUUM

Z~"~_

10-7

;

r

o

lO-e

4, "

+

?:I

"8~ 10-g

o",~"

Z

iI- i i iii

lo-ZO

°°"

ml R

6K

10-11

~

+

i

,

CT04

CT05

CT06

0.1 I

0.5 •

o. 1 ,i

i

,,rl

t

i

I

AK (MPa~)

10

Fig. 7. Fatigue crack growth rate of long cracks under vacuum at two load ratios R = 0.1 and 0.5 using [001]T[010]p CI" specimens vs. A K or A K e . .

,

r

~

i

,,,

I

,

r~

,

i

3.3. Metallographic observations

T = 6 5 0 "C

oI

Z x2

I0-°

°:x. o



10-10 CT VACUUM SEN AIR 10-11

,

, ,

, t , , I

mmo • ,

I0

oo

Av ,

J

,

AK (MPo',Fm)

Fig. 8. Comparison of fatigue crack growth rate in air (SEN specimen) and under vacuum (CT specimen) vs. A K or AKeu for a load ratio R = 0 . 1 and a crack growth direction along

[010]~.

(Fig. 7). Such a behaviour has often been observed in vacuum in various alloys and recently in Astroloy [24, 25, 15]. The intrinsic crack growth curve corrected for closure seems to fit a single straight line with no evidence of a threshold as was observed in air for the [110] orientation. The FCGR curves using either nominal or effective AK for short cracks in SEN specimens along [010] in air are compared with the results for CT specimens under vacuum in Fig. 8. The da/dN-AK curves for [010] short cracks (which correspond to an FCGR higher than 1 0 - 8 m cycle-1, i.e. above threshold) are in good agreement with those under vacuum. However, the da/ dN-AKeff curve in air is slightly above the intrinsic curve in vacuum. Therefore the major differences between tests in air and vacuum are the disappearance of the R dependence of FCGR and that of the threshold in vacuum, at least for the investigated range of stress intensity factor.

The surfaces of all SEN specimens with initially short cracks were observed by scanning electron microscopy. The crack path in [ll0]p specimens is rather smooth on a macroscopic scale (Fig. 9) at all load ratios. However, the crack path on the average perpendicular to the load direction is more irregular for [010] specimens. More crystallographic cracking is visible, which occurs along {111} slip planes as shown by slip band traces emanating from the crack at high AK (Fig. 10). At the end of the test the crack path significantly deviates from the horizontal plane. The range of low crack growth rates (below 5 x 10-9 m cycle-1) was mainly studied on the CT specimens. The [210]v specimens tested in air show a smooth crack surface along the {001} plane. Cracking on this plane was previously

Fig. 9. Crack path on the surface of a [001]T[110]p SEN specimen in air for a load ratio R = 0.1 (crack grows from left to fight).

60

Fig. 10. Crack path on the surface of a [001IT[010It SEN specimen in air for a load ratio R = 0.7 (crack grows from left to right).

Fig. 11. Fracture surface in air for a load ratio R=0.1 at a crack growth rate of about 10 -8 m cycle l (CT specimen; crack grows along the [210] direction from left to right).

r e p o r t e d b y V i n c e n t a n d R 6 m y in c o a r s e - g r a i n e d M A R - M 0 0 4 [17]. 7' precipitates r e v e a l e d by oxidation s h o w u p as s q u a r e s o n the flat f r a c t u r e surface (Fig. 11). T h e [010]p s p e c i m e n s tested in v a c u u m s h o w a f a c e t e d f r a c t u r e surface with facets along {111} slip planes (Fig. 12). T h e s a m e effect of e n v i r o n m e n t o n the c r a c k p a t h was previously r e p o r t e d on cast M A R - M 0 0 4 [18].

Fig. 12. Fracture surface under vacuum for a load ratio R=0.5 and a crack growth rate of about 10 -~° m cycle -~ (CT specimen; crack grows along the [010] direction from left to right).

4. Discussion This investigation of the influence of crack g r o w t h direction in [001] single crystals of C M S X 2 has s h o w n that the F C G R along the [110] direction is a l m o s t 10 times that along the [010] direction. T h i s influence of orientation is mainly an effect of closure, h o w e v e r , since the

61

intrinsic fatigue crack growth curve seems almost independent of orientation. In the range of da/dN higher than 5 x 10 -9 m cycle -~, the crack path in [110] specimens is much smoother than in [010] specimens. The specimens which show more crack closure exhibit less roughness of the fracture surface. This seems to be a paradox since numerous papers on polycrystals claim surface roughness to be a major cause of crack closure [26]. This was pointed out by Vincent and R6my previously using the coarse-grained superalloy MAR-M004. This alloy exhibited a transition from a {001} fracture surface (no surface roughness) to {111} crystallographic cracking (large surface roughness) with decreasing crack growth rate under vacuum [25]. The large grain size enabled the monitoring of crack growth in one grain after another. They observed that the transition from a surface with very little roughness ({001} surface) to a surface with large surface roughness (with {111} crystallographic cracking) gave rise to no change i n Kop level when this transition occurred in a single grain. Crack closure was modified only when the crack grew across a grain boundary. The paradox of more crack closure corresponding to less surface roughness in the present study could be due to several reasons. Surface roughness may have to be considered on a very microscopic scale. {111 } facets occurring for crack growth along the [010] direction may give rise to a smaller density of contact points behind the crack tip than a macroscopic flatter surface as for crack growth along the [110] direction (see Fig. 13). Oxide debris behind the crack tip may also induce crack closure and its formation kinetics might be dependent upon the crystallography of the reacting surfaces. Plasticity-induced closure may also be a predominant phenomenon as originally postulated by Elber [16] and modelled by Newman [27]. Residual stresses might be dependent upon the crystallographic orientation of the crack front, [110] and [010] corresponding to maximum and minimum Young's modulus in the (001) crack plane respectively. These considerations are speculative in the absence of computations of the stress-strain singularity at the tip of a growing crack in an anisotropic elastic-plastic medium. No significant difference between CT specimens with long cracks and SEN specimens with short cracks was observed. This finding is in agreement with previous results of Hicks and

Fig. 13. Details of the crack path in air for short cracks of different orientations: (a) SEN B16 [001]x[010]p, R=0.1, da/dN=l.4×lO -v m cycle-J; (b) SEN B20 [001]x[ll0] P, R=0.1, da/dN--2xlO 7 m cycle-I; (c) SEN B2 [00111-[llO]p,R=O.7, da/dN--4xlO Smcycle i

Brown on SRR 99 single crystals tested at room temperature [23]. This behaviour is rather different from that observed in polycrystalline superalloys, where a short-crack effect was found [10, 11]. Crack closure explains most of this shortcrack effect in polycrystalline alloys, since Kop, the stress intensity factor at crack opening, was found to show a strong crack length dependence.

62 The variation of gop for a [001]T[11011, shortcrack specimen vs. crack length could suggest a crack length dependence of closure (Fig. 14). However, this apparent length dependence of Kop is due to the variation of Km~x (Fig. 15) since the ratio Kop/Kma x is constant whatever the crack length, i.e. the parameter U defined by Elber [16] as AKeyy/AKis constant (for a given value of R). The present experiments in air do not provide any information concerning the R dependence of U, since for a load ratio of 0.7 closure is no longer observed. Experiments on CT specimens [001]T[010]p under vacuum show that Kop/Kma x increases with R and that U seems to be independent of the load ratio (Fig. 16). The results using CT specimens have shown the influence of the environment on the FCGR in CMSX 2 single crystals at 650 °C. In vacuum the R dependence of the FCGR disappears and the threshold behaviour is absent. This gives strong

30

. . . .

. . . .

I

I

. . . .

CMSX2

SEN

I

B20

"C

T-650

. . . .

I

. . . .

I

. . . .

[OO|]T[IIO]p

R=O.

AIR

I

L•E2O 0 EL :E v

10

le

,

,

,

,

I

f

,

i

,

I

.5

. . . .

1

I

. . . .

1.5

2

2. a

(turn)

Fig. 14. Crack-opening stress intensity factor Kop for a short crack along the [110]p direction in air as a function of crack length a (specimen S E N B20).

support to an oxide-blocking mechanism producing a threshold in air as proposed previously for other superalloys [12, 17]. This difference in FCGR curves does correspond to a difference in fracture surfaces at low crack growth rates (below 5 x 10 -9 m cycle-I): a flat {001} fracture surface in air and a {111 }plane faceted fracture surface in vacuum. This transition was previously reported in MAR-M004 superalloy [18] and in Astroloy [13]. Faceted crack growth along {100} planes was also reported in an aluminium alloy by Garett and Knott [28]. They suggest that this can be achieved by a combination of displacements along {111}. The absence of {111} cracking in air implies an environmental effect. Oxygen can diffuse from the crack tip into the alloy. At low temperatures such as 650 °C it may diffuse along dislocations, pin dislocations and reduce their slip distance with respect to that in vacuum. Their movement along {111} planes might be restricted to the channel of matrix between X' precipitates and thus give rise to an average {001} fracture surface. Such an effect can be enhanced by interfacial dislocations at 7 - ) ,r interfaces which can act as diffusion short circuits. Pieraggi and Dabosi have shown that the Y phase and Y-7' interfaces of CMSX2 can oxidize rapidly [29]. Oxidation embrittlement could favour cracking along (001) channels of 7 matrix in this superalloy in air. The variation of fatigue fracture modes between tests in air and vacuum explains the peculiar features of the fracture surfaces of shallow notch specimens in Part I of this investigation [1]. When crack initiation occurred at an internal micropore, a circular area was often

i -' ' ' ' I ' ' ' ' I . . . . . . . .

I

. . . .

I

. . . .

I

. . . .

I

'

"

I ....

"

CMSX2 SEN B20 [OO]]r[liO]e T=650"C R=O. 1 AIR

.75

• 75

T~C

:nmlmlm



.5



.5 IE

~

T

-550 'C

°~ 25

.25

,~ Kop/KIIx •

0

,

0

,

,

,

I

.5

,

,

i

r

I

l

. . . .

I

1.5

. . . .

I 2

. . . .

a

2.5 (ram)

Fig. 15. Ratio of the crack-opening stress intensity factor to the maximum stress intensity factor, Kop/Kmax,as a function of crack length a for a short crack along the [110]p direction in air (specimen SEN B20 as in Fig. 14).

. . . .

I

• 25

,

,

i

U-

,

(l-Kop/K.=x)

I

,

,

.5

,

/ (l-R)

,

I

J

,

i

J

. 75 Load

ratio

Fig. 16. Ratio of the crack-opening stress intensity factor to the maximum stress intensity factor, Kop/K. . . . and Elber's U parameter vs. load ratio for a long crack along the [010]p direction under vacuum.

63

observed which was tangential to the notch root surface and the crack then grew into a semicircular shape from the outer surface. This circular shape of internal cracks (as well as the semieUiptical shape of external cracks) is worth emphasizing, since a marked orientation dependence was actually evidenced for two-dimensional crack growth. There is no anisotropic shape similar to that reported by Anton in an experimental single-crystal superalloy [30]. This implies that three-dimensional crack growth involves an averaging of crack increments in the various orientations along the crack front. Only small differences are therefore expected for specimens with three-dimensional corner cracks such as those used by Howland and Brown [2] or for natural penny-shape cracks initiating at a notch root. In the latter case only a weak influence of the eventual anisotropic stress field on fatigue crack growth is expected. This explains why variations in size and position of the crackinitiating micropore have masked any influence of notch secondary orientation on the LCF life of CMSX2 single crystals [1]. Furthermore, the FCGR curves determined in the present study for two-dimensional cracks can be used to compute the LCF life of these notched specimens provided a suitable anisotropic stress analysis is employed. Such an analysis is currently under way.

5. Conclusions This study of the fatigue crack growth behaviour of two-dimensional short cracks in [001] single crystals of CMSX2 superalloy at 650°C has shown a strong influence of secondary orientation on crack growth rate. Crack growth along the [110] direction proceeds at a much slower rate than in the [010] direction. This difference in FCGR is mainly due to variations in crack closure. Closure-corrected effective curves do not exhibit significant orientation dependence. Precise mechanisms for these differences remain to be identified. There is no significant difference between long and short cracks in air for [001] single crystals. For a given orientation the ratio of Kop t o Kma x is constant. The FCGR is different between air and vacuum. These differences, especially at low growth rates, are associated with different crystallographic fracture paths along {001} planes in air

and {111} planes in vacuum. This is in agreement with previous observations on superalloys.

Acknowledgments Single crystals of CMSX 2 were suppfied by Turbomeca. The authors are indebted to the Direction des Recherches, Etudes et Techniques (DRET) of the French Ministry of Defence for financial support of this work. They would like to thank the engineers of Turbomeca for their continuous interest in the present study.

References 1 A. Defresne and L. R6my, Mater. Sci. Eng., A129(1990) 45. 2 C. Howland and C. W. Brown, in C. J. Beevers (ed.), Fatigue 84, Proc. 2nd Int. Conf. on Fatigue and Fatigue Thresholds, Birmingham, September 3-7, 1984, Vol. II1, EMAS, Warley, 1984, pp. 1349-1359. 3 G.K. Bouse, in M. Gell, G. S. Kortovich, R. H. Bricknell, W. B. Kent and J. F. Radavich (eds.), Superalloys 1984, Proc. 5th Int. Symp. on Superalloys, The Metallurgical Society of AIME, Warrendale, PA, 1984, pp. 751-759. 4 J. S. Crompton and J. W. Martin, Metall. Trans. A, 15 (1984) 1711-1719. 5 A. Diboine, J. M. Peltier and R. M. Pelloux, in P. Bensussan et al. (eds.), High Temperature Fracture Mechanisms and Mechanics, Proc. Int. Seminar, Dourdan, October 13-15, 1987, MECAMAT, MoissyGramayel, pp. IV.71-95. 6 K. S. Chan and G. R. Leverant, Metall. Trans. A, 18 (1987) 593-602. 7 K. S. Chan, J. E. Hack and G. R. Leverant, Metall. Trans. A, 18 (1987) 581-591. 8 D. C. Wu, D. W. Cameron and D. W. Hoeppner, in D. N. Duhl, G. Maurer, S. Antolovich, C. Lund and S. Reichman (eds.), Superalloys 1988, Proc. 6th Int. Syrup. on Superalloys, The Metallurgical Society of AIME, Warrendale, PA, 1988, pp. 605-614. 9 J. Telesman and L. T. Ghosn, in D. N. Duhl, G. Maurer, S. Antolovich, C. Lund and S. Reichman (eds.), Superalloys 1988, Proc. 6th Int. Symp. on Superalloys, The Metallurgical Society of AIME, Warrendale, PA, 1988, pp. 615-624. 10 J. L. Breat, F. Mudry and A. Pineau, Fatigue Eng. Mater. Struct., 6 (1983) 349-358. 11 E Soniak and L. R6my, in K. J. Miller and E. R. de Los Rios (eds.), The Behaviour of Short Cracks, EGF PubL 1, Mechanical Engineering Publications, London, 1986, pp. 133-142. 12 F. Soniak and L. R6my, in R. O. Ritchie and E. A. Starke (eds.), Fatigue 87, Proc. 3rd Int. Conf. on Fatigue and Fatigue Thresholds, Charlottesville, Virginia, June 28-July 3, 1987, Vol. I, EMAS, Warley, 1987, pp. 351-36(I. 13 E Soniak, Thesis, Ecole des Mines de Paris, 1989. 14 J.E. Strawley, Int. J. Fract., 12 (1976) 475-476.

64 15 K. S. Chan and T. A. Cruse, Eng. Fract. Mech., 23 (1986) 863-874. 16 W. Elber, Damage Tolerance in Aircraft Structures, ASTM Spec. Tech. PubL 486 (1971) 230-242. 17 J. N. Vincent and L. R6my, in J. B~icklund, A. E Blom and C. J. Beevers (eds.), Fatigue Thresholds, Proc. Int. Conf., Stockholm, June 1-3, 1981, Vol. I, EMAS, Warley, 1982, pp. 441-454. 18 J. N. Vincent and L. R6my, in S. R. Valluri (ed.), Advances in Fracture Research, Proc. 6th Int. Conf. on Fracture (1CF6), New Delhi, December 4-10, 1984, Vol. 4, Pergamon, Oxford, 1984, pp. 2513-2520. 19 S. Pearson, Eng. Fract. Mech., 7(1975) 235-247. 20 J. Lankford, Fatigue Eng. Mater. Struct., 5 (1982) 233-248. 21 J. Lankford and D. L. Davidson, in R. O. Ritchie and J. Lankford (eds.), Small Fatigue Cracks, Proc. 2nd Engineering Foundation Int. Conf/Workshop, Santa Barbara, CA, January 5-10, 1986, The Metallurgical Society of A1ME, Warrendale, PA, 1986, pp. 51-71. 22 C. W. Brown, J. E. King and M. A. Hicks, Met. Sci., 18 (1984) 374-380. 23 M. A. Hicks and C. W. Brown, in C. J. Beevers (ed.), Fatigue 84, Proc. 2nd Int. Conf. on Fatigue and Fatigue

24

25

26 27

28 29 30

Thresholds, Birmingham, September 3-7, 1984, Vol. III, EMAS, Warley, 1984, pp. 1337-1343. C. J. Beevers, in J. B~icklund, A. E Biota and C. J. Beevers (eds.), Fatigue Thresholds, Proe. Int. Conf., Stockholm, June 1-3, 1981, Vol. I, EMAS, Warley, 1982, pp. 257-275. J. N. Vincent and L. R~my, in C. J. Beevers (ed.), Fatigue 1984, Proc. 2nd Int. Conf. on Fatigue and Fatigue Thresholds, Birmingham, September 3-7, 1984, Vol. II, EMAS, Warley, 1984, pp. 697-706. S. Suresh and R. O. Ritchie, Metall. Trans. A, 13A (1982) 937-940. J.C. Newman Jr., in J. B. Chang and C. M. Hudoon (eds.), Methods and Models For Predicting Fatigue Crack Growth under Random Loading, ASTM Spec. Tech. Publ. 748 (1981) 53-84. G. G. Garett and J. E Knott, Acta Metall., 23 (1975) 841-848. B. Pieraggi and E Dabosi, Werkstoffe und Korrosion, 38 (1987) 584-590. D. L. Anton, in C. J. Beevers (ed.), Fatigue 84, Proc. 2nd Int. Conf. on Fatigue and Fatigue Thresholds, Birmingham, September 3-7, 1984, Vol. III, EMAS, Warley, 1984, pp. 1361-1369.