Low temperature diffusion bonding of steels

TWI Bulletin, November/December 1988

3 Diffusion bonding below the A ₁ temperature

Ian Bucklow, MA(Cantab), PhD is Head of Surface Technology in the Materials Department at The Welding Institute.

In parts 1 and 2 of this series of articles, it was maintained that all welding processes essentially require the conjunction of atomically clean surfaces. Such surfaces have an inherently high energy and so it was argued that it should be possible to use that energy to assist in the diffusion bonding of steels without massive deformation and at temperatures below the onset of metallurgical damage. An investigation into surface cleaning was described, and this further article reports on a series of diffusion bonding experiments designed to test the ideas presented.

Having established a technique for the production of a sufficiently clean surface, ^[1] some factors governing the bonding of low alloy ferritic steels below the A ₁ temperature were investigated under standardised bonding conditions (700°C, 20 N/mm ², 30min). Bonding trials were followed by tensile tests on specimens machined from the joints.* Among the factors influencing bonding, the following were investigated: surface cleanliness, surface finish, and steel composition.

Two steels were used in the bonding study: 080M40 and 709M40. ^[1] The former was chosen as being representative of a general engineering steel, and was used as a standard material for much of the bonding work; the latter is a widely used heat treatable 1%Cr-Mo 0.4%C steel and was introduced to study the possibility of bonding a low alloy steel below the subcritical annealing temperature. Finally, a study was made of the effect of contaminant gases on the bonding of previously cleaned steel surfaces.

Experimental details

UHV bonding chamber

The principal equipment ( Fig.1) consisted of an ultra high vacuum (UHV) chamber that was used for both cleaning and joining; it was needed because pressures around 10 ^-9torr allow an adequate 'clean time' for experimentation before recontamination by adsorbed layers. The specimens to be joined were in the form of rods, 8.5mm diameter by 40mm long, whose endfaces were machined accurately normal to their lengths. They were held horizontally, in line, with their endfaces initially separated by 12mm and were surrounded by a molybdenum mesh resistance heater. Accurate and extensive calibration established that the reduction in temperature in the parent metal on either side of the bond line was not more than 5°C at 700°C, and that the furnace temperature was controlled to ± 3°C at a mid specimen temperature of 700°C. Two 10kV, 100µA ion beam guns were mounted such that each bombarded one specimen face during the cleaning cycle before the specimens were brought together under pressure for bonding.

Experimental procedure

Specimens of 080M40 steel were machined to size and vacuum stress-relieved (S/R) at 600°C for 1 hr to avoid doming of the mating surfaces during bonding, and the 709M40 specimens were heat treated according to Table 1 before machining. Final preparation consisted of surface grinding to RA 0.35µm, or medium grinding on 7µm SiC paper to RA 0.035µm, or fine grinding on 3µm SiC paper to RA 0.025µm, or polishing on 0.25µm diamond to RA 0.016µum. Bonding was generally carried out at 700°C under a pressure of 20 N/mm 2 for 30min afterion bombardment cleaning at 700°C in the bonding chamber, and the standard procedure is given in Table 2. Parent metal rod was subjected to the same heat treatment and bonding cycle (B/C) as the bonded pairs.

Table 1 Heat treatments of 709M40 steel

Table 2 Standard bonding cycle

The bonding conditions of 700°C at 20 N/mm 2 for 30min were chosen somewhat arbitrarily and were adhered to virtually throughout the work to reduce the number of variables explored. As high a temperature as possible was required to achieve fast diffusion rates, but to remain below the Ac 1 for the material used; hence 700°C was adopted. The bonding pressure of 20 N/mm 2 was chosen as being safely below the level at which bulk deformation would occur. The time of 30min represents the shortest that could be expected (on the basis of diffusion data) to produce any worthwhile result at 700°C.

The achievement of satisfactory tensile properties in joints made under the above conditions encouraged a further reduction in bonding temperature, and so a series of runs at 600°C under a higher pressure of 100 N/mm 2 for 1hr was instituted with steel 709M40.

Four pairs of rods were bonded per condition; one was sectioned for metallographic examination, and the remaining three were machined for tensile testing. If there was a wide scatter of results from the tensile test, further bonds were made and tested until a satisfactory agreement was reached between the majority of tests.

The same material batch was used throughout each series of runs.

Specimens for tensile testing were machined to a gauge diameter of 4.5mm by 16mm gauge length (Hounsfield No. 12) and were pulled in an Instron at a crosshead speed of 2 mm/min at room temperature.

Results

080M40 steel

The tensile test results for the parent metal and bonded pairs are given in Table 3. Bonds made under the standard conditions between polished surfaces failed on tensile testing at the bond line at the onset of necking, whereas bonds between medium ground surfaces gave results similar to those from parent metal specimens with ductile failure occurring in the parent metal, away from the bond line. Specimens with medium ground surfaces bonded for 20hr were more ductile, and they too failed away from the bond line.

Table 3 080M40 steel: tensile test results of parent rod and bonded pairs

Metallographic sections of a standard condition bond (medium abraded finish) ( Fig.2) showed that there was no recrystallisation across the bond line and that pearlite colonies persisted up to the line. Both these features indicate that the bonding temperature remained below the Ac 1.

709M40 steel

The tensile test results for parent metal and bonded pairs are given in Tables 4 and 5. The response of parent material to bond cycling was consistent with a heat treatment that was in effect, at the top end of the tempering range, nevertheless the values obtained were within the acceptable range as defined by BS970, 1972.

Table 4 709M40 steel: tensile properties of parent rod and bonded pairs, bonded at 700°C

Bonds made at 700°C in as-received material between medium abraded surfaces ( i.e. S/R + B/C only) gave fully ductile tensile specimens which were virtually equivalent to parent rod that had undergone the same heat treatment ( Table 4).

Bonds between annealed and re-heat treated ( i.e. quenched and tempered) material were however, sensitive to surface roughness. All surfaces resulted in joints with much the same yield and tensile levels as parent rod of the same heat treatment, but the ductility figures discriminated sharply between them in that only bonds between polished surfaces produced ductility comparable to that of the parent metal. Rods thus bonded were fully ductile, and it may be significant that failures occurred outside the bonded zone to give the same reduction of area as in parent metal, but a lower elongation overall i.e. the presence of the bonded zone appeared to localise necking deformation. Surfaces prepared by fine abrasion or surface grinding failed across the bond line virtually at the onset of necking.

Bond cyling at 600°C of re-heat treated parent rod had no influence on the tensile properties, but joints made between medium ground surfaces at 600°C under 100 N/mm 2 for 1hr ( Table 5) failed across the bond line at the onset of necking.

Table 5 709M40 steel: tensile properties of parent rod and bonded pairs, bonded at 600°C

See Table 1 for heat treatments
Bonding conditions: Medium ground, 600°C, 100 N/mm 2, 1 hr

Fractography of bonded surfaces

Those specimens that failed at the bond line, irrespective of surface finish or steel composition, all showed a 'thumb nail' region of ductile dimpling occupying some 25% of the bond area ( Fig.3a); few dimples were seen to have inclusions within ( Fig.3b). The remaining fracture surface was characteristic of a fast brittle cleavage which, other work had shown, [2] ran through grains on either side of the interface ( Fig.3c).

This behaviour was remarkably consistent in all failures, even to the extent to which the fine abrasion scratches were emphasised in the dimpled region, and the same effect is seen in bond line failures of joints made at temperatures as high as 1000°C. ^[3]

Bonding of contaminated 080M40

The influence of chemical contamination on bondability was investigated by exposing abraded (7µm SiC) surfaces of 080M40 to carbon monoxide or ethylene immediately before joining. These two gases were chosen to simulate the contaminants present in a vacuum system and their adsorption to cleaned steel surfaces was reported in parts 1 and 2 of this series of articles. ^[1]

Experimental procedure

The standard procedure was followed up to the point where two specimens of 080M40 were about to be joined for bonding at 700°C. Ion beam cleaning was then halted and the contaminant gas was admitted to maintain a dynamic total pressure of 10 ^-6torr for 10min. The specimens were then brought into contact and pressurised to 20 N/mm ² for 30min as usual.

Tensile test results

Although one bond made in carbon monoxide appeared to be unaffected by the gas, five others failed at the bond line with virtually no ductility, four failing to reach the yield stress. Joints made in ethylene were little better, but they did yield away from the bond line and showed a small degree of ductility before fracturing across the interface.

Discussion of bonding results

Contaminated surfaces

Surface analysis indicated ^[1] that the commonly occurring vacuum chamber gases adsorbed rapidly on to cleaned steel surfaces. The parallel series of bonding experiments confirmed that contamination of surfaces immediately before bonding was detrimental. Carbon monoxide was the most harmful in this respect and it should be noted that the surface analysis studies showed it to be the most strongly and actively adsorbed of the gases.

Cleaned surfaces

Bonding at 700°C was sensitive to surface finish in that either polished or medium ground surfaces were required, but a wholly consistent pattern was not observed. Thus, medium ground surfaces bonded well on 080M40 and on as-received 709M40, but re-heat treated 709M40 required polished surfaces to be used.

Fractography gave no indication of the origin of these differences, but it did demonstrate that the microvoid coalescence dimples in bond line failures were somewhat uncharacteristic of those seen in association with ductile tearing (in that they were rather larger than normal), and there appeared to be an unusual lack of inclusions within them.

Both these observations suggested that the dimples were nucleated by existing voids at the interface, rather than by inclusions. Some confirmation of that supposition may be found in the exaggerated appearance of grinding scratches in the fractured surface ( Fig.3a).

The ductile area of bond line failure surfaces was always surrounded by a region of fast brittle cleavage which ran through grains on either side of the bond line, thereby implying that the bond line was stronger than the parent metal under the existing conditions. However, there were combinations of material and surface finish which gave bond lines that survived the maximum tensile load applied to the specimen. In these cases, ductile failure always occurred in the parent metal well away from the bond zone.

Gross plastic deformation across the bond zone itself was never observed, although microductility within the bond was always evident. Although it appears to be relatively easy to reach the onset of necking in a joint made by bonding cleaned surfaces, in all cases the bond line retained distinctive properties which were unlike those of the parent material.

The above observations can be accounted for by postulating the presence of interface voids which generate a largely elastic constraint condition in the contact regions surrounding them. It follows that the number of voids, and probably their shape, controls the overall joint properties but no evidence has been found to date to account for the influence of surface finish on this particular aspect.

The influence of the various heat treatments (and possibly of batch impurities) on bond line toughness is also unknown, but nitrogen, and to a lesser extent, oxygen is known to diffuse to a surface at elevated temperatures. ^[4] Furthermore, there are strong indications ^[5,6] that minor impurities in the bulk (rather than contaminants on the surface) diffuse to the interface voids during bonding and prevent their closure. If such an effect is operative, it presents a fundamental difficulty to the low temperature diffusion bonding of simple steels.

The relationship between surface cleaning and bonding

The results indicate that effective surface cleanliness is obtained by mild ion bombardment immediately before joining because many of even the poorest bonds survived the stresses of machining. It follows that there must have been extensive interface diffusion which can only have occurred (given the time/temperature combination and absence of recrystallisation across the bond line) at relatively clean surfaces. The work also demonstrates that the adsorption of contaminant gases on to cleaned steel surfaces is rapid and that their presence in the interface can be highly detrimental to bonding. It is therefore suggested that the bonding of steels under the relatively mild conditions described here requires that the bonding surfaces be continuously cleaned right up to the moment of contact between them.

Few other workers have examined the diffusion bonding of steels at temperatures in the region of the Ac ₁, those who did used simple cleaning procedures prior to bonding in conventional vacuum systems at pressures similar to those adopted in the present work. For example, Thornton and Wallach ^[6] bonded a pure iron, a 0.2%C steel, and 080M40 at 700-1000°C but were unable to obtain tensile fracture in the parent metal at bonding temperatures below 850°C. Elliott ^[3] also worked with 080M40 at 720°C and upwards but found that parent metal fracture could not be achieved below 1000°C. The fact that parent metal tensile properties were achieved after bonding at 700°C in the present work demonstrates that surface cleanliness does beneficially influence low temperature bonding and that the original premises were correct. But, equally, the fact that necking was not seen across a bond line neither in this, nor in Thornton and Wallach's, nor in Elliott's work demonstrates that the bond line retained unusual properties notwithstanding the cleanliness or otherwise of the bonding surfaces.

The evidence for the postulate of elastic constraint by interface voids is circumstantial, but strong, and Thornton has provided a theoretical justification. ^[7] The question remains, however, as to why the voids persist rather than closing completely as predicted by diffusion theory. ^[8] Work by Lea and Seah, e.g. Ref. ^[9] , demonstrated that impurities from the bulk segregate to cavities in certain temperature/time regimes and reduce sintering rates. Furthermore, examination of the present diffusion bonds (and those of Thornton and Wallach) by Lea at NPL, showed ^[5] that the surfaces of the interface voids were indeed partially covered by mono-molecular layers of impurities derived from the bulk, (P, S, N, Sb, Sn).

It is therefore suggested that two effects are at work in the diffusion bonding of steels. One is that surface contaminant films will prevent true metal to metal contact and must be removed before bonding can begin. (Removal is either by self-dissolution above 900°C or so, or by, for example, ion bombardment below 900°C). The second effect, in line with the conclusions drawn in Ref. ^[6] , is the diffusion, during bonding, of impurities from the bulk to the surfaces of interface voids and the subsequent interference by those impurities with the processes that otherwise cause void closure. It is the persistence of these voids which is believed to be responsible for the mechanical properties of the interface. Counteracting this effect presents a different problem from that posed by surface cleanliness.

Ion bombardment is evidently a successful way to clean surfaces before bonding, but it can be of no assistance in preventing the diffusion of bulk impurities to the walls of interface voids. On the contrary, it may even assist that process in low temperature bonding, in that the void surfaces will themselves have been cleaned before they were formed and so will reduce any surface contaminant barrier to segregation on to the walls.

The prospects for low temperature bonding

Although the tensile properties of joints made at 700°C can be close to properties of similarly treated parent metal, a bonding temperature of 700°C is still too high if full parent metal strengths are to be achieved (080M40 parent rod is considerably softened, and 709M40 is further tempered). The work done to date indicates that lower bonding temperatures (such as 600°C) are feasible, but the success of such joints would probably be even more dependent on surface cleanliness and would certainly require some method of preventing the contamination by diffusion of interface void surfaces. A diffusion barrier coating is one possibility.

Part 4 of this series will describe experiments designed to attack the problem of interface voids; methods of preventing their formation or eliminating their effects will be considered.

Conclusions

1. Effective elimination of surface contaminants is an important factor for the successful diffusion bonding of steels. A combination of heating at 700°C with ion bombardment appears to be sufficient for effective cleaning.
2. Parent metal tensile properties have been achieved in joints made by bonding cleaned surfaces of both a plain 0.4%C steel, and a heat treatable 1%Cr-0.25%Mo-0.4%C steel, at 700°C for 30min under 20 N/mm ², although the latter steel was appreciably tempered during bonding.
3. The bondability of steel at 700°C was destroyed by exposing previously cleaned surfaces to contaminants, specifically carbon monoxide or ethylene.
4. It is deduced that the presence of interface voids controls the tensile properties of the joint.
5. The present research supports the proposition of other workers that impurities in the bulk diffuse to interface voids during bonding and prevent their dispersal.
6. Further work is necessary to establish the influence of surface finish on bondability.
7. The joining at 600°C of cleaned surfaces of the steels studied here is considered to be feasible provided that attention is concentrated on eliminating interface voids.

References