2519

LOW TEMPERATURE EMBRITTLEMENT OF ALLOYS Neil Paton, Rockwell International/Rocketdyne Division, Canoga Park, CA., USA

Abstract

Titanium alloys are well known to be subject to embrittlement by hydrogen, either as internal hydrogen or as hydrogen in the environment. Most of these embrittlement studies have been conducted at or near room temperature,, and low temperature be.:. havior has received little attention. Titanium structures fre­ quently operate somewhat below room temperature, however, and recent studies on embrittlement of titanium alloys at tempera­ tures from -ao0 to 20°c have shown that embrittlement can occur at levels of internal hydrogen well within the limits of stan­ dard specifications. It will be shown that this embrittlement can be attributed to diffusion of hydrogen to a region of high triaxiality, and subsequent hydride formation can lead to em­ brittlement, cracking, and finally . Both theoretical studies and experimental results will be described in the paper, and it will be shown that there is a good correlation between the predictions of .the theoretical model and the experimental data. Of particular importance is the experimental observation and theoretical prediction of a maximum in crack growth rate at about 100°C below room temperature. The importance of these results and their implications in structural design and poten­ tial limits to acceptable hydrogen concentration levels in titanium alloys will be described.

Introduction

Embrittlement of titanium in alloys by hydrogen has been the subject of extensive research over the last 20 years. Embrit­ tlement may either take the form of internal hydrogen embrittle­ ment due to hydrogen introduced into the metal during various processing steps, or it may take the form of environmental hydrogen embrittlement where hydrogen is introduced into the metal from the environment during service.

The level of understanding of internal hydrogen embrittlement of titanium alloys.is such that it is no longer a problem under normal -circumstances. Control of this form of embrittlement is achieved by keeping hydrogen levels low during melting and processing of titanium alloys. If particularly low levels of titanium are acquired, then vacuum annealing can be used to further reduce internal hydrogen levels.

Titanium alloys produced in the United States are generally fabricated to a specification requiring less than 125 parts per million (ppm) hydrogen in the alloy. More frequently, however, the actual levels,encountered in service are of the order of 50 to 60 ppm and can be controlled to as low as 10 ppm with special processing. At these levels, the effect of hydrogen on mechanical properties of titanium alloys such as tensile and 2520 properties is generally not measurable and of no real concern in service. There is some indication, however, that hydrogen levels above 50 ppm can have a negative influence on . (1) In addition, there is mounting evidence that under conditions of sustained load cracking or fatigue with a dwell time on the opening cycle, severely detrimental effects can be observed with hydrogen levels as low as 50 to 100 ppm. (2-5).

This paper discusses in detail circumstances under which embrit­ tlement can be encountered even at these low levels of hydrogen.

Sufficient experimental work has been conducted to indicate that embrittlement of titanium does not occur if the hydrogen level is maintained below the equilibrium solubility level. However, under certain circumstances, local concentrations might rise above this level. This can occur, for example, under condi­ tions of high triaxiality, such as at a crack tip, or at low temperature, where the driving force for hydride precipitation is above that at room temperature.

Theoretical Background

To establish the conditions under which embrittlement might be anticipated with nominally low hydrogen concentration levels, two factors must be understood on a theoretical basis: the first is the effect of stress on the equilibrium solubility limit, and the second is the effect of misfit on terminal solid solubility. Both of these factors could have a profound effect on the propensity for titanium alloys to precipitate hydrides locally, and it is this local precipitation which can, under the right circumstances, cause embrittlement.

Effect of Stress Ori Equilibrium Solute Coricentration. A hydrogen .atom in solution in titanium causes a volume increase that can be expressed as a partial molar volume of hydrogen in solution, VH. If a stress is applied to the material, the hydrostatic component of the stress, P, does an amount of work, PVH and the hydrogen solubility is represented by PVH CH = c~ exp RT 0 where CH is the solubility. . under zero stress. For t h e case o f a tensile stress, the concentration of hydrogen in equilibrium with the standard state will increase, and in the case of. a compressive stress, the hydrogen concentration will decrease. This means that at the tip of a crack, for example, where there is a hydrostatic tensile component, the equilibrium hydrogen concentration will increase while that remote from the crack will tend to decrease, providing a hydrogen flux to the crack tip. In other words, hydrogen will diffuse up the stress grad­ ient and up the concentration gradient toward the crack tip.

This situation is shown schematically in Fig. 1. The flux of hydrogen toward the crack tip has been calculated by Pardee and 2521

Paton in Ref. 5. The concentration of hydro­ gen at the crack tip is dependent on distance from the crack tip and on the time after ap­ plication of the load and temperature, and has been calculated in Ref. 5 for specific circumstances which show that concentra­ tions of the order of Fig. 1: Schematic of a crack growing 2 to 3 times the aver­ into a region of high triaxiality con­ age concentration in taining a region of precipitating the bulk material can hydrides. easily be obtained after 10 hours at 20°c. Thus, concentrations exceeding the terminal solubility limit are reached in a short period of time for titanium alloys containing as little as 50 ppm hydrogen where the solubility limit might be, for example, 150 ppm. The situation is further complicated, however, by the fact that the solubility limit itself is affected by stress, as discussed in the following paragrpahs.

The Effect of Misfit on Terminal Solubility. The precipi­ tation of hydrides in the metal containing hydrogen in solution is significantly influenced by the fact that the hydride has a large misfit with respect to the matrix. When titanium contain­ ing hydrogen in solid solution is converted to a hydride, there is a volume increase of about 18 to 23 percent associated-with this transformation. (5) The misfit must be accommodated either by elastic strains around the hydride or by plastic deformation of the matrix itself. In either case, mechanical work must be done. A tensile stress will therefore tend to assist hydride precipitation by reducing the amount of work done by the hydride, while a compressive stress will increase the amount of work re­ quired to be done by the precipitating hydride and therefore increase the apparent terminal solubility. The thermodynamics associated with this transformation were initially discussed by Paton et al. (6) and have since been further developed by numer­ ous workers. (7,8,9)

Energetically, the work required to expand the matrix to accom­ modate the additional volume of hydride must be provided by the chemical energy of the phase transformation from titanium to titanium hydride. The work.done on the matrix can either take the form of elastic energy if the hydride is coherent, which is unlikely except for very small precipitates at this large mis­ fit, or take the form of plastic work, W , if the accommodation is by plastic punching of dislocation logps in the matrix, which is the more usual situation.. The plastic work, W , is irrever­ sible and results in hysteresis in the transforma~ion, while the elastic work, We, is fully reversible and does not result in any hysteresis. The extent of the hysteresis depends on the 2522 magnitude of the term Wp. These issues have been discussed in detail elsewhere. (6-9)

The WP term is only large where the matrix is strengthenec by solid solution strengthening of additional elements such as aluminum in the matrix titanium alloy, or perhaps by severe cold work.

0 Thus for the standard stress-free state, where CH is the equili­ brium concentration and llGQ is the standard free energy change for hydrogen going from hydride into solution, then

0 CH = exp - llG /RT • 0 In a condition where self-stresses are permitted to build up around the particle, and C is the fraction of dissolved hydro­ gen, then 0 + w + w exp e p RT where We and WP both result in an increased apparent solubility.

Under conditions of zero-applied stress, self-stresses around a growing hydride can grow to considerable levels resulting in a barrier to further hydride growth and supersaturation of hydro­ gen in solid solution before any hydrides can nucleate and grow to a significant size. This is particularly true where the matrix is strengthened by solid solution elements such as alum­ inum, where the matrix can build up large back stresses around incipient hydrides without deformation and therefore prevent hydrides from growing. Introducing a tensile plastic strain through the application of an external stress can now relieve the self-stresses around an incipient hydride permitting it to grow, reducing the solubility toward the unconstrained value.

One further point that should be made concerning alloying ef­ fects is the influence of alloy composition on hydride habit plane. It has been pointed out recently(lO) that aluminum ad­ ditions cause the hydride habit plane to change from (lOTO) planes in pure titanium toward (10T7) and (0001) in titanium 6.6 percent aluminum. This is caused by the change in slip mode to basal slip at higher aluminum concentrations~ Perovic et a1.(10) have examined hydride precipitation in zirconium alloys and suggested that the observed (10T7) habit plane in the case of zirconium can be accounted for in terms of a growth mechan­ ism involving stacks of microscopic hydrides that lie on basal (0001) planes. Whatever the explanation in the case of titanium alloys, the observation of (10T7) and (0001) hydrides in alloys containing aluminum levels typical of commercial alloys coincides with the observed basal or near-basal cracking planes frequently re­ ported under conditions of internal hydrogen embrittlement in such alloys. ( 3 l 2523

Results and Discussion

Sustained Load Cracking. Levels of hydrogen significantly above the nominal 100 to 125 ppm permitted in most specifica­ tions do not cause any noticeable change in monotonic or cyclic mechanical properties of titanium alloys. Even at concentra­ tions up to 300 ppm, tensile properties of most commercial alloys are essentially unaffected. However, recently, more subtle effects on mechanical properties such as sustained load cracking, , and hold-time effects in fatigue have been observed. (11)

These effects are all time and temperature dependent in nature, since diffusion of hydrogen toward some incipient failure point is a key element. Boyer and Spurr(l2) were the first to show experimentally that crack growth rate in a sustained load crack­ ing test reaches a maximum of about -5o0 c. The peak was very abrupt with maxi­ mum crack growth rates of 4 x 10-6 Ti-6Al-6V-2Sn /35;ppm H2 mm/sec, declin­ I (RE~ 1~ ing to immeasur­ Tl-SAl-2.SSn ably low rates L STRESS INTENSIT1Y 75°c above and I H2 ENVIRONMENT So-60 MPa-m'h below this peak 10-5 I (REF. 13) temperature. ::c The results of ~ 10-6 a: Boyer and Spurr Cl are reproduced in Fig. 2 along Ti-6AI with recent 100 ppm H2 theoretical (REF. 5, THEORY) calculations by Paton and -200 -100 0 +100 +200 Pardee. (5) It TEMPERATURE, C will be noted Fig. 2: Crack growth rate for various that the form titanium alloys in sustained load of the latter cracking. theoretical curve is the same as that of Boyer and Spurr, although the peak tem­ perature is somewhat higher. The occurrence of the peak is readily explained in terms of a C or TTT curve frequently em­ ployed to characterize precipitation phenomena. The precipita­ tion rate N is low at high temperature because of a low driving force, high at intermediate temperatures and low again at low temperatures because of slow kinetics. In the case of hydride precipitation at a crack tip, with low bulk hydrogen concentra­ tions of the order of 100 ppm, the peak precipitation rates can be below room temperature because slightly above room tempera­ ture all the hydrogen is in solid solution.

More subtle effects can be envisaged if temperature is cycled while maintaining a constant applied load. At high tempera­ tures above the solubility limit, rapid diffusion can occur, resulting in a hydrogen "cloud" at the crack tip. This will 2524 not result in sustained load cracking, however, as long as the maximum concentration within the cloud is everywhere below the solubility limit appropriate for the local stress state. If the temperature is lowered, however, precipitation is highly probable, since the solubility limit is now lowered. Tempera­ ture cycling such as might occur from night to day, or low to high altitude, would be more than enough to promote such a mechanism.

An example of what is believed to be this type of cracking has been observed recently in commercial purity titanium hydraulic fittings. The range of hydrogen concentrations was very low, being on the order of 30 to 70 ppm, but small cracks attributed to sustained load hydride cracking were observed in several in­ stances after some 2 to 4 years. The fractography found to be evident in this type of fracture is shown in Fig. 3 and was re­ markably similar to that found in other cases of sustained load cracking. It should be noted that cracking of titanium in an external hydrogen environment also results in similar fractoq­ raphy as has been demonstrated by Williams and Nelson. (13) Furthermore, crack growth rates reported by Williams and Nelson

Fig. 3: Fracture surface of a commercially pure titanium sample containing -so ppm hydrogen subjected to sustained load cracking at room temperature. for Ti-5Al-2.5Sn are very similar in the low temperature regime to those for Ti-6Al-4V as plotted in Fig. 2. This would appear to be good evidence that the kinetic processes that control crack growth are the same in each case.

Although sustained load cracking rates in alpha titanium alloys such as commercial purity and Ti-6Al are generally quite slow, (5) alpha beta alloys having a continuous beta phase can be extremely rapid. The maximum sustained load cracking rates observed and calculated by Pardee and Paton(S) had a maximum value of the order of lo-5 mm/sec. These results are substan­ tiated by the data of Boyer and Spurr(12 l for an alpha alloy with discontinuous beta phase and are also in general agreement with crack growth rates quoted by those investigating sustained load cracking in alpha zirconium.(7) 2525

Recently, however, much higher crack growth rates have been mea­ sured in a Ti-6Al-6V-2Sn alloy with a microstructure exhibiting continuous beta phase by Moody and Gerberich. (14) Crack growth rates were as much as three orders of magnitude higher than those generally encountered in alloys containing either no beta phase or a continuous beta matrix. Moody and Gerberich(l4) measured growth rates up to io-2 mm/sec, and rates approaching these values have also been measured in the same alloy by Lederich et al. (11) All of these ·results were obtained with relatively low hydrogen concentrations of the order of 35 to 38 ppm and temperatures in the vicinity of room temperature. Moody and Gerberich also showed that thick specimens with high triaxiality at the crack tip favored high crack growth rates, providing as much as two orders of mangitude further increase in growth rate when specimen thickness increased from 2.5 to 20 mm. The detrimental effect of a continuous beta phase in this type of loading is further substantiated by results in microstructure effect on Ti-6Al-4V, discussed in the following section. · Hold-Time Effects on Fatigue Crack Propagation. The influ­ ence of a hold or dwell time applied during the opening cycle at peak load in a fatigue test has been shown, under some cir­ cumstances to cause considerable acceleration of the crack prop­ agation rate in titanium. (2,3,15) This effect has generally been ascribed to stress-induced hydride precipitation at the crack tip in a manner similar to that recognized as the cause of sustained load cracking in these alloys.

Chesnutt and Paton(2) showed that in the recrystallization an­ nealed (RA) condition, where both a and 8 phases have a contin­ uous path, acceleration of up to 30 times the normal crack growth rate was observed, while in the beta quenched (BQ) con­ dition no acceleration was observed. In the latter cas~ neither phase is continuous, so that a short circuit diffusion path through the beta phase is not available. These results are sum­ marized in Table 1. Table 1 shows both the observed accelera­ tion in da/dN in comparison to the anticipated growth rate with­ out a hold time.

TEMPERATURE, Kmax da/dN MICROSTRUCTURE Oc MPa•mV2 INCREASE

RA -70 21.4 30X

RA 20 30 <2X

ea -70 34.6 NONE

ea 20 27.7 NONE

Table 1: Increase in fatigue crack growth da/dN for a 5- minute hold time in Ti-6Al-4V with 100 ppm hydrogen. It can be seen from Table:! that both low temperature, tci pro­ vide a driving force for hydride precipitation, and an RA microstructure are required to cause acceleration. At higher hydrogen concentrations (300 ppm) some acceleration was also 2526

noted by Chesnutt and Paton(2) at 20°c, but these high concen­ trations are well beyond those normally encountered. Lowering the hydrogen content to the order of 10 ppm eliminated any ac­ celeration at all temperatures and stress intensities tested.

Conclusions

1. Rapid crack growth by a process of stress-induced hydride precipitation can occur in titanium alloys at hydrogen con­ centrations as low as 35 ppm. 2. This fracture process is aided by high triaxiality as might be found in thick sections, low temperatures (v - 7ooc) and continuous beta phase in the microstructure to provide a diffusion path for hydrogen. 3. Hold-time effects in fatigue of titanium alloys can also be explained in terms of sustained load cracking due to hydride precipitation. 4. Removal of hydrogen down to levels of the roder of 10 ppm appears to eliminate this phenomenon in most titanium alloys.

References

(1) D. A. Meyn: Met. Trans. 5 (1974) 2405. (2) J. C. Chesnutt and N. E. Paton: "Titanium '80" AIME (1980) 1855. (3) A. W. Sommer and D. Eylon: Met. Trans. 14A (1983) 2178. (4) J. C. M. Li, R. A. Oriani, and L. S. Darken: 2. Phys. Chem. 4 9 ( 19 6 6) 18. (5) w. J. Pardee and N. E. Paton: Met. Trans. llA (1980) 1391. (6) N. E. Paton, B. S. Hickman, and D. H. Leslie: Met. Trans. 2 (1971) 2791. (7) M. P. Puls, L. A. Simpson, and R. Dutton: Fracture Prob­ lems and Solutions in the Energy Industry, Pergamon Press, N.Y. (1982) 13. (8) M. L. Grossbeck and H. K. Birnbaum: Acta. Met. 25 (1977) 135. (9) N. E. Paton and R. A. Spurling: Met. Trans. 7A (1976) 1769. (10) V. Perovic, G. C. Weatherly, and C. J. Simpson: Acta. Met. 31 (1983) 1381. (11) R. J. Lederich, S. M. L. Sastry, and P. S. Pao: Met. Trans. 13A (1982) 497. (12) R. R. Boyer and W. F. Spurr: Met. Trans. 9A (1978) 23. (13) D. P. Williams and H. G. Nelson: Met. Trans. 3 (1972) 2107. (14) N. R. Moody and W. W. Gerberich: Met. Trans. 13A (1982) 1055. (15) J. E. Hack and G. R. Leverant: Met. Trans. 13A '(1982) 1729.