At odds with ODS...

TWI Bulletin, July - August 2003

Philip Threadgill joined TWI in 1976 to work on ferrous metallurgy under the guidance of Norman Bailey. Following a four year secondment to EWI to develop and lead their Materials Department in the late 1980s he returned to TWI's Materials Department in 1989 to specialise in joining advanced materials, including intermetallic alloys, metal matrix composites and ODS alloys. In 1995 he transferred to the Friction and Forge Processes Group, and has been the R&D Manager for that section since 1999.

Joining iron based Oxide Dispersion Strengthened alloys has been largely uncharted territory until recently. TWI has been exploring the options using fusion welding, diffusion bonding and friction welding.

Iron based ODS alloys have been the subject of considerable interest for high temperature usage, where their high strengths, good creep resistance and resistance to many forms of corrosion offer considerable advantages over other materials. As Philip Threadgill explains there are two basic families of alloys, namely the well established alloys based on Fe-20Cr-5Al-Y ₂O ₃, ie Inco MA956 and Plansee PM2000, and the more recently developed, and still largely experimental alloys based on iron aluminides.

Two such intermetallic alloys have received considerable attention, namely the FeAl40 Grade 3, made by CEREM in France, and the Fe3Al-ODS alloy developed by ORNL in the USA. The former has a composition on approximately Fe ₆₀Al ₄₀, and is thus closer to FeAl than Fe ₃Al. Intermetallic alloys have been the subject of many investigations to exploit their excellent high temperature mechanical properties and oxidation resistance. Recent studies at CEREM and ORNL have developed improved alloys based on the Fe ₆₀Al ₄₀ and Fe ₃Al systems, ^[1,2] whereby these alloys are used as the basis for ODS alloys.

ODS alloys are made by a powder attrition and compaction method, and contain a very fine dispersion of oxide particles (usually Y ₂O ₃) which improve the creep resistance by pinning dislocations, and a stable oxide layer of Al ₂O ₃, which provides excellent oxidation and corrosion resistance. Such materials are of great interest for new generation heat exchangers and other high temperature equipment. While development of such alloys continues, it is important to remember that techniques have also to be developed to enable them to be processed and fabricated into useful components. It should be self evident that any industrial uptake of these alloys would be greatly assisted by the availability of simple fabrication and processing methods, and welding is obviously one of the most important aspects. Earlier work on welding ODS alloys by several authors, reviewed by McKimpson and O'Donnell ^[3] has established that the use of fusion welding techniques will seriously compromise the high temperature mechanical properties of the alloys, as the fine dispersion of oxides in the alloy will be destroyed. However, many ODS alloys can be welded without problems of cracking or porosity, even though the high temperature properties are poor. Non-fusion techniques are therefore intrinsically more attractive, as these will not remove the critical oxides from the joint. There are also several references in the literature to brazing, which can give useful results. However, this work will not be discussed further in this paper.

ODS alloys such as MA956 and PM2000 have been successfully joined by continuous drive rotary friction welding and diffusion bonding ^[4,5,6,7] . Recent work has also reported the successful joining of an ODS alloys based on Fe ₆₀Al ₄₀ ^[8] and Fe ₃Al ^[9] . Studies on welding ODS alloys to date have concentrated on room temperature mechanical properties and on understanding the microstructures formed. There appears to be little data in the public domain on the high temperature performance of these alloys.

The welding of iron aluminides has been investigated for a number of alloys, mostly using fusion processes. ^[3,10,11,12] No major problems with welding this class of alloy have been reported using TIG and electron beam welding. However, these investigations were primarily to demonstrate the feasibility rather than achieve fully optimised welding parameters. No special filler metals are known to have been developed, and all fusion welds so far reported in iron aluminides have been autogenous.

The purpose of this work is to review the available welding methods for use with iron based ODS alloys, and this includes work on the well established alloys such as MA956 and PM2000, as well as more recent work on iron aluminide based ODS alloys.

Fusion processes

TIG Welding

Ideally, the final structure should be coarse grained to improve creep strength. Post weld heat treatment of fusion welded joints made in the fine grained condition gives unpredictable, and generally unsatisfactory results. The grain size in the welded region generally remains finer than the parent material grain size, and this is believed to be a consequence of the reduced stored energy in the weld metal, which will reduce the driving force for grain growth. The orientation of the coarse grains in the fusion zone of the weld is typically determined by the solidification direction of the weld, unlike the parent material, where it is determined by the working direction of the material. Thus, in weld metal, the orientation of the coarse grained welds changes across the weld, resulting in many grains being banana shaped. This is clearly a less than ideal solution.

A feature which can be overlooked is that some ODS alloys are made in a hydrogen atmosphere, which may increase the risk of hydrogen cracking during welding, particularly during flux shielded processes. Even if argon is used, the gas may become trapped, contributing to porosity on melting.

Laser welding

Work by Bucklow et al ^[4] on gas free PM2000 showed that although seemingly good quality welds could be obtained by CO ₂ laser welding, better results were obtained only at higher welding speeds. Lower speeds promoted hot cracking, although the reasons for this are not clear. Examination of the microstructure showed epitaxial growth from the coarse grained parent material into the fusion zone, but it is also clear from Fig.2 that extensive nucleation of equiaxed grains occurred in the solidifying weld pool. Some directionality was evident close to the upper and lower surfaces. As far as is known, there is no high temperature property data in the public domain, but there is no expectation of good results as the Y ₂O ₃ dispersion will be lost. Thus, although laser welding is perfectly feasible for certain geometries, the performance of the welds is not expected to be high, and the component design would need to reflect this.

Solid state processes

Diffusion Bonding

Some of the more recent work in diffusion bonding of ODS alloys has been reviewed by Cam and Koçak. ^[13] These authors stress the difficulty of achieving grain growth across the interface with either conventional or transient liquid phase methods. The work by Bucklow et al ^[4,7] provided an elegant solution to the development of a controlled coarse grained structure across the interface in MA956 and PM2000 alloys. In this work, diffusion bonds were made between fine grained materials in either MA956 or PM2000. Special care was needed to make certain that no cold worked material remained at the interface after grinding flat, and this was achieved by carefully controlled electro-polishing. After bonding, the joint was placed in a furnace with a very small hot zone, and the joint was passed through the hot zone slowly. As the fine grained area became heated to a temperature at which recrystallisation occurred, epitaxial growth occurred on the coarse grains as the zoning proceeded, leading to a complete coarse grained structure, in which the bond line was often difficult to detect. Some preference for growth on particular orientations was observed, but no quantitative measurements were made. The authors noted significant batch to batch variations in the response of these alloys to diffusion bonding. Although the approach provides the best end result of any joining process in terms of microstructure, it is a very slow process, and batch to batch repeatability would complicate production usage of this process. Further work would be required to develop a procedure which could be used in commercial practice. Typical examples of the microstructures obtained in diffusion bonds after the zoning heat treatment are given in Fig.3. The authors also found similar results when joining coarse to fine grained material, followed by a zoning heat treatment. It is interesting to note that a similar approach was unsuccessful on nickel based ODS alloys. Bucklow has suggested that success in the iron based alloys is related to the self diffusion rates at temperatures below the recrystallisation temperature, whereas the diffusion rates in nickel based alloys are much lower.

Table 1. A summary of published data on friction welding of iron based ODS alloys..

Friction welding

Friction welding is a totally solid state process, and therefore problems associated with solidification and fusion, such as segregation, evaporation of certain elements, floating out of particles, porosity etc are avoided. There have been several publications dealing with friction welding of ODS alloys, and these have investigated both iron based and iron aluminide based alloys.

Figure 4 shows typical macrosections from two rotary friction welds in iron aluminide based ODS alloys, and the similarity is obvious. Comparison with the work of Kang et al ^[5,6] shows an almost identical appearance with MA956 material welded by the same process.

The microstructural features for friction welds in fine grained MA956, Fe ₆₀Al ₄₀-ODS and Fe ₃Al-ODS have been characterised, ^[5,6,8,9] and found, not unexpectedly, to be very similar. At the bond line, the structure consists of fine, equiaxed grains, generally with a low dislocation density. Although no measurements are known, it is suspected that the grain boundaries are predominantly high angle, and it is generally agreed that these grains are formed by dynamic recrystallisation during the friction welding process. The grain size is typically a little larger at the bond line than at the outer edge of the recrystallised region, typically by a factor of no more than two. Between the recrystallised area and the parent material there is no widespread recrystallisation, although the high aspect ratio grains in the parent material are progressively bent through almost 90 degrees as the bond line is approached. Close examination shows a greater frequency of small equiaxed subgrains as the bending angle increases, and it is assumed that these are due to local recovery mechanisms. Neither the strain, nor the strain rate in these regions is as high as in the recrystallised area at the weld centre. A typical weld region is shown in Fig.5. Close up views of the recrystallised areas are shown in Fig.6.  

An important observation on the redistribution of the Y ₂O ₃ particles has been made by Kang et al ^[5,6,8,9] in MA956, and subsequently confirmed by Inkson and Threadgill in Fe ₆₀Al ₄₀-ODS ^[8] . These studies have shown that in the fine grained equiaxed area close to the bond line, agglomeration of Y ₂O ₃ particle has occurred. This has led to the formation of fewer, but larger particles, with a consequent increase in the interparticle spacing which will have a profound effect on the pinning of dislocations, and hence strength at high and low temperatures. This is illustrated in Fig.7. Kang has observed some association of the agglomerated Y ₂O ₃ particles with either Al ₂O ₃ or Ti(C,N) particles, but this effect was not observed in the Fe ₆₀Al ₄₀-ODS alloy.

The mechanical properties of friction welds in ODS alloys have been investigated at ambient and at elevated temperatures, although the database is small, particularly so for the intermetallic based alloys. It has been shown that tensile strengths at ambient temperatures which approach that of the parent material can be obtained in MA956 alloys, although the strength achieved is influenced, as expected, by the welding parameters used. For MA956, higher forge pressures and lower weld times appeared beneficial. ^[5,6] The database for Fe ₆₀Al ₄₀-ODS and Fe ₃Al-ODS is more limited, and therefore it is not possible to draw the same conclusion, although the data available ^[8] are not inconsistent with the observations of Shinozaki et al. ^[6] At high temperatures, Shinozaki et al. ^[6] have shown a substantial reduction in the creep strength of the alloys.

Better results were obtained with high forge force conditions, and failure occurred away from the bond line in the deformed region of the HAZ. Tests were only carried out on fine grained material, and these authors have suggested that in the region of failure, many grains will be oriented at about 45 degrees to the tensile axis. It is clear that this is the point at which the resolved shear stresses will be maximized, allowing a greater probability of grain boundary sliding.

Sketchley et al. ^[9] have shown the possibility of joining ODS to alloys to other materials. In this work, an Fe ₃Al-ODS intermetallic alloy was joined successfully to a nickel based alloy (Haynes 230), and it seemed to make little difference whether the ODS alloy was in the fine grained or coarse grained condition. Examples of dissimilar material rotary friction welds are shown in Fig.8.

The effect of a post-weld recrystallisation treatment is shown in Fig.9 for a friction welded MA754 alloy. Although this alloy is nickel based, the figure serves to illustrate the very similar microstructural behaviour of the two alloys. Recrystallisation for one hour at 1315°C has resulted in substantial grain growth, but the orientation of the large grains at the bond line is parallel to the bond line, instead of the ideal perpendicular. Similar effects have been recorded in iron based ODS alloy

Discussion

Weld quality:

It has been observed that several processes can be successfully used to weld iron based ODS alloys. However, fusion based processes run a high risk of porosity, due to release of shielding gases which have become occluded into the material during the attrition process. It is also believed that the fine Y ₂O ₃ and other small particles coalesce and float out during welding. Processes such as TIG welding are probably more suited to local repairs, or to structural welds in locations where high service temperatures and pressures are not encountered. Solid state processes give defect free welds without difficulty, and friction welding in particular has the advantage of enabling joints between dissimilar materials. It is also noteworthy that intermetallic alloys in general respond well to rotary and linear friction welding, and there are several instances of using the technique to join alloys based on Ti ₃Al ^[14,15,16] , TiAl ^{[17,18,19,20]} and Ni ₃Al ^[21,22] .

Weld microstructures

Fusion weld microstructures show a large dependence on the original microstructure. Grain growth is largely epitaxial, although some nucleation of new grains occurs in the solidifying pool. The direction of grain growth is determined principally by the direction of solidification. Any directionality in the original material will be lost, and impossible to recover. The grains in the weld are in a predominantly annealed condition, and much of the Y ₂O ₃ will have been lost.

In friction welds, there is a gradual transition through deformed, and partly recovered grain structures to fully recrystallised grains at the bond line. Although the lack of melting prevents loss of Y ₂O ₃, there is strong evidence that the particles will agglomerate, which is an irreversible process. Furthermore, recrystallisation of friction welds gives large grains which have unfavourable orientations at the bond line, and low dislocation densities. For these reasons, it is unlikely that parent material properties can be achieved at the bond line, in particular at high temperatures.

Mechanical properties

Although this article has not reviewed mechanical properties of ODS welds in detail, examination of the literature establishes two main facts. First, ambient temperature mechanical properties never equal those of the parent material (although strengths of 80-90% are easily obtained), and secondly, the volume of data at elevated temperatures, both tensile and creep, is minimal. As for ambient temperatures, parent material properties cannot be matched at elevated temperatures.

The reasons have been discussed, and are related to irreversible changes to the microstructure, namely loss or agglomeration of second phase particles, reduction in dislocation density, and unfavourable grain size and orientation. It is likely that the only process where microstructural changes can be avoided is diffusion bonding, but this is slow and expensive. Explosive bonding may also preserve the microstructure better than other processes, as the plastically deformed region will be minimised. However, little information exists to substantiate this point of view.

Although friction welding has undoubted attractions, it is believed that attainment of full parent material strength in friction welds in ODS alloys will not be achieved, as the microstructure at the bond line will be irreversibly damaged. The same phenomena will also adversely affect the high temperature tensile and creep behavior of the welds for the same reasons, and it is considered unrealistic to expect parent material properties in friction welds. However, sound welds can be made with relative ease, and the prospects for solid state processes look far better than for fusion processes, where the Y ₂O ₃ particles are believed to agglomerate and float out of the weld pool.

Although friction welding has great potential for joining ODS alloys, and is arguably the most suitable process, designs employing the process will have to allow for the reduction in mechanical properties.

Conclusions

This assessment of previous studies on welding of iron based ODS alloys has suggested that the area in general has been inadequately studied. The following conclusions can be drawn.

• Fusion welding techniques are feasible, but the fine dispersion of Y ₂O ₃ is irreversibly destroyed. Gas occluded in the attrition stage can also be released, giving a high risk of porosity.
  
  
• Diffusion bonding is slow, and the batch to batch variation in process parameters is a cause for concern. Nevertheless, welding fine grained material to coarse grained material, followed by a zoning heat treatment, has led to anall coarse grained structure where the bond line is difficult to detect.
  
  
• Friction welding is probably the most successful method available at present. The elongated grain structure is irreversibly destroyed at the interface, and there is strong evidence of particle agglomeration at the bond line.Despite this, the welds are sound, and can be made with relative ease.
  
  
• Friction welding can also be used successfully to make dissimilar material welds.
  
  
• In all cases (with the possible exception of diffusion bonds), microstructural changes will cause irreversible microstructural changes at the bond line which will compromise ambient and high temperature mechanical properties.Designers need to be aware of this, and design components accordingly.

Acknowledgements

The significant contributions of several colleagues to this paper is acknowledged, namely Dr I A Bucklow (TWI/Cambridge University), S B Dunkerton (TWI), Dr B J Inkson (Oxford University), P D Sketchley (TWI) and Dr I G Wright (ORNL). The work was supported by:

(1) Industrial Members of TWI, as part of the TWI Core Research Programme,
(2) EC funded Concerted European Action on Structural Intermetallics (CEASI) project.
(3) US DoE contract DE-AC05-96OR22464.

References