Materials Science & Engineering A 586 (2013) 25 30

Contents lists available at ScienceDirect

Materials Science & Engineering A

journal homepage: www.elsevier.com/locate/msea

A new experimental method for identifying the conditions necessary for diffusion bonding in free cutting steels
S. Afshan a, D. Balint a,n, D. Farrugia b, J. Lin a
a b

Department of Mechanical Engineering, Imperial College London, South Kensington, London SW7 2AZ, UK Swinden Technology Centre, Tata UK Ltd, Moorgate Rd, Rotherham S60 3AR, UK

art ic l e i nf o
Article history: Received 1 March 2013 Received in revised form 1 July 2013 Accepted 11 July 2013 Available online 31 July 2013 Keywords: Diffusion bonding Void closure Free cutting steel Porous plasticity Densication

a b s t r a c t
Porosity in cast bloom steel may lead to reduced strength or inconsistency of nal rolled bar products and components if not properly closed and healed during the hot rolling process. Partial or complete recovery of strength in such porous materials can be achieved by diffusive healing processes at elevated temperatures. Devising an appropriate healing process that does not cause discontinuity in the microstructure, and in the mechanical properties at the bonding sites, whilst preventing distortion of the component during bonding requires an accurate choice of thermo-mechanical processing parameters. Despite work carried out on optimising diffusion bonding in materials such as titanium alloys, aluminium alloys and copper, the diffusion bonding process in free cutting steels has received relatively little attention. To support thermomechanical process optimisation (e.g. rolling) and calibration of theoretical models, a new experimental method was developed to determine combinations of load, temperature and time sufcient for complete diffusion bonding in as-cast Free Cutting Steels (FCS). The extent of diffusive healing and bond strength were examined by tensile testing to failure corroborated by SEM examination of the bond line. This enabled optimal loading conditions for the formation of a complete, strong bond to be identied. & 2013 Elsevier B.V. All rights reserved.

1. Introduction Continuous casting is used to solidify most of the 750 million tons of steel produced in the world every year [1]. The process reduces the number of required milling stages and results in semi-nished products such as billets, blooms and slabs which will later be rolled into more specic shapes and nished products (e.g. bars, rods, plates, etc.). Extending the range of nished product sizes produced from a given concast bloom or billet section is often limited by the minimum area reduction required to ensure effective central consolidation and nal mechanical properties. Porosity in as-casts billets could be caused by factors such as entrained air during lling, solidication shrinkage, mould wall reactions and dissolved gases [2]. Predicting effective consolidation or level of remnant porosity for a range of steel grade (function of solidication regime), billet size, pass schedule/roll design and thermo-mechanical conditions has always been an important issue for steel producers as it will affect the mechanical properties of nal products (strength, ductility, etc.) [3]. It is known that partial or complete recovery of strength in such porous materials can be obtained by pore closure and diffusive healing processes at elevated temperatures. Healing by reducing material defects such as voids and cracks to the point where diffusive bonding may occur, resulting in a

Corresponding author. Tel.: +44 2075 947084; fax: +44 2075 947017. E-mail address: d.balint@imperial.ac.uk (D. Balint).

stiffer and stronger material with enhanced mechanical properties, has been studied for many material applications. This method of healing has proven very successful in applications involving polymers and composites [4,5], biomaterials [6,7] and also for recovery of concrete [8,9]. Metal healing has been mostly studied in terms of sinter powder metallurgy where mass transfer of metal occurs at high temperatures in the range of 0.70.9 of the melting temperature, producing either solid state or liquid phase bonding across powder interfaces [10]. Void healing has been studied as the nal stage of crack healing [11,12] and metal bonding [13,14]. In crack healing, pore formation is due to crack splitting, whereas in metal bonding, voids form as a result of contact and deformation of surface asperities. There have been many attempts to model the void shrinkage process; however these models use different approaches. Derby and Wallach [13] developed a mathematical model for the bonding process (elimination of bond interface voids of cylindrical shape), predicting the overall bonding rate and also the dominant mechanisms as bonding progresses. They studied the diffusion bonding of copper experimentally, where the extent of bonding was determined using optical microscopy, and the results were compared with the predictions of their model. Later Guo and Ridley used the same approach to develop a mathematical model for elimination of bond interface voids of irregular shapes [14]. They used the experimental data from Derby and Wallach to validate their model. The optimal conditions to create high quality diffusion bonds have already been reported for some titanium alloys, aluminium alloys and magnesium alloys [1517].

0921-5093/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.msea.2013.07.056

26

S. Afshan et al. / Materials Science & Engineering A 586 (2013) 25 30

To improve the efcacy of the void healing process and calibrate diffusion bonding models for free cutting steels, it is essential to identify the conditions that lead to complete diffusion bonding. The main factors affecting the quality of the bonds are temperature, pressure and time. The effect of these parameters on damage recovery has been investigated experimentally by Han et al. [18] on crack recovery in 20MnMo steel, which indicated that damage recovery could be achieved without plastic deformation, just by increasing the temperature and prolonging the holding time. However, another experimental study on crack healing in 1045 steel showed that even after a 120 min heat treatment at 1100 1C, micro-voids were left in the crack healing area [19]. It is well known that void closure is more easily and quickly obtained if plastic deformation takes place under the inuence of compressive hydrostatic pressure [20]. Wang et al. [21] found that pore closure occurs at a certain level of hydrostatic pressure and accelerates at high temperatures. It was also shown that the holding period of the pressure in the compressive state affects the degree of healing, together with the state of oxidation in the vicinity of the porosity. Although void closure and healing have been widely studied for different alloys, the conditions required to create self-bonding of FCS cast steel porosity has never been investigated, and is the main focus of this paper. It is also noteworthy that in the majority of studies on diffusion bonding, an examination of the bond strength has not been carried out. In most cases SEM and optical microscopy have been used, focusing on observation of the bonding line. The approach described in this study offers a way of identifying combinations of load, temperature and time that lead to complete diffusion bonding by testing the mechanical strength of the resulting bond. SEM analysis has also been used to corroborate the mechanical testing by examining the extent of healing at the bond line. The optimal conditions for diffusion bonding with respect to the requirements of a thermomechanical processing method (e.g. rolling) were identied from those that led to complete diffusion bonding in the experiments; load applied in rolling is linked to the cross sectional area reduction, and a higher rolling speed increases efciency and output, hence it is desirable to reduce both the load and time required for complete bonding at a given temperature whilst ensuring a strong bond is formed. A secondary aim of this work was to provide a test method that can be used for calibration of void elimination models, such as that presented in [22], based on void closure according to Gurson Tvergaard porous plasticity, and healing by creep and diffusion according to the Pilling model of diffusion bonding [23].

were aligned, under various combinations of load and time at a given temperature. The load and the holding time were varied in a systematic way to ensure that accurate estimations of the required load and healing time were obtained. The specimens were subsequently tested in tension to assess the extent of the healing process by inspecting the resulting ow curves. The tests were conducted using a Gleeble 3800, which is a fully integrated digital closed-loop control thermal and mechanical testing system. Desired temperatures were achieved by direct resistance heating, with temperature at the bond line measured and controlled using thermocouples. To ensure the same temperature was achieved on both sides of the bond line, thermocouples were welded at a distance of 2.5 mm on either side of the bond, and their readings were maintained within 20 1C throughout the tests. The Gleeble was operated under load control with copper grips. The direct resistance heating mechanism of the Gleeble requires sufcient initial contact between the two halves of the specimen. To achieve this, the specimens were preloaded. It was found that an applied load of -0.5 kN, corresponding to a pressure of 6.4 MPa for a 10 mm diameter sample, maintained for a minimum of 30 s produced the contact conditions necessary to achieve a temperature of 1000 1C by direct resistance heating. The 30 s preloading time also served as a soaking period to achieve a uniform temperature along the gauge of the specimen. Following the preloading step, the force was increased until one of three target values was reached, and the specimens were held at that load for varying time periods. In all cases the loading rate was adjusted by the control system such that a constant strain rate of 0.01 s 1 was maintained. Since the sample temperature was uniform, cases where the applied load divided by the original cross sectional area (nominal stress) was in excess of the yield stress incurred plastic deformation of the entire sample gauge and associated barrelling.

Table 1 Chemical composition of leaded free cutting steel (LFCS). Element % Weight C 0.07 Si 0.01 Mn 1.00 P 0.05 S 0.30 Other(Pb) o 0.40

2. Experiments
Fig. 1. Test specimen dimensions (mm).

2.1. Material and specimens Cylindrical samples manufactured from as-cast FCS billets by TATA Steel were used for the experimental investigation. The samples were cut from the non-porous section of the billet (billet surface). The composition of the material is shown in Table 1. The specimens were machined into cylinders having 10 mm diameter and 111 mm length with threaded ends. The cylinders were then cut in half (see Fig. 1), producing two 55.5 mm long cylindrical samples, threaded at only one end. The cut surfaces were ground to a smooth surface nish, with roughness no greater than 9.1 m, nishing with P2500 abrasive paper. 2.2. Bonding procedure The test procedure comprised compressing the two halves of each specimen, which were put back together such that their axes
Table 2 The load-time combinations tested for 1000 1C. Test no. 1 2 3 4 5 6 7 8 9 10 11 12 Load (kN) 1.0 1.0 1.0 1.0 3.0 3.0 3.0 3.0 5.0 5.0 5.0 5.0 Nominal pressure (MPa) 12.7 12.7 12.7 12.7 38.2 38.2 38.2 38.2 63.7 63.7 63.7 63.7 ( E 0.4sy) ( E 0.4sy) ( E 0.4sy) ( E 0.4sy) ( E 1.3sy) ( E 1.3sy) ( E 1.3sy) ( E 1.3sy) ( E 2.1sy) ( E 2.1sy) ( E 2.1sy) ( E 2.1sy) Time 2.5 min 4 min 5 min 10 min 20 s 30 s 1 min 2 min 10 s 20 s 30 s 1 min

S. Afshan et al. / Materials Science & Engineering A 586 (2013) 25 30

27

Table 2 provides a summary of the load and time combinations studied for 1000 1C. The nominal pressure at the bonding interface is also shown in Table 2 in terms of the yield stress of the material at 1000 1C and 0.01 s 1, sy 30 MPa, taken from [24]. A schematic diagram of the time, temperature and pressure variation is also illustrated in Fig. 2.

2.3. Tensile tests Tensile tests on the bonded specimens were subsequently carried out to measure the strength of the bonded region. The tensile test specimens were turned into dog bone shapes; the geometry and dimensions of the specimens are shown in Fig. 3. The tensile tests were performed at room temperature and a uniform strain rate of 0.1 s 1 was employed. A c-gauge was utilised to measure the change in diameter, hence calculate the diametrical measure of true strain. The ow curves obtained for bonds formed at 1000 1C under compressive loads of 1 kN, 3 kN and 5 kN are shown in Fig. 4. It is

Temperature (C) 1000 960

48 Pressurepreload Pressure Pressure (MPa)

68

98

98+t

Time (s)

Fig. 2. Schematic of the time, temperature and nominal pressure variation during the test (positive pressure is downwards).

Fig. 3. Uniaxial test piece geometry and dimensions (mm).

Fig. 5. Position of fracture in samples subjected to 1 kN load at 1000 1C for (a) 2.5 min, (b) 4 min, (c) 5 min, (d) 10 min.

700 600 Stress (MPa)

700 600 Stress (MPa) 500 400 300 200 100

0.1 Strain 0.2 0.3
20 s 30 s 1 min 2 min

500 400 300 200 100 0 0

2.5 min 4 min 5 min 10 min

0.0

0.1

0.2 Strain

0.3

0.4

600 500 Stress (MPa) 400 300 200 100 0 0 0.05

10 s 20 s 30 s 60 s

0.1 Strain

0.15

0.2

Fig. 4. Flow curves (true stress versus true strain) obtained for bonds formed at 1000 1C and (a) 1 kN, (b) 3 kN, (c) 5 kN.

28

S. Afshan et al. / Materials Science & Engineering A 586 (2013) 25 30

apparent from Fig. 4(a) that under a compressive load of 1 kN, holding times of 2.5 min and 4 min are not sufcient for a strong bond to form; the bonds broke under considerably lower stress and strain

values under these conditions. This is also evident from Fig. 5, where it can be observed that the samples held for 2.5 min and 4 min (Fig. 5 (a) and (b), respectively) fractured at the bond line. This is supported

Fig. 6. Cross-sections of bonded samples after fracture. Bonds were formed under 1 kN, 1000 1C and holding times of (a) 2.5 min, (b) 4.5 min, (c) 5 min, (d) 10 min.

S. Afshan et al. / Materials Science & Engineering A 586 (2013) 25 30

29

by the presence of at regions in the cross-sections of the bond interfaces of the 2.5 min and 4 min samples, as shown in Fig. 6. Although the 5 min sample fractured in the middle of the specimen (see Fig. 5(c)), the fracture surface shown in Fig. 6(c) clearly indicates that a strong bond was formed. The 10 min sample (see Fig. 5(d)) fractured away from the bond line, indicating that the bond line was stronger than the adjacent material; this is supported by the fracture surface shown in Fig. 6(d). The stress-strain response of the specimens with 2.5 min and 4 min holding times for the 1 kN load (see Fig. 4(a)) also shows that the bonds formed for these holding times fractured differently from those formed at 5 min and 10 min holding times. In contrast to holding times of 5 min and 10 min where the nal stage of fracture is represented by a sudden drop in stress, the 2.5 min and 4 min holding time curves exhibit a nal stage of fracture consisting of a gradual, almost linear, decrease in stress followed by an abrupt reduction to zero. The initial gradual stress drop in these cases is consistent with a fracture sequence consisting of initial separation of the weakly bonded part of the cross-section, followed by fracture of the strongly bonded part, as shown in Fig. 6. The failure strains for the 5 min and 10 min samples for the 1 kN load are also similar, at least to within the level of scatter expected from testing different samples under the same conditions (this is explored further in Fig. 7), giving further indication that holding times greater than approximately 5 min are sufcient to form a strong bond at this pressure and temperature. Fig. 4(b) shows that increasing the compressive load from 1 kN to 3 kN signicantly decreased the holding time required for a strong bond to form. For a load of 3 kN, a holding time of 20 s was sufcient to create a strong bond.

As shown in Fig. 4(c), increasing the load further to 5 kN did not signicantly alter the required holding time; however, comparing Fig. 4(b) and (c), a relative decrease in the peak strength can be observed for the higher load. The peak strengths of the bonded

700 600

Stress (MPa)

500 400 300 200 100 0 0.00 0.10 0.20 Strain 0.30 0.40 2.5 min 10 min 5 min

Fig. 7. Flow curves obtained from tensile testing of samples soaked for different times.

Table 3 Maximum strengths of bonded specimens. Load (nominal pressure) 1 kN ( E 0.4sy) 3 kN ( E 1.3sy) Holding time 5 min 10 min 20 s 30 s 1 min 2 min 20 s 30 s 1 min Max strength (MPa) 573.55 602.19 631.00 614.02 607.33 619.53 522.46 527.88 518.90

5 kN ( E 2.1sy)

Fig. 8. SEM images of the bond line. Bonds were formed under 3 kN load, 1000 1C and holding times of (a) 10 s, (b) 20 s and (c) 30 s.

30

S. Afshan et al. / Materials Science & Engineering A 586 (2013) 25 30

specimens are listed in Table 3, for the ranges of loads and holding times tested. Table 3 shows that for loads insufcient to cause yield, the peak strength of the material is lower than those having a nominal bond line pressure approximately equal to the yield stress. It is also evident that increasing the applied load to a value well beyond what is necessary to cause yield led to a decrease in the peak strength of the material. This decrease in peak strength is attributed to the signicant amount of barrelling that occurs in these tests under such large compressive loads, creating damage in the material as a result and therefore softening the material; it is predominantly an artefact of the experimental method in that further increase of the load may actually improve the strength of the bond line somewhat in an actual thermomechanical process (e.g. rolling) by causing further strain hardening, however this leads to little or no improvement in bonding time. In a practical rolling process, adding further strength to an already complete and strong bond is undesirable because it would require modications to the rolling stage design in order to achieve a greater area reduction, hence higher load. Furthermore, greater area reduction in a single pass may exacerbate other damage mechanisms such as edge cracking. Therefore, the conditions that lead to a complete, strong bond in the least time are optimal in this context. Although a strong bond formed for the 1 kN load and 5 min holding time, Fig. 4(a) indicates a higher failure strain for the bond formed under 10 min holding time. This can also be seen in Fig. 4(b), where for different holding times the material exhibits a range of failure strains with no discernible trend. One possible explanation is the difference in soaking time experienced by the samples. To investigate this, several contiguous cylindrical samples were machined from the same batch of billet surface material and tested using the same procedure as that shown in Fig. 2. The specimens were heated to 1000 1C and a compressive load of 1 kN was applied for holding times of 2.5 min, 5 min and 10 min. (Fig. 4 shows that failure strain variability for bonded samples decreased with increasing load, hence examining the effect of soaking time for the lowest load is sufcient.) The specimens were then machined into dog-bone shapes as in Fig. 3 and tested in tension to investigate the resulting stress-strain behaviour of each specimen. Fig. 7 compares the ow curves obtained from these tensile tests. The highest failure strain is associated with a holding time of 2.5 min, the second highest ductility is exhibited by the 10 min holding time and the 5 min holding time has the minimum ductility. A relationship between soaking time and ductility cannot be established, hence this can be ruled out as the cause of the variability in failure strain shown in Fig. 4. This implies that the range of failure strains shown in Fig. 7 is simply the result of the natural variability associated with testing different samples taken from the same batch of material. 2.4. SEM images of the bond In order to further assess the bonded interfaces, bonded samples were cut longitudinally and images were taken from the bonding zones using a Hitachi S3400N Scanning Electron Microscope (SEM). Images were taken of the joint interface for bonds formed at 1000 1C, under a compressive load of 3 kN. Fig. 8(a)(c) shows the bonds formed at this condition for holding times of 10 s, 20 s and 30 s, respectively. As shown previously for the tensile testing of the 5 kN bonded samples, for a holding time of 10 s the bonds were broken under considerably lower stress and strain values, indicating an incomplete bond (the same must be true for any smaller compressive load). The crack line appearing at the bond interface in Fig. 8(a) is consistent with this nding. Increasing the holding time to 20 s resulted in better healing of the bond line; as shown in Fig. 8(b), the crack was reduced to predominantly a series of separated voids. Increasing the holding time to 30 s (Fig. 8(c)), no trace remained of the bond line conrming a fully bonded sample.

3. Conclusions

 A new experimental method was developed based on applied

loading under direct resistance heating to identify conditions (temperature, loading, time) required for complete diffusion bonding of free cutting steel. Different combinations of load and time were tested for 1000 1C. Tensile testing to failure of the bonded specimens was carried out to measure the strength of the bonded region and assess macroscopically whether full bonding occurred. The position of fracture on the specimen, the nature of the fracture surface and SEM images of the bond line were also used to assess the extent of bonding and corroborate the tensile testing results. Conditions leading to complete bonding of free cutting steel at 1000 1C were 1 kN for 5 min, 3 kN for 20 s and 5 kN for 20 s; the loads corresponded to nominal bond line pressures of 12.7 MPa, 38.2 MPa and 63.7 MPa, respectively. Variability in failure strain for bonded samples can be attributed to natural specimen material variability, not an effect of soaking time on the material response. Maximum bond line strength was achieved in the experiments for an applied load of 3 kN, which gave a nominal bond line pressure approximately equal to the yield stress at 1000 1C. This is attributed to lower loads being insufcient for strain hardening to occur, and larger loads causing signicant barrelling and associated material damage in these experiments. Although somewhat greater bond line strength could be achieved by increasing the load further beyond the yield point in an actual process (e.g. rolling), the experiments suggest there would be no associated further improvement in bonding time. Hence the load that produces a bond line pressure approximately equal to the yield stress is optimal in the context of a rolling application, as it leads to a complete, strong bond in the least time with the least modication to the sequence of area reduction in a multi-stage rolling process.

 

References
[1] B. Thomas, in: O. Yu (Ed.), Modeling for Casting and Solidication Processing, Marcel Dekker, New York, 2001, pp. 499540. [2] R. Monroe, AFS Trans. 113 (2005) 519546. [3] G.J. Watts, D. Farrugia, P.M. Ingham, Proceedings of the First International Conference on Component Optimisation, University of Wales, Swansea, 1999. [4] B.J. Blaiszik, S.L.B. Kramer, S.C. Olugebefola, J.S. Moore, N.R. Sottos, S.R. White, Ann. Rev. Mater. 40 (2010) 179211. [5] R.P. Wool, K.M. Oconnor, J. Appl. Phys. 52 (10) (1981) 59535963. [6] P. Fratzl, R. Weinkamer, S. Zwaag, Self Healing of Materials, Springer, Netherlands (2008) 323335. [7] F. Vermolen, M. Rossum, E. Perez, J. Adam, S. Zwaag, Self Healing Materials, Springer, Netherlands (2008) 337363. [8] V. Li, E. Yang, S. Zwaag, Self Healing Materials, Springer, Netherlands (2008) 161193. [9] H.W. Reinhardt, M. Jooss, Cem. Concr. Res. 33 (7) (2003) 981985. [10] R.N. Lumley, I.J. Polmear, Proceedings of the First International Conference on Self Healing Materials, 1820 April 2007, Noordwijk aan Zee, The Netherlands. [11] H. Wang, P. Huang, Z. Li, Self Healing Materials, Springer, Netherlands (2008) 255277. [12] D. Wei, J. Han, K. Tieu, Z. Jiang, Scr. Mater. 51 (6) (2004) 583587. [13] B. Derby, E.R. Wallach, J. Mater. Sci. 19 (10) (1984) 31403148. [14] Z.X. Guo, N. Ridely, Mater. Sci. Technol. 3 (11) (1987) 945953. [15] Y. Huang, N. Ridley, F.J. Humphreys, Mater. Sci. Eng. A 266 (1-2) (1999) 295302. [16] M.T. Salehi, J. Pilling, N. Rideley, D.L. Hamilton, Mater. Sci. Eng. A 150 (1) (1992) 16. [17] H. Somekawaa, H. Hosokawab, H. Watanabec, K. Higashia, Mater. Sci. Eng. A 339 (1-2) (2003) 328333. [18] J. Han, G. Zhao, Q. Cao, Sci. China Ser. E: Technol. Sci. 40 (2) (1997) 164169. [19] D. Wei, J. Han, Z.Y. Jiang, C. Lu, A.K. Tieu, J. Mater. Process. Technol. 177 (1-3) (2006) 233237. [20] H. Keife, U. Stahlberg, J. Mech. Work. Technol. 4 (2) (1980) 133143. [21] A. Wang, P.F. Thomson, P.D. Hodgson, J. Mater. Process. Technol. 60 (1996) 1996. [22] S. Afshan, D. Balint, J. Lin, D. Farrugia, J. Multiscale Modell. 3 (1 and 2) (2011) 7990. [23] J. Pilling, Mater. Sci. Eng. 100 (1987) 137144. [24] D. Farrugia, Leaded Steel, Tata Steel UK, 2010.