ELSEVIER

Composites Science and Technology 51(1997) 697-702

0 1997 Elsevier Science Limited

PII: SO266-3538(97)00029-S

Printed in Northern Ireland. All rights reserved

0266-3538/97/$17.00

THE HIGH-TEMPERATURE CREEP BEHAVIOUR OF ALUMINIUM-MATRIX COMPOSITES REINFORCED WITH

Sic, A1203 AND TiB2 PARTICLES

S. C. Tjong & Z. Y. Ma

Department of Physics and Materials Science, City University of Hong Kong, 83 Tat Chee Avenue, Kowloon, Hong Kong

(Received 24 September 1996; revised 9 December 1996; accepted 16 December 1996)

Abstract Tensile creep tests have been carried out on particulate composites with matrices of aluminium and A1/3*2 wt% Cu alloy at 573 and 623 K. Particulate Sic was incorporated in these alloys which also contained particles of ALO.? and TiBz grown in situ during processing. The results showed that the pure aluminium-based composites exhibited an apparent stress exponent of 8.9 and 9.5, respectively, at 573 and 623 K, and an apparent activation energy of 177Wmol-‘. The composites based on the Al-Cu alloy exhibited an apparent stress exponent of 11.9 and 13.2 at 573 and 623 K, respectively, and an apparent activation energy of 323 kJ mol-‘. The creep resistance of the Al-Cu alloy composite was about two orders of magnitude higher than that of the pure aluminium composite. The creep data for the two composites were rationalised by using a substructure- invariant model with a stress exponent of 8 together with a threshold stress. The threshold stresses obtained from the extrapolation technique tended to decrease with increasing temperatures for these two composites. 0 1997 Elsevier Science Limited

Keywords: metal-matrix composites (MMCs), particle- reinforced composites, creep, aluminium

1 INTRODUCTION

Increasing interest is being shown in the development of metal-matrix composites (MMCs) consisting of aluminium containing discontinuous reinforcements for applications where high specific modulus and high specific strength are important aspects. The potential use of these composites for elevated-temperature applications has attracted many researchers to study their high-temperature creep behaviour.‘-lo Extremely high values of the apparent stress exponent, IZ, and the apparent activation energy, Q, have been reported for these aluminium-based MMCs. However, these values

697

are inconsistent with existing theoretical or phenome- nological models for dislocation creep and under- standing of the steady-state creep mechanism for the material is rather poor.

Nieh et a1.2V4 reported that composites consisting of 6061Al reinforced with Sic particles and whiskers exhibited very high values of y1 and Q. However, they give no theoretical or phenomenological justification for their results. Nardone and Strife” used the concept of effective stress, (+ - (+0 (where u is applied stress and (T,, is threshold stress), in the interpretation of creep data for an Al/Sic composite. They indicated that the high work-hardening rate of the composite and load transfer from the matrix to the ceramic reinforcement contribute to a high value of threshold stress. In this context, the steady-state creep data can be interpreted in a consistent manner with existing creep models for particle-strengthened alloys.3 Park et al.” and Mohamed et a1.9 investigated the creep behaviour of a 6061Al/SiC composite; they demonstr- ated that the composite exhibits a threshold stress behaviour and the true stress exponent of the minimum critical strain rate is equal 5. Mishra and Pandeyi’ re-analysed the steady-state creep data for 6061Al/SiC composite reported by Nieh et a1.2T4 and Morimoto et a1.,5 employing the substructure-invariant model developed by Sherby et al.12 On the basis of this model, they predict a stress exponent of 8 and a subgrain size exponent of 3 for the composites. In addition, the creep strain rate is controlled by matrix-lattice diffusion. By using this model, they have satisfactorily explained the steady-state creep be- haviour of Al/Sic and A1/TiB2 composites.7~s51” Similarly, Gonzalez-Doncel and Sherby13 indicated that the creep behaviour of Al/Sic composites can be described by substructure-invariant models. Very recently, Cadek et a1.14,15 re-analysed creep data for 30 vol% SiC,/Al, 20 ~01% SiC,/AlF 20 ~01% SiC,/2124Al, 30 vol% SiC,/6061A1,5 20 vol% SiC,/6061A14 and 26 ~01% A1203-fibre/Al-5 Mg16 composites. They found that the relationships between & ‘l/n and (T for these composites can be well fitted by

698 S. C. Tjong, Z. Y. Ma

straight lines with n = 5. They suggested that this stress exponent of 5 results from lattice-diffusion- controlled creep.

In this work, tensile creep tests have been carried out on composites containing particulate Sic in- corporated into aluminium and Al/Cu alloys with particles of A&O3 and TiB2 grown in situ. The aim was to investigate the operative creep mechanism of these composites.

2 EXPERIMENTAL

In this work, two composites containing particulate Sic, A&O3 and TiB2 were used. They were fabricated by the powder metallurgy route from Al-Sic-Ti02-B and Al-Sic-TiO*-B-CuO systems. Particulate Sic with a size of 3.5 ,um was used for additions while the A&O, and TiB2 particles were formed in situ through the reaction between aluminium, Ti02 and boron. In powder blending, the addition of raw materials was carried out so that the composites produced had the reinforcement contents shown in Table 1, provided that the in situ reaction between titanium and boron went to completion. For sample 2, the Al-3.2 wt% Cu alloy matrix was formed through the reaction between CuO and Al. The powders were ball-milled in alcohol for 8 h, and then dried. The cold-compacted powder mixture was heated to above 800°C in vacuum, held for lOmin, and then cooled down to 600°C and hot pressed. The as-pressed billet was extruded at 693 K at an extrusion ratio of 2O:l. As-extruded sample 1 was annealed at 773 K for 20 h, whereas sample 2 was given a T6 heat treatment (solutionised at 803 K for 2 h, water quenched, and then aged at 443 K for 5 h). X-ray diffraction analyses and metallographical observation were carried out on polished specimens 2.

Constant-load creep tests were conducted in air at 573 and 623 K. In order to determine the activation energy, the creep rate of sample 2 at 598 K and a constant stress of 60MPa was measured. Specimens with 4 mm gauge diameter and 20mm gauge length were used. The temperature of the sample, which was monitored by placing two separate thermocouples near the two ends of the gauge length, was controlled to within +1 K. The creep strain of samples was measured with linear variable displacement transdu- cers (LVDTs) connected to an amplifier.

Table 1. Designed composition of composites (~01%)

Sample Reinforcement Matrix l-IO.

Sic Al,@ TiBZ

1 10 4.70 5.30 80” 2 10 4.53 5.47 80b

n Pure Al matrix. ’ Al-3.2 wt% Cu alloy matrix.

Fig. 1. Optical micrograph of sample 2 ( X 800).

3 RESULTS

Figure 1 shows a typical structure of sample 2. The SIC particles and in situ A1203 and TiB2 particles are distributed uniformly in the aluminium alloy matrix. The Sic particles exhibit an angular shape, whereas the A&O1 and TiB, particles are equiaxed. The mean size of the A&O3 and TiB, is about 0.31 pm,” much smaller than that of the SIC. The X-ray diffraction analyses proved the existence of Al&u precipitates in sample 2 (Fig. 2). The formation of in situ phases in sample 2 was discussed in detail elsewhere.”

Figure 3 shows the steady-state creep rate, 8, of the two composites as a function of applied stress, (T, in double logarithmic coordinates. One can see that the steady-state creep rate of both composites varies linearly with applied stress. This plot yields apparent stress exponent values of 8.9 and 9.5 at 573 and 623 K,

30 40 50 60 70

2Wdeg)

Fig. 2. X-ray diffractograph of sample 2.

High-temperature creep behaviour

- 513 Kn= 8.9 I 10-3 c

i - 623 Kn =9.5

lo-4 z -+ 573Kn=11.9

%

s

tz

g c

Stress, MPa

Fig. 3. Variation of steady-state creep rate with applied stress for samples 1 and 2 (solid symbols, sample 1; open

symbols, sample 2).

respectively, for sample 1 and 11.9 and 13.2 at 573 and 623 K, respectively, for sample 2. Figure 4 shows the dependence of the steady-state creep rate on the reciprocal of temperature (103/T) for sample 1 at 36MPa and sample 2 at 60 MPa. The apparent activation energies calculated from the slopes of the straight lines are 177 and 323 kJ mol-’ for samples 1 and 2, respectively.

Figure .5(a-c) shows the relationship between t.“8, & **“, and i1’3 and applied stress for samples 1 and 2 in double linear coordinates. It can be seen that the relationship is linear only for the 8”’ versus u plot. This result indicates that samples 1 and 2 both obey the substructure-invariant model proposed by Sherby et al.,” i.e.

C = S(D,/b2)(A/b)3[(~ - cr,,)/E]’ (1)

where 8 is the steady-state creep rate, DL is the lattice

\ \

\ \ 0=60MPa

b=36MPa

Qa = 177 kJ/mol

‘L-4-a-J 1oi’.SS 1.60 1.65 1.70 1.75 1.80

1000/T, K-’

Fit 4. Arrhenius plot of steady-state creep rate versus 10 /T for samples 1 and 2 (solid symbols, sample 1; open

symbols, sample 2).

(a) 0.5

0.4

5 0.3

b e

*-w 0.2

0.1

- sample 1

-- sample 2

. 573 K

. 623 K

t 0.0’ ’ ’ ’ I

0 30 60 90 120 1

(b) 0.3

Stress, MPa

0 A sample 1

0 A sample 2

l 0 573 K

A A 623 K a

0 AA .

A 4 . 0

. . l

.A 0

A.@ o” 0.0 I t , ,

30 60 90 120 I

Stress, MPa

(C> 0.10

0.08 0 A sample. 1

0 A sample 2

l 0 573 K . A

A A 623 K

s 0.06

v1 .

c .w 0.04

0.02

0

AA .

0.00

.& l 0 .

l . 0 ..k 43 0

699

i0

0

10

Stress, MPa

Fig. 5. Variation of (a) L?‘“, (b) iIn and (c) ti1’3 with applied stress, (T, for samples 1 and 2 in double linear scales.

self-diffusion coefficient, A is the subgrain size, b is the Burgers vector, (T is applied stress, (T” is the threshold stress, E is the Young’s modulus and S is a constant. Thus the creep rate depends on the size of the subgrains which develop during creep tests.

Figure 6 shows the relationship between the

700 S. C. Tjong, 2. Y. Ma

"E 1011 A A 623 K

- 6 1010

’ W 109

10s

107

omT&E

Fig. 6. Variation of diffusivity-normalised steady-state creep rate, ri/DL, with modulus-compensated effective stress, (a - fl,,)lE, in double logarithmic coordinates (solid

symbols, sample 1; open symbols, sample 2).

steady-state creep rate, 8, normalised with respect to the matrix-lattice coefficient of diffusion, DL, with the effective stress, g - co, normalised with respect to the modulus, E, in double logarithmic coordinates. In the creep data calculation, values of the coefficient of lattice self-diffusion for aluminium were obtained from the relationship’s

& (m2 s-‘) = 1.71 X lo4 exp(-142_12/RT)

and values of Young’s modulus, E, from the equations’”

and

9G (MPa) = 3.0 X lo4 - 16T

E=ZG(l+v)

where T is temperature, R is the universal gas constant, and Y is Poisson’s ratio. One can see that the data points of samples 1 and 2 can both be fitted by straight lines with a slope of approximately 8. This means that the creep strain rates of two the composites are controlled by matrix-lattice diffusion. Alternatively, values of (i/D$‘s against (a - go)/E in double linear coordinates for samples 1 and 2 are shown in Fig. 7. The data points for the two composites at all temperatures can be well fitted by single straight lines and these straight lines nearly pass through the origin.

4 DISCUSSION

In the present work, the observed values of the apparent stress exponent range from 8.9 to 9.5 for sample 1, and from 11.9 to 13.2 for sample 2, while the values for apparent activation energy are 177 and 323 kJ mall’ for samples 1 and 2, respectively. These values are much higher than those predicted by any theoretical or phenomenological models for

- sample 1 -- sample 2

l 0 513 K P

A A 623 K

01 4 ’ , ’ * 1 ’ 0 4 8 12 16 :

Fig. 7. A plot of (k/D,)“* versus (u - a,,)/ E in double linear coordinates (solid symbols, sample 1; open symbols,

sample 2).

dislocation-climb-controlled creep. Such high values of the apparent stress exponent and apparent activation energy have been reported for Al/SiC composites,2-9 as mentioned earlier. The present work indicates that the apparent stress exponent for both samples tends to increase with increasing temperature. Similar be- haviour has been reported for 2124Al/2Ovol% Sic, composite3 and Al/Al,O,fibre composite. i6

We now consider the effect of the metal matrix on the composite creep behaviour. From Fig. 3, it is seen that the creep resistance of sample 2 is much higher than that of sample 1. This is attributed to the dispersion strengthening of Al&u precipitates in sample 2, though the precipitates have coarsened considerably at elevated temperat~e.2* Moreover, the apparent stress exponent and apparent activation energy of sample 2 are much higher than those of sample 1. The obviously high apparent activation energy in sample 2 may be associated with the formation of AlzCu precipitates. On the other hand, the estimated values of the apparent activation energy are affected by the stress level selected owing to the variation of the apparent stress exponent with temperature. If the steady-state creep rates at higher stress are used for estimating the apparent activation energy for sample 1, the estimated value of the apparent activation energy should increase.

Pandey et al. ‘J have suggested that the subgrain boundaries are pinned by the reinforcing particles in particulate MMCs, and the subgrain size remains constant with applied stress. They assumed that the average interparticle spacing is equal to the subgrain size. In this case, they can explain satisfactorily the creep data of AlfSiC and Al/TiBz composites’.“’ by using the substructure-invariant model proposed by Sherby et a1.12 To examine the validity of the substructure-invariant model for the present creep data, the steady-state creep rate raised to the

High-temperature creep behaviour 701

one-eighth power is plotted versus the applied stress (Fig. 5(a)). It appears that the relationship at 573 and 623 K is linear for both composites. The observed linearity between L? and (T suggests that the creep rate versus stress data follow a power-law behaviour for these composites. In addition, it is worthwhile examining the applicability of other models to the present composites. For example, Cadek et aZ.14.15 suggested that a lattice-diffusion-controlled creep model with a stress exponent of 5 fits the creep data of Al/Sic composites. In this context, the present data are plotted as the creep rate raised to the one-fifth power, g”‘, versus applied stress, (T, in double linear scales (Fig. 5(b)). These figures clearly indicate that the relationship is non-linear for these two composites at 573 and 623 K. Finally, the present data are also examined for evidence that they fit the viscous drag mechanism with a stress exponent of 3 (Fig. 5(c)). Apparently, the data for both composites at 573 and 623 K exhibit curvature, so that these models cannot be used to explain the present creep data.

The threshold stress for creep can be estimated from Fig. 4(a) by extrapolating the linear regression line to zero creep strain rate. The values of threshold stress obtained are listed in Table 2. It is seen that the threshold stress for both composites tends to decrease with increasing temperature. These results agree well with those for 17.4~01% SiC,/6061A1,‘2 Al/Sic and Al/A&O,-fibre composites.14’15 On the other hand, Pandey et af.x,2’ have reported that the threshold stress is independent of temperature for the 20 ~01% Sic, (1.7 pm)/Al and 10~01% SiC,/Al-4Mg com- posites. Gonzalez-Doncel and Sherby13 indicated that the modulus-compensated threshold stress decreases with increasing temperature for SiC-whisker- reinforced and particulate aluminium composites. However, there is no threshold stress at temperatures above 743 K. The modulus-compensated threshold stress for this work is also listed in Table 2. One can see that the modulus-compensated threshold stress decreases with increasing temperature, as reported by Gonzalez-Doncel and Sherby.13

The threshold stress for discontinuously reinforced aluminium-matrix composites has been the subject of discussion by a number of investigators.3’8.9*22 The threshold stress could originate from (1) dislocation- particle interaction, (2) internal stress associated with

Table 2. Values of threshold stress for composites determined by extrapolation technique

Sample T (K) u. (MPa) LTJE (X10-") IlO.

1 573 9.19 1.67 623 7-18 1.36

2 573 29.56 5.37 623 22,22 4.38

the subgrain size, or (3) load transfer from the matrix to the stiffer reinforcement phase.8 The dislocation- particle interaction has often been quoted as the origin of a threshold stress for dispersion-strengthened materials. This is related to Orowan bowing stress, and is given by following equation

(T,, = 0+34Gb/(A - d,) (2)

where u,, is the Orowan stress, G is the shear modulus, b is the Burgers vector, h is the planar spacing between particulates, and d, is the average particle size. When the Sic particle is considered as a barrier, the Orowan stress calculated is too small to be considered as the origin of threshold stress.3,6,8 Park et a1.6 have presented an alternative view for the origin of the threshold stress in a 6061Al/SiC, composite. They suggest that the threshold stress bears no relation to the SIC particles, but rather arises from the presence of ultra-fine oxide particles in the composite as a result of fabrication of the MMC by the powder metallurgy route. However, the view of Park et a1.6 was criticised by Pandey et al.,’ who reported that the threshold stress depends on the volume fraction of Sic.

Furthermore, Arzt and Wilkinson have proposed a model for attractive dislocation-particle interactionz3 which gives the detachment stress, gd, as

(TV = 0.29(Eb/h)(l - k2d)1’2 (3)

where kd is the relaxation parameter, which can take values between 0 and 1.

The second possibility is related to the presence of subgrains and associated internal stress. The internal stress, pi, associated with subgrains can be estimated from24

Ui = 3G8(1 - y)(b/eA)*” (4)

where 8 is the misorientation angle between adjacent subgrains, Y is Poisson’s ratio, and h is the subgrain size.

It should be noted that uor, (Td and Bi, estimated from eqns (2)-(4), respectively, are directly propor- tional to the modulus. This shows that the ratios u,,IE, udlE and ui/E are independent of tempera- ture. Apparently, they do not agree with the present results (Table 2) and those of Gonzalez-Doncel and Sherby.‘”

The third possibility could be load transfer to the reinforcement phase. However, all of the existing models predict that the magnitude of load transfer is dependent on applied stress. Thus, the observed stress exponent remains unaltered. Apparently, this also conflicts with the experimental data. Pandey et al.8 have suggested there is a need for a satisfactory model based on an applied-stress-independent load-transfer mechanism to explain the origin of the threshold stress for steady-state creep of composites.

From Fig. 6, it can be seen that the data points of

702 S. C. Tjong, Z. Y. Ma

sample 1 at two different temperatures can be well fitted by a single straight line with a slope of 8, whereas the data points of sample 2 at two different temperatures are merged together on a single straight line with a slope of approximately 8. Thus, the creep rate is controlled by lattice diffusion. It is noted that the creep resistances of samples 1 and 2 are approximately equal after normalisation. From Figs 6 and 7, it is indicated that the substructure-invariant model can provide a satisfactory explanation for the present creep data.

7.

8.

9.

10.

5 CONCLUSIONS 11.

1. The Sic-A1203--TiB2/A1 composites exhibit an apparent stress exponent of 8.9-9.5 and an apparent activation energy of 177 kJ mall’, whereas the Sic-A1203-TiB2/A1-Cu composite exhibits an apparent stress exponent of 11*9- 13.2 and an apparent activation energy of 323 kJ mol-‘.

12.

13.

2. The apparent stress exponents of the two composites increase with increasing temperature.

14.

15. 3. The threshold stress obtained by a linear

extrapolation method exhibits a strong tem- perature dependence, which cannot be ex- plained by the existing threshold stress models.

4. The creep behaviour of these composites can be explained by the substructure-invariant model. After normalisation, the two composites exhibit almost equal creep resistances.

16.

17.

REFERENCES 18.

19. 1.

2.

3.

4.

5.

6.

Webster, D., Effect of lithium on the mechanical properties and microstructure of Sic whisker reinforced aluminum alloys. Metall. Trans., 13A, 1982, 1511-1519. Nieh, T. G., Creep rupture of a silicon carbide reinforced aluminum composite. Metall. Trans., 15A, 1984,139-146. Nardone, V. C. and Strife, J. R., Analysis of the creep behavior of silicon carbide whisker reinforced 2124 Al (T4). Metall. Trans., 18A, 1987, 109-114. Nieh, T. G., Xia, K. and Longdon, T. G., Mechanical properties of discontinuous Sic reinforced aluminum composite at elevated temperatures. J. Eng. Mater. Technol., 110, 1988, 77-82. Morimoto, T., Yamaoka, T., Lilholt, H. and Taya, M., Second stage creep of Sic whisker/6061 aluminum composite at 573 K. J. Eng. Mater. Technol., 110, 1988, 70-76. Park, K. T., Lavernia, E. J. and Mohamed, F. A., High temperature creep of silicon carbide particulate

20.

21.

22.

23.

24.

reinforced aluminum. Acta Metall. Mater., 38, 1990, 2149-2159. Pandey, A. B., Mishra, R. S. and Mahajan, Y. R., Creep behaviour of an aluminum-silicon carbide particulate composite. Scripta Metall. Mater., 24, 1990, 1565-1570. Pandey, A. B., Mishra, R. S. and Mahajan, Y. R., Steady state creep behaviour of silicon carbide particulate reinforced aluminum composites. Acta Metall. Mater., 40, 1992, 2045-2052. Mohamed, F. A., Park, K. T. and Lavernia, E. J., Creep behavior of discontinuous Sic-Al composites. Mater. Sci. Eng., AlSO, 1992, 21-35. Pandey, A. B., Mishra, R. S. and Mahajan, Y. R., High-temperature creep of Al-TiB, particulates com- posites. Mater. Sci. Eng., A189, 1994, 95-104. Mishra, R. S. and Pandey, A. B., Some observation on the high-temperature creep behavior of 6061Al-SiC composites. Metall. Trans., 2lA, 1990, 2089-2090. Sherby, 0. D., Klundt, R. H. and Miller, A. K., Flow stress, subgrain size, and subgrain stability at elevated temperature. Metall. Trans., 8A, 1977, 843-850. Gonzalez-Doncel, G. and Sherby, 0. D., High temperature creep behavior of metal matrix aluminum- SIC composites. Acta Metall. Mater., 41, 1993, 2797-2805. Cadek, J., Sustek, V. and Pahutova, M., Is creep in discontinuous metal matrix composites lattice diffusion controlled?. Mater. Sci. Eng., A1’74, 1994, 141-147. Cadek, H., Oikawa, J. and Sustek, V., Threshold creep behaviour of discontinuous aluminum alloy matrix composites: An overview. Mater. Sci. Eng., A190, 1995, 9-21. Dragone, T. L. and Nix, W. D., Steady state and transient creep properties of an aluminum alloy reinforced with aluminum fibers. Acta Metall. Mater., 40,1992, 2781-2791. Ma, Z. Y. and Tjong, S. C., In-situ ceramic particulates reinforced aluminum matrix composites fabricated by reaction pressing in the TiO,-(Ti)-Al-B-(B,O,) systems. Metall. Mater. Trans., in press. Lundy, T. S. and Murdock, J. F., J. Appl. Phys., 33, 1962, 1671. Bird, J. E., Mukherjee, A. and Dorn, J. E., Quantitative Relations Between Properties and Microstructure. Israel University Press, Jerusalem, 1969. Krajewski, P. E., Allison, J. E. and Jones, J. W., The influence of matrix microstructure and particle rein- forcement on the creep behavior of 2219 aluminum. Metall. Trans., 24A, 1993, 2731-2741. Pandey, A. B., Mishra, R. S. and Mahajan, Y. R., Effect of a solid solution on the steady-state creep behavior of an aluminum matrix composite. Metall. Mater. Trans., 27A, 1996, 305-316. Rosler, J. and Arzt, E., A new model-based creep equation for dispersion strengthened materials. Acta Metall. Mater., 38, 1990, 671-683. Arzt, E. and Wilkinson, D. S., Threshold stresses for dislocation climb over hard particles: The effect of an attractive interaction. Acta Metall., 34, 1986, 1893-1898. Argon, A. S. and Takeuchi, S., Internal stresses in power-law creep. Acta Metall., 29, 1981, 1877-1884.