1 Introduction

The China Spallation Neutron Source (CSNS) consists of a negative hydrogen linac, a rapid cycling proton synchrotron (RCS) accelerating the beam to 1.6 GeV, a solid target station, and instruments for spallation neutron applications [1]. The transverse beam collimation system, with a primary collimator and four secondary collimators, is the key device in RCS. The halo particles, with increased emittance after scattering from the beam scraper in the primary collimator, are absorbed by the downstream secondary collimators after some phase shift [2, 3]. Different materials were studied as a candidate for scraper material by the physics department of CSNS. Tungsten, copper and tantalum materials impart a large scattering angle to the halo particles. They are of high melting point, good thermal conductivity, and a moderate price. From the three materials, tungsten is chosen to fabricate the primary collimator scraper, with a thickness of 0.17 mm [4, 5]. Thin and fragile, the tungsten slice is subjected to energy deposition by the particles. An extended power deposition heats the slice up, and an effective heat dissipation design is to weld the tungsten foil onto an oxygen-free copper block, which is connected to a Φ34 mm copper rod that transfers the heat to cooling structure outside the vacuum chamber.

As shown in Fig. 1, the beam scraper of the primary collimator was designed as a 0.17 mm tungsten foil brazed on a double face-wedged copper block of 121.5 mm×20 mm. Tungsten and copper differ greatly in thermal properties, especially, thermal expansion coefficient and thermal capacity (Table 1). It is hard to weld them together [6, 7], and the two materials of this structure contrast sharply in thickness, making the task even harder. [8]

Fig. 1Full-screen View

Structure of the W/Cu assembly.

The high thermal expansion ratio, and large shrinkage during cooling, causes great thermal stress, which tends to create brazing cracks [9]. Methods to minimize stress include: improving the structure technology, using brazing filler metals with the lowest possible fusion point, and inserting an adequate cushion layer between the base metals [10-22]. In Ref.[8], we found that the recrystallization temperature of a 0.17 mm tungsten foil was 1025°C and the best brazing strength could be achieved by vacuum brazing technology, using Ag-Cu-Ti as brazing filling metals, and holding at 890°C for 10 min. In this paper we focus on minimizing W/Cu brazing thermal stress by improving the structure and inserting a transition layer, by finite elemental analysis (FEA) and experimental tests, towards a proper brazing and processing plan for heat treatment of the tungsten foil, which led to the successful acceptance of the primary collimator R&D.

2 Methods

To simplify the analysis, we did not consider plastic or brittle deformations, and effects of material welding, and just focus on reducing thermal stress by structure improvement and buffering inlet. Welding was simulated by a heat-up-to-cooling -down process of tungsten foil and copper block. Transient Thermal-Structural Coupled Analysis was utilized to obtain thermal stress in the cooling down process from welding temperature to room temperature.

The loading and constraint conditions of the analysis were:

(1) Thermal loading The copper block and tungsten foil were allowed to cool in vacuum, and temperature of the entire model dropped from 890°C to 25°C in 25 s and maintained at 25°C for 5 s; and

(2) Fixed loading Structural analysis was performed on the copper block bottom. As all the models were symmetrical about the XOY plane, to save computation time, half of the model was considered in the analysis and symmetrical constraints were applied to the plane of symmetry.

From the FEA results, a proper plan was established for test research so as to validate the reliability of analysis.

3 Improvement of the structure technology

As thermal stress is mainly concentrated on the joint, the tungsten foil is of minimal thickness, and tungsten material is hard and brittle, so there is little possibility of structural improvement at its joint structure. Therefore, structural improvement is only possible at the copper block joint.

The assembly consisting of different copper structures at the joint and the tungsten slice were analyzed. The temperature drop trends of all the models in the thermal analysis are shown in Fig. 2(a). In 0–25s, the temperature difference of the models remained basically the same, while in 25–30 s the temperature difference narrowed down to 0°C. Fig. 2(b) shows a typical analysis models and the constraint conditions. Temperature distribution at different seconds was treated as a loading in structural analysis and was introduced into the model. A fixed loading was applied to the bottom of the copper block. Fig. 2(c) shows the maximum stress at different seconds for the model in Fig. 2(b). The higher the minimum temperature is, the larger the temperature difference and the maximum stress of the model. Fig. 2(d) shows the stress analysis results at 1.9 s loading time with the temperature distribution being used as transient structural analysis loading. The maximum stress fluctuates from the beginning and finally stabilizes around 400 MPa.

Fig. 2Full-screen View

Thermal and structural analysis results.

Fig. 3 show stress distribution in the copper and the tungsten assembly for different copper structures at the joint under 2 ms loading corresponding to the same time point. It is clear from that the structural analysis located the maximum stress on the tungsten foil, on the joint to be specific. Different structures at the joint had considerable impact upon the stress distribution in the W/Cu assembly. The thermal stresses in Structures 3 and 4 were far less than those in Structures 1 and 2.

Fig. 3Full-screen View

Stress distribution in different copper structures at the joint and in the tungsten foil, under 2 ms loading at the same time point.

To validate feasibility of the FEA, copper blocks of different structures were fabricated and welded with a 0.17 mm tungsten foil, as shown in Fig. 4. The double face-wedged head for tungsten foil welding was in structures of no glooves, fine grooves, nonuniform grooves, 1-mm groove in every 2 mm, and 1-mm groove in every 3 mm.

Fig. 4Full-screen View

Brazing test specimens of different copper structures. The double face-wedged head to weld with the tungsten foil was in structures of (a) no glooves, (b) fine grooves, (c) nonuniform grooves, (d) 1-mm groove in every 2 mm, and (e) 1-mm groove in every 3 mm.

We found that the W/Cu joint structure did have a great influence upon the W/Cu brazing quality. When the stress-relief grooves were fine (Specimen b), no fold appeared on the tungsten foil, but less brazing wire was fused into the void between the tungsten and copper. The smaller the W/Cu contact area was, the poorer the brazing quality was. By changing structure and size of the stress-relief groove, we had Specimens c, d and e, The brazing quality of Specimen c and d was somewhat improved, but the tungsten foil folded apparently; while a not-so-bad balance was achieved between the tungsten foil deformation and brazing quality of Specimen e. However, the small folds of tungsten foil could not meet CSNS's physics requirements. If a 0.17 mm tungsten foil was welded to a 20 mm copper block, the foil could hardly resist the shrinkage stress caused by the thermal expansion of the two materials. Improvement ought to be made by other methods.

4 Insertion of a transition layer

To substantially reduce brazing stress in the tungsten foil, a transition layer has to be introduced in addition to the copper structure improvement. As two pieces of the same metals do not have the above brazing problems, the most reliable way to reduce the brazing stress in the tungsten foil is to weld a transition layer of tungsten sheet in suitable thickness on the copper block, and the transition layer was later welded with the tungsten slice. Based on this philosophy, tungsten sheets in different thicknesses welded on copper blocks of the same structure were analyzed. The stress distributions of the general model with a transition tungsten sheet of 1.5 and 2 mm thick are shown in Fig.5. Finite element analyses were performed on the assembly with the tungsten foil and a tungsten sheet of different thicknesses. With the sheet thicknesses of 1, 2, and 3 mm, the maximum stresses were 0.80, 0.87 and 1.65 MPa, respectively. So, when the tungsten sheets do not differ much in thickness, the thickness effect is small enough to be neglected.

Fig. 5Full-screen View

Analysis results of a copper structure with tungsten transition layer of different thicknesses, under 2 ms loading at the same time point.

Based on the FEA results, a brazing test was conducted. To ensure brazing quality of the tungsten foil, it was necessary to fabricate a number of grooves on the transition layer. This would be a tough machining task for the transition layer was too thin. Therefor, 2 mm tungsten sheet was used as the transition layer. The sheet was brazed with the copper block, grooves were machined on it, and the tungsten foil was brazed on the transition layer. Fig. 6 shows the specimens of W/Cu brazing structure with a 2 mm tungsten transition layer. Brazing quality of the W/Cu structure with 0.17 mm tungsten foil well meets the physics requirements of the CSNS project.

Fig. 6Full-screen View

W/Cu brazing specimens with a 2 mm tungsten transition layer.

5 Conclusion

We have explored on W/Cu brazing of large difference in size of the parts, and come up with a solution whereby the copper block structure is modified at the W/Cu joint and a tungsten transition layer of adequate thickness is inserted to cope with the problems of dissimilar metals and disparate thermal expansion coefficients. The feasibility of this plan has been validated by tests, ensuring the successful acceptance of the CSNS primary collimator. This study may be of help for developing a brazing plan for similar structures.