Low temperature diffusion bonding of steels

TWI Bulletin, September/October 1988

2 Surface cleaning - results and discussion

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

Part one of Ian Bucklow's work on low temperature diffusion bonding of steels looked at experimental procedures of surface cleaning of steels by autodissolution and ion bombardment. Here he reveals the results of that work and some of the conclusions he reached.

Experimental results

Cleaning by heating in a vacuum

The results for 4N5 iron are shown in some detail as an example of the raw data and how they are treated, but all other results are plotted as surface element concentration versus the experimental variable.

4N5 iron (<0.001%C), abraded surface

The wide scan spectra for 20, 300 and 500°C are shown in Fig.3. No elements other than those shown were present in a concentration sufficient for reliable detection ( i.e. all less than 1at%). There was little detectable change between 20 and 300°C, but at 500°C the Fe peaks sharpened somewhat and were higher, the nitrogen peak was more obvious, but the Si peaks were barely detectable. Again, no elements other than those shown were detected. All these observations are indicative of an increased degree of surface cleanness at 500°C.

In Fig.4, the Fe(2p ^3/2) envelope is expanded in narrow span spectra at various temperatures. The Figure illustrates both the change of oxide from magnetite at room temperature to haematite at 300°C and above, and the emergence of metallic Fe as the dominant form at the higher temperatures.

Figure 5a shows the variations with temperature of the integrated peak areas of representative iron, oxygen and carbon peaks; no distinction is made between the various compounds in which these elements occur on the surface. All other elements in the surface totalled no more than 3at%. Figure 5b shows the corresponding surface element concentrations derived from Fig.5a with uncertainty bands drawn about the mean lines. A comparison of the two figures emphasises the necessity to apply analytical corrections before conclusions can be drawn about the composition of a surface.

The total quantity of oxygen on the surface dropped slightly with increasing temperature up to 250°C - this is consistent with surface outgassing - and thereafter fell slowly. The Fe content of the surface was constant at 20% up to about 300°C, but above this temperature it started to rise. The carbon signal began to increase above 100°C and reached a maximum at 250°C, but thereafter fell steadily. The apparent initial increase in C was probably connected with the early loss of physisorbed gases, etc, which thereby exposed more of the underlying carbonaceous contaminants, but there were other factors operating which will be discussed later.

It appears that surface contaminants were fairly stable up to about 300°C, but above that temperature they began to decrease somewhat although the process was not complete at 500°C (the maximum at which the instrument could be used). The major constituents were carbonaceous in nature, with oxide present in considerable quantities.

All materials, abraded and polished surfaces

The extent of metallic Fe exposed by vacuum heating of abraded and polished surfaces is shown in Fig.6. The steels were generally cleaner than the irons, and abraded surfaces were initially cleaner than the polished, but the latter appeared to be more responsive to heating and the curves suggested that they would probably become the cleaner above 500°C.

The corresponding surface carbon concentrations ( i.e. all forms of carbonaceous contaminants) are shown in Fig.7 where, again, it is seen that there was less contamination on the steels, and that polished surfaces were rather more responsive to heating. Generally, however, the carbonaceous contaminants were fairly stable over the temperature range studied.

Ion bombardment

The atomic percentage of exposed metallic iron as it changed with cumulative ion bombardment time at 20°C is shown in Fig.8a, and with bombardment after rapid heating to 500°C, in Fig.8b. Corresponding curves for total carbon are shown in Fig.9.

At 20°C there were rapid early changes as loosely bound contaminants were removed - carbonaceous materials evidently forming the major loss which constituted some 10 to 15 monolayers. The loss of these layers exposed the more firmly bound surface films which were eroded more slowly, but stability was reached after some 10min bombardment time. The steels consistently displayed a local maximum in the Fe curves after 3-6min, with a corresponding dip in the carbon curves, but the one iron studied showed no such tendency.

At 500°C the initial levels of contamination were rather lower than at 20°C, and the general pattern of erosion was the same, but the eventual cleanness at 500°C was much higher on the impure iron and on the three steels. The impure iron behaved in a similar way to the steels, but the high purity iron underwent little cleaning after the initial loss of loosely bound contaminants - the surface carbon level was particularly high.

Adsorption of contaminant gases

The results of trials to investigate the condensation rates of chamber gases on to a clean 080M40 steel surface are described below.

1. Control run: Fig.10 shows the maximum heights of the carbon, oxygen, Fe (metal) and Fe (oxide) peaks (note that the graphs are not converted to surface concentration). It is evident that stability was not achievedafter 1hr exposure.
2. Carbon monoxide: Fig.11a demonstrates that the same pattern of behaviour as above occurred, but that carbon and Fe (oxide) levels were higher than the control run, and Fe (metal) fell more rapidly.
3. Ethylene: Fig.11b shows the behaviour of the peak height signals with continued exposure, and Fig.11c shows the shape of the C(1s) peak at various times. The rise in the carbon signal in Fig.11b, although apparently small, is significant in view of the relative insensitivity of the technique to carbon. The significance is illustrated in Fig.llc where changes in the shape of the C(1s) envelope indicate thatprolonged exposure led to a decreasing amount of elemental carbon on the surface and its replacement by (CH ₂)n.
4. Water vapour: a behaviour pattern similar to the control run was seen, but the oxygen level was higher and the Fe (metal) level correspondingly lower.

Discussion

Surface cleaning

A comparison of cleaning by heating with cleaning by bombardment was complicated by the procedure of following the former in stages because the resultant slow heating in vacuum to 500°C took some 5hr, whereas heating for high temperature ion bombardment took less than 30min. In view of the rate of collision with the surface of residual gases at 10 ^-8torr, it is not surprising that slowly heated surfaces carried more impurities at 500°C than those rapidly heated. Nevertheless, although heating without bombardment to 500°C was not particularly effective in cleaning, it does appear that the higher carbon steels were beginning to lose their contaminants at an increasing rate with temperature, which suggests that the surface would be much cleaner at a bonding temperature of 700°C. The appearance ( Fig.3) of nitrogen on the surface at 500°C is in keeping with the observations of Honda and Hirokawa who suggested that it had diffused to the surface from the bulk at a faster rate above 300°C than it was lost by desorption from the surface. This observation constitutes a significant parallel with others made in diffusion bonding experiments to be described later.

The question remains as to why the steels were more effectively cleaned by vacuum heating than the irons, and, particularly, why carbonaceous contamination on high purity iron was so persistent even under ion bombardment. One suggestion is that carbon in solution reduces the surface reactivity for carbonaceous materials. The impure 2N6 iron with 0.03%C would be saturated and would therefore behave rather more like a steel than the purer 3N8 iron. Whereas the high purity 4N5 iron, with a carbon content in the 0.001% range, might be sufficiently reactive to form much stronger bonds than those found on steel surfaces. The presence of other impurities in the bulk or of cementite in the surface may also influence surface reactivity for carbonaceous contaminants.

Fluctuations in the Fe(2p ^3/2) signals under ion bombardment as shown in Fig.8, and possibly also in the carbon signals, may in some part result from the ion bombardment itself. It is known ^[6] that ion bombardment can partially reduce some metal oxides, and there is also evidence ^[7] to suggest that the oxygen thus released may be forced back into the substrate to a slight extent. If these effects are in operation, they could account for the similarities in the shapes of the composition/depth profiles which are seen in all the ion bombardment studies, both at 20°C and at 500°C. The 20°C surfaces generally have high initial levels of oxygen and carbon signals from airformed contaminant films, whilst the surfaces rapidly heated to 500°C initially have lower, but nevertheless significant, levels. Under bombardment, the carbon signals generally drop sharply within the first 30sec ( i.e. 5 monolayers) as the airformed contaminants are swept away, thus exposing the underlying ones. The oxide layer begins to be eroded but, at the same time, its surface becomes chemically reduced as a result of bombardment, and some of the release oxygen is driven back into the film. The rate of erosion is, however, faster than that of other effects and so the oxygen rich layers of oxide (created by bombardment burial) become exposed leading to a rise in the oxygen signal and a fall in the Fe emission. Continued erosion eventually begins to expose the underlying iron metal so that the total Fe signal increases again to a steady value which probably represents a dynamic equilibrium between erosion and recontamination. In many cases, the carbon signal also displays a secondary maximum which may also be due to bombardment burial.

In the case of 'conventional' diffusion bonding when two surfaces are pressed into contact at room temperature in air, the surface contaminants will be those illustrated in Fig.1, but the gases trapped in the voids will consist largely of nitrogen, oxygen and water vapour, a major proportion of which must be absorbed by the metal after the surface contaminants have dissolved. In the case of free surfaces heated in vacuum, however, the observed cleaning rates represent a balance between solution and recontamination and therefore may not be equivalent to the trapped gas state. These points must be borne in mind when comparing the two conditions.

Adsorption of contaminant gases

At 5 X 10 ^-8torr, the conventional monolayer recombination time at 20°C is about 30sec ( Fig.2), but this figure is calculated by assuming a sticking coefficient of unity, i.e. it assumes that each molecule that collides with the surface adheres to it. Sticking coefficients are usually much less than unity and depend upon the various species involved, thus the control run of Fig.10 suggests that monolayer coverage may not have been achieved until times much longer than 30sec. However, coverage probably occurred well within 500sec i.e. before the Fe (metal) and oxygen (total) curves began to change their slopes. A qualitative interpretation of Fig.11 is that the most active constituent in the residual gases was oxygen, either as water or as CO. Analysis of the O(1s) peaks indicated that oxide increased by up to about 200sec, but that adsorbed OH (from water in the residual gas) continued to build up steadily with exposure on top of the oxide. The low sensitivity factor of carbon gives rise to the misleading impression of the percentage of carbon present in Fig.11b. The recorded intensity probably represents a surface concentration in the region of 30at% after 1 hr exposure.

Comparing the curves for carbon monoxide ( Fig.11a) with the control run ( Fig.10), it is seen that the carbon, Fe (oxide) and oxygen signals all increase more rapidly in the presence of carbon monoxide and that the surface was virtually stable after some 1500sec, implying that the saturated surface no longer adsorbed CO or even water from the chamber gases.

Exposure to ethylene, Fig.11b, produced a more complex behaviour; the low sensitivity to carbon again gives a misleading impression but Fig.11c suggests that a high proportion of the surface carbon was in the elemental form. The presence of elemental carbon in small quantities was detected in the early stages of CO contamination and its presence was suspected in the water vapour contamination runs, but no trace of it was seen in the control run.

The interpretation of the foregoing observations is that an initially clean surface of steel quickly takes up the principal gaseous contaminants in its vicinity. If the contaminant is water vapour, the reaction produces an oxide which probably forms a monolayer rapidly but thereafter thickens only slowly in spite of continued adsorption of water vapour on its surface. If the contaminant is carbon monoxide, again an oxide is formed and the resultant reduced carbon remains on the surface. If the contaminant is ethylene, there again appears to be a chemical reaction with the iron which produces a comparatively high level of elemental carbon, but reaction with other chamber gases also generates oxides and layers of physisorbed bound oxygen.

It must be remembered that the instrument cannot detect hydrogen, so nothing is known about the possible solution, by the metal, of hydrogen from water vapour or from ethylene.

Summary

The work demonstrated that ion bombardment was a more effective cleaning method than simple heating to 500°C, and was even more effective when a hot substrate was treated. It is therefore probable that surfaces ion bombarded at 700°C and immediately joined (the actual procedure used in the bonding experiments to be described) were much cleaner than the present analyses suggest, but in all the cases studied, steels appeared to be easier to clean than high purity irons.

The contamination experiments highlight the necessity, for the low temperature bonding of steel, to devise some form of surface cleaning that can be applied in vacuum and continued right up to the time of contact of surfaces to be joined. It must be remembered that the levels of vacua in industrial bonding equipment will be poorer than those used in these experiments, and contamination will be that much faster.

Subsequent articles in this series will describe the diffusion bonding at 700°C of steels cleaned under the conditions described here, and discuss the relationship between surface cleaning and bonding.

SI units - 1 torr = 133 N/m ² or Pascals.

References