Low temperature diffusion bonding of steels

TWI Bulletin, July/August 1988

1 Surface cleaning - experimental

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

This two-part article, the first in a series on low temperature diffusion bonding, describes surface cleaning of steels by autodissolution and ion bombardment. Following articles will report how such cleaning affects diffusion bonding below the A I temperature. Subsequently, work will be described on attempts to overcome the problem of the interface voids which characterise diffusion bonds.

Work carried out at The Welding Institute to examine the possibility of joining steels without inflicting substantial metallurgical and mechanical damage indicated that some form of diffusion bonding below the A I temperature was the most promising approach provided that the surface contaminants that impede diffusion, and the inevitable interface voids, could both be removed to allow bonding to proceed.

The objectives of the research programme devised to investigate this were:

a) To explore the physical and chemical properties of the surface of steel with the object of exploiting surface energy in joining processes;
b) To develop non-fusion joining techniques for steel that will substantially reduce the time/temperature combination required by present methods, yet without the necessity for large scale deformation.

Welding by fusion is now the dominant method of joining steel, but fusion techniques have introduced a host of problems that are making increasingly severe demands on steel production and fabrication. These problems largely stem from the practice of joining two cold, generally wrought, pieces of metal by what amounts to casting in situ; but even solid state joining techniques such as diffusion bonding or pressure welding require high temperatures, long times, or large deformations, which may have detrimental effects. There are, however, high inherent energies in a truly clean metal surface which, if they are properly used in a joining process, can do much to reduce the external energy requirements and to reduce the incidence of welding problems. The part of the project which this article describes was therefore designed to study the various factors involved in obtaining truly (or, at least, sufficiently) clean surfaces.

In essence, all fusion or solid state welding processes consist initially of surface cleaning followed, or accompanied, by conjunction of the two surfaces to form a bond. The surface of a metal which has been exposed to a normal atmosphere consists of a number of layers, chemically and physically attached ( Fig.1), which must first be removed before atomic contact is made with another metal. The physisorbed layers can be removed by standard high vacuum techniques, but the reacted oxide and chemisorbed layers are removed only by more drastic methods such as chemical etching or mechanical disruption.

Conventional diffusion bonding of steels is carried out at 1000° or 1100°C and relies for surface cleaning wholly upon the fact that iron begins to dissolve its own oxide and surface contaminants at a sensible rate at temperatures over, say, 800°C. Cold pressure welding, on the other hand, uses such massive deformations that the surface is stretched sufficiently to break up the surface oxide layers into small islands and thereby allow the two metal surfaces to come into atomic contact, the contaminants remaining finely dispersed at the interface. Whatever the method of cleaning, when two atomically clean solid iron surfaces are brought into contact, the unbalanced bonding energy that appears at those surfaces helps to unite the two components and to assist the processes by which the inevitable interface voids are removed. We can therefore consider the two processes of cleaning and joining separately, bearing in mind that one will always influence the other.

Surface studies

Materials

Three pure irons and three steels were used; bulk analyses are given in the Table. The irons are designated by a shorthand which abbreviates their total iron content, thus: high purity iron of 99.995% is referred to as 4N5 iron, equally, 99.98% is 3N8 iron, and 99.6% is 2N6 iron; they were chosen to examine the influence, if any, of impurity levels on surface cleaning. The steels were two plain C-Mn steels of 0.2% and 0.4.%C, and a heat treatable low alloy steel, also of 0.4%C. They were chosen to examine the influence of carbon and of alloying elements on surface reactions and also as being representative of general engineering steels.

Table 1 - Materials analysis, wt %

Table 1 - Materials analysis, wt % - continued

In all cases, the same specimen was used for each of the surface cleaning series (being re-prepared between each series), and was taken from the same batch of material as used in the later bonding studies.

Surface analysis

The various chemical species on the surface, and the way that they changed during heating and ion bombardment, were followed in the sample chamber of an X-ray photoelectron spectrometer at some 5x10 ^-8 torr. X-ray photoelectron spectroscopy (XPS, also known as ESCA) is sensitive only to the immediate surface ( e.g. three to ten molecular layers) of a sample. ^[1,2] The instrument detects and identifies all elements from lithium upwards and, by analysis of the spectral peaks, gives information concerning their chemical state ( e.g. the instrument can be used to distinguish between surface iron present both as the metal and as its different oxides). The technique examines an area of some 20-50mm ² and thus gives an averaged analysis.

Quantitative comparison of surface element concentration was such in this work that it was not possible to define an analysis accuracy over the whole range of elements but, as a general figure, +10% of the stated value can be regarded as a reasonable approximation. The lines drawn in the surface concentration graphs should therefore be regarded as indicating trends rather than as definitive values. Surface concentrations, given in atomic percent (at%), were derived by integrating the area under each spectral peak, applying various correction factors, and then proportionately summing to 100%.

Surface cleaning

The exploitation of surface energy requires that a steel surface be rendered atomically clean, or as near so as possible, and that the cleanness be preserved until the moment of contact of the surfaces to be joined. The research therefore studied the effectiveness of two factors; self-cleaning of pure irons and simple steels in vacuum, which is, in essence, what happens during the diffusion bonding of steels, and of cleaning by ion bombardment.

Cleaning by heating in vacuum

The conventional procedure for diffusion bonding of steels at about 1050°C is to clean the faces to be joined by degreasing and then to stack the parts so that the bonding surfaces are in close contact and are under pressure before being heated in a vacuum chamber. The bonding faces are thus effectively sealed off from the chamber atmosphere, and the oxides and absorbed gases, plus the gases trapped within the voids, are dissolved into the steel when heated without being renewed from the background gases within the chamber.

It is evident that such cleaning does take place at the comparatively high temperatures used in conventional diffusion bonding, but little is known about steel surface contaminants and their rate of dissolution at 700°C or lower. Although there is some information about impurity loss from steel surfaces by desorption ^[3] and of impurity diffusion from the bulk to the surface of steel in this temperature range, ^[4] there is little, other than that derived from simple gas kinetics, on the rate of recontamination of cleaned steel surfaces exposed to background chamber gases at these temperatures.

If cleaning by simple vacuum heating is to be used as part of a practical process below the A I temperature, its characteristics must be known so that its effects can be separated from others influencing void dispersal.

Ion beam cleaning

In addition to auto dissolution, there are many ways to clean freely exposed steel surfaces in vacuum, but of all the methods that are feasible as a means of achieving, or at least approaching, an atomically clean surface, ion bombardment is the most attractive for a research project. It is not temperature dependent, it is non-contacting and therefore can be used on any shape, the surface layer erosion rate is controllable, subsequent removal of the cleaning agent is simple, and it can be combined with intermediate layer joining techniques.

Bombardment cleaning with an ion beam, rather than a glow discharge, was chosen for this work because ion beams are well characterised in terms of their energy, current distribution, and specimen dosage rate, and the power in such a beam is so low that its heating effect can be ignored.

Contaminant gases

The pressure in the analysis chamber is of the order of 10 ^-8 torr but, even at this low level, there is an appreciable amount of gas present (chiefly water vapour and hydrocarbons). Figure 2 shows the relationship between gas pressure and some related variables and it indicates that, at 25°C and 10 ^-8 torr, the rate at which gas molecules impinge upon a surface is equivalent to the formation of a monolayer within 100 sec. Because a cleaned metal surface is reactive, residual chamber gases will rapidly adsorb on collision, rather than bounce off again, to form a layer of contamination which is substantial in surface analytical terms.

The surface analysis work on cleaning by heating and by ion bombardment did indeed suggest that contamination by gases present in the analysis chamber atmosphere played a part in the results obtained, particularly in those cases where ion bombardment was not used. Furthermore, as part of the bonding programme, it was necessary to determine the rates of contamination of steel surfaces exposed to a vacuum before bonding, and therefore a short study was undertaken to monitor the condensation rates of chamber gases on to a clean 080M40 steel surface. The work, which was regarded as a preliminary study, was confined to ambient temperatures only and involved the three reactive gases most commonly found in clean vacuum conditions, i.e. carbon monoxide, water vapour, and to simulate pump-oil vapours, ethylene (C ₂H ₄).

Experimental procedure

Heating in vacuum: surface analysis

Specimens were cut, longitudinally, from bar in the form of strip 15 x 7 x 2mm; they were metallographically polished on 1µm diamond powder and some were left in that condition while others were lightly abraded on 7µm silicon carbide paper; all were ultrasonically cleaned prior to insertion in the analysis chamber.

Surface analyses were performed by recording wide scan spectra from 500-1500eV (kinetic energy) at 20°C, and then narrow scan spectra of high resolution for each of the XPS peaks of interest. The temperature was raised in 100°C steps up to the limit of the instrument at 500°C, and spectral recordings were taken at each step to determine the surface changes that were taking place. The time taken for a complete series was some 5hr.

Ion bombardment: surface analysis

Specimens were prepared as above to the 7µm SiC 'abraded' condition, and experiments were carried out at 20°C and at 500°C.

Surfaces to be cleaned were bombarded in the analytical chamber by a beam of some 10mm diameter. The surface erosion rate generated by ion bombardment cannot be exactly defined because it varies by up to a factor of 10 according to the species being removed (metals fastest, oxides slow, elemental carbon slowest). The erosion/composition profiles for ion bombardment cleaning are therefore plotted in terms of bombardment time. A good general working rule however is to assume that bombardment for 600sec by a 6keV Ar-ion beam of 30 µA/cm ² (the conditions used in the later bonding work) results in an erosion rate of 3 nm/min, which is equivalent to about six monolayers of iron removed per minute. Erosion depths probably totalled no more than 50nm under the bombardment conditions used ^[5] and therefore had no influence on surface roughness.

Wide scan and narrow scan spectra were taken as before of the as-prepared surfaces at 20°C, and then each specimen was subjected to argon ion bombardment. After 30 sec, the ion beam was switched off and the scans repeated. A further 30 sec bombardment was followed by a third series of spectra, and so on at various intervals up to a cumulative total of about 15 min bombardment. An identical procedure was used for the series of experiments in which the temperature was first stabilised at 500°C after being heated, in approximately 20 min, from cold.

Adsorption of contaminant gases

The chamber was evacuated to approximately 5 x 10 ^-8 torr and the specimen surface was cleaned by 6keV argon ion bombardment for 15min according to the conditions already described. The surface was then analysed for carbon, iron and oxygen by narrow scans of the main peaks; each scan took 100 sec, this being the shortest time for data acquisition consistent with minimising contamination effects. The residual gases present in the chamber at 5x10 ^-8 torr were monitored by a mass spectrometer which showed that they consisted predominantly of nitrogen with small traces of oxygen, argon, water vapour and carbon monoxide. Nitrogen and argon were not detected on the steel surface by analysis.

A control, or blank, run was first carried out under continuous pumping by repeatedly analysing the undisturbed surface until it had been exposed to the residual gases in the chamber for 1 hr. The specimen was removed, re-abraded and washed, re-introduced into the chamber and the cleaning procedure repeated. The chosen contaminant gas was then admitted to the vacuum chamber under controlled conditions at a constant rate to achieve a partial pressure of 7.5 x 10 ^-7 torr (approximately 7 x 10 ¹⁰ molecules/cm ³, equivalent to an impingement rate of 7x10 ¹⁴ molecules/cm ²/sec, Fig.2) for 200 sec under continuous pumping. The gas was admitted to the chamber from a point such that it flowed over the specimen before entering the pump. After 200 sec, the gas flow was halted, narrow scans of the three elements were taken as before, and a further 200 sec exposure was given, followed by analysis until a total cumulative exposure time of 2000 sec was reached. The specimen was then removed, recleaned, etc and the whole procedure repeated with another contaminant gas on the same specimen.

Part 2 looks at the results and the conclusions reached.

Acknowledgements

Funding was provided by Research Members of The Welding Institute, the Engineering Materials Requirements Board, and the Commission of European Communities.

References