Joining of graphite and copper for nuclear applications
TWI Bulletin, May/June 1995
Wendy Hanson joined TWI in 1992 and is a Senior Research Engineer in the Ceramics Group in the Advanced Materials and Processes Department. Before this, she studied Metallurgy and Engineering Materials at the University of Strathclyde. She gained a PhD at Strathclyde for a study of the fabrication and properties of silicon carbide ceramic matrix composites. Her work at TWI involves joining ceramics to metals, polymer derived ceramics, high temperature inorganic adhesives and glass-ceramics.
An Industrial Member from the nuclear industry recently posed TWI with a tough graphite-copper bonding problem. Wendy Hanson takes up the story.
On 9 November 1994, the first muons were produced from a new £10 million beamline facility on ISIS, the world's most powerful pulsed neutron and muon source, at the Central Laboratory of the Research Council's Rutherford Appleton Laboratory, (RAL). To increase the yield of muons the engineers specified a change in target design. This design change caused the targets to fail during manufacture, which hindered work at the facility, and TWI was approached to solve the problem. A modification to target shape, joint design and joining procedure were developed by TWI in under four weeks of the problem first arising. Copper-graphite assemblies are used as targets for pulsed proton beams. Among the products of this interaction are muons which are subsequently collected for analysis. The graphite used is pyrolytic grade and is engineered to have a lamellar structure, perpendicular to the proton beam, such that the heat generated within the target during operation flows vertically through the graphite to the copper base, Fig.1. Copper is used since it has a high thermal conductivity and is subsequently water cooled to maintain efficient heat removal.
Fig. 1. Design of proton target using a rhombohedral cross-section
To alter the yield of muons, the designers/physicists at the ISIS facility specified a design change to produce a range of graphite thicknesses, where a greater thickness gives a greater muon yield. The shape of the copper heat-sinking disk was to remain relatively unchanged and a graphite rhombohedral section was required in three thicknesses: 3.5, 5 and 7 mm, which would then offer the scientists a range of operational parameters.
To join the copper and graphite, a range of techniques could be considered. A mechanical fit is not possible, since intimate contact is required between the graphite and copper to gain effective heat transfer; adhesives are unsuitable since they cannot be used in an ultra high vacuum environment, cannot withstand the operating temperatures required nor does it give good heat transfer; diffusion bonding offers a solution where no interlayers are required, but the high pressures required may distort the components during bonding, and it is also economically less viable.
The method used which allows the most versatility and simplicity is brazing. This is a clean, commercially attractive process which forms a good, thermally conductive bond between the copper and graphite.
When the brazing of these new shapes was first attempted at RAL, the graphite sections all failed as a result of cracking at the sharp-edged comers. Initially RAL felt that their own procedures/equipment could be at fault and had identical experiments carried out at other organisations. Results were the same. RAL visited TWI at the end of November 1994 to discuss the problem and potential solutions. A quick response to the problem was paramount, since the new muon target designs had to be operational before Christmas 1994.
Examination of the failed bonds produced at this stage showed classical residual stress cracks emanating from each corner of the graphite, Fig.2. This was due to both the large difference in coefficient of thermal expansion (CTE) between graphite and copper (0.1 x 10 -6/°C and 18 x 10 -6/°C respectively) and also the presence of sharp corners which act as stress concentrators.
Fig. 2. Example of cracking in a copper-graphite bond
The potential ways of resolving this problem were:
- a change in braze alloy - to reduce the brazing temperature and hence reduce the thermal stresses induced during the heat treatment.
- modification to the copper heat sink - either by changing the composition to reduce the CTE (by alloying with a low CTE metal such as tungsten), or a change of dimensions to reduce the cross-sectional thickness.
- modification to the design of the graphite - removal of the sharp edges which act as local stress concentrators.
Braze material
The current graphite targets used within the ISIS facility have been brazed with a commercial Ni-Cr-P alloy. This was selected since it was known to wet onto graphite satisfactorily. The required brazing temperature is a minimum of 940°C and the intermetallics formed between the braze and the graphite, which enable wetting, are based on chromium carbide, a relatively brittle compound.
An active metal braze alloy containing Ti which is capable of wetting on to both graphite and copper is Ag-Cu-Ti alloy. This alloy brazes at 850°C and forms a titanium carbide with the graphite. The combination of reduced brazing temperature and the formation of a slightly more elastic carbide phase could alleviate some of the cracking problems encountered. Experimentally, this one change in brazing procedure (and no change to either the graphite or copper), reduced the cracking in each component by 50%. Therefore, this method could be used in conjunction with the other design modifications.
A further alteration made to the use of the braze alloy was the use of 'stop-off' compound (based on titanium oxide) near the joint area. Previous joints appeared to have failed at a point of initiation close to areas of excess braze material. These areas of excess braze wetting form due to flow of the braze as it wets onto the two substrates.
To inhibit this excessive braze wetting, stop-off solution was painted around the joint area, so that braze could only wet on to the bond area itself.
Modification to the copper heat sink
The proposed modification to the composition of the copper by alloying it with a lower thermal expansion material such as tungsten or molybdenum was unacceptable, since the thermal conductivity of the material decreased. Therefore, the only acceptable modification to the copper was a reduction of the cross-sectional thickness at the joint area. This was reduced from 3mm to 1.5mm on the underside,
Fig.3. This reduction in cross-sectional area is the maximum allowable such that sufficient heat sinking is still possible.
Fig. 3. Copper heat-sinking disc (recessed by a depth of 1.5mm on the underside)
In addition, advantage was taken of the CTE mismatch between the materials by slotting a groove (0.4mm deep) in the copper, Fig.4, such that on cooling after brazing, the copper contracts on to the graphite thus forming a mechanical bond as well as that of the chemical bond with the braze.
Fig. 4. Upperside of copper heat-sinking disc having a 0.4mm deep groove (note that the lowerside is identical to that shown in Fig.3)
Modification to graphite target design
The pulsed proton beam only uses a central ellipse of approximately 30 x 40mm 2 on the target; therefore, small modifications to the joint area were acceptable, since this would not affect its operating capacity.
To inhibit the formation of cracks, it was important that there were no sharp contours on the graphite at the joint interface. For this, part of the basal contact area was machined away and then rounded off through a polishing procedure to produce the shape shown in Fig.5. The designers at RAL worked with TWI at this stage to establish the minimum contact area that was acceptable such that efficient heat transfer would still be achieved.
Fig. 5. Modified shape of graphite used for new targets
The hermeticity of the bond was further improved by fine polishing of the graphite joint area prior to bonding. In this way all sharp edges were removed from the joint area and a surface finish of 0.15 microns was achieved on the graphite. The roughness and surface finish can influence the strength of bonding, as a rough surface can prevent complete contact at an interface and can damage the ceramic (in this case graphite) because of severe residual stresses caused by deep scratches or poor surface finish. Conversely, there will be an effect of anchoring on a rough interface due to mechanical keying. In reality, it appears that it is a combination of these two properties which influence the mechanical properties of the joint. Wherever possible, a fault-free surface is preferred in order to inhibit cracks manifesting from residual flaws or surface imperfections.
A combination of the above three modifications allowed the production of sound graphite-copper bonds for proton targets. In summary these were:
- use of a lower temperature braze alloy
- reduction of the copper thickness at the bond areas and slotting to produce mechanical keying as well as chemical bonding
- polishing of the graphite to remove surface irregularities and removal of sharp corners from the joint area.
The successful graphite-copper bonds were produced and delivered to RAL two weeks before Christmas 1994, and are now operating within the CLRC ISIS facility.
