Microstructure evolution of innovative thermal bridge ...

Materials and Design 137 (2018) 68–78

Contents lists available at ScienceDirect

Materials and Design journal homepage: www.elsevier.com/locate/matdes

Microstructure evolution of innovative thermal bridge composite (i-TBC) for power electronics during elaboration Hiba Fekiri, Vladimir A. Esin ⁎, Vincent Maurel, Alain Köster, Yves Bienvenu MINES ParisTech, PSL Research University, Centre des Matériaux (CNRS UMR 7633), BP 87, Evry cedex 91003, France

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• A new substrate, innovative Thermal Bridge Composite (i-TBC), for high temperature electronic systems was developed. • i-TBC was characterized through all elaborations steps: first cold rolling, heat treatment and second cold rolling. • Ultra-fine and coarse grained areas can be formed during the cold rolling due to heterogeneous plastic deformation. • Interface adherence and strain hardening were found to be dependent on rolling direction.

a r t i c l e

i n f o

Article history: Received 23 June 2017 Received in revised form 30 September 2017 Accepted 4 October 2017 Available online 05 October 2017 Keywords: i-TBC Cold rolling Thermal bridge Strain hardening Cold welding EBSD

a b s t r a c t To improve the reliability of the power electronic modules for the high temperature applications, an innovative Thermal Bridge Composite (i-TBC) was designed. It has the architectured structure consisting of perforated FeNi36 sheet inserted between two Cu sheets. Due to simultaneous use of Cu and FeNi36, i-TBC possesses both a good thermal transverse conductivity and a limited longitudinal coefficient of thermal expansion. Different characterisations of i-TBC are required to understand the formation of its microstructure leading to the final properties. Therefore, the aim of this study was to analyse the integrity of Cu-Cu and Cu-FeNi36 interfaces as well as copper microstructure evolution throughout all elaboration steps: (i) first cold rolling, (ii) heat treatment and (iii) second cold rolling. First cold rolling did not lead to a bonding of Cu-Cu interfaces in the thermal bridge. Moreover, heterogeneity of Cu grain microstructure was observed with formation of ultra-fine grained structure close to junctions of Cu and FeNi36. The heat treatment led to a degradation of different interfaces adherence and to a complete copper recrystallization. Finally, the second cold rolling ensured an efficient solid welding of Cu-Cu interfaces and led to a heterogeneity of strain hardening of copper. © 2017 Elsevier Ltd. All rights reserved.

1. Introduction A new generation of semiconductors (GaN, SiC) with a wide band gap designed for high power (N50 kW for a 0.5 cm2 die) and high frequency (N 10 MHz) applications makes possible to use the power ⁎ Corresponding author. E-mail address: [email protected] (V.A. Esin).

https://doi.org/10.1016/j.matdes.2017.10.009 0264-1275/© 2017 Elsevier Ltd. All rights reserved.

electronic systems (modules) in harsh environments with large temperature variations (for example, from −55 to 200 °C in aircraft landing gear electronics) [1–3]. Such a temperature variation leads to a generation of thermo-mechanical stresses due to the coefficient of thermal expansion (CTE) mismatch between different materials found in power electronic modules (silicon and GaN, solders, substrate (copper and/or ceramic) and aluminium based radiator-convector) and, thus, conditions module lifetime. In addition, substrate materials based on pure

H. Fekiri et al. / Materials and Design 137 (2018) 68–78

69

Fig. 1. Architectured structure of the i-TBC: perforated FeNi36 sheet situated between two U-shaped Cu sheets for Invar positioning.

copper are no longer compatible with the die and solder materials for such temperature cycles amplitudes and to resist these severe service conditions, the structure of power electronic modules must be redesigned, in order to minimize the CTE mismatch between successive layers. Recently, the innovative Thermal Bridge Composite (i-TBC1) was proposed as a substrate material for power electronic systems [4]. iTBC is an architectured material based on Cu and Invar™2 FeNi36. Due to the simultaneous use of Cu and FeNi36, i-TBC possesses both a good thermal transverse conductivity and a limited longitudinal CTE. Therefore, it can provide an efficient solution for electronic packaging issues for the high temperature applications (up to 250 °C). The architectured structure of the i-TBC is schematized in Fig. 1. The central part is reinforced by perforated FeNi36 and devoted to the positioning of active devices (semiconductors GaN, SiC); the Cu-Cu part serves for external connections and cooling through power cables. It is worth noting that both Cu sheets have initially U-shaped geometry (with thickness of 1.79/1.37/1.79 mm) to be able to place FeNi36 (initial thickness of 1.25 mm). The formation of Cu-FeNi36 interfaces ensures a reasonably low CTE of the assembly, of the order of 10 · 10−6 K−1 [4], intermediate between that of Si or Si-GaN (about 4 · 10−6 K−1) and that of Cu (about 18 · 10−6 K−1) or Al alloys (N 20 · 10−6 K−1). The main innovation of the i-TBC consists in the formation of a bond area between the Cu sheets through the perforations in FeNi36 sheet. These bond areas, called thermal bridges (Fig. 2), provide a good transverse thermal conductivity to the i-TBC from 50 W·m− 1 K− 1 to 250 W·m−1 K−1 [4]. The i-TBC, in terms of CTE and thermal conductivity is quite well positioned against existing alternative substrate materials although in the future composites based on copper with carbon nanotubes might be better positioned. A second originality of the i-TBC is that it can be produced as explained below, continuously by “cold” (actually up to 120 °C due to the heat generation associated with plastic deformations) rollbonding which is a rather low cost process. To fill the perforations of the FeNi36 sheet, we have previously also considered different options like “cold spraying” of Cu powder, leading to a better Cu\\Cu bond quality in the thermal bridges at the expense of a fabrication cost increase linked to the use of Cu powders and of a lower productivity than roll bonding. Hot roll bonding was not considered because of the detrimental effect of Cu oxide layers on the thermal conductivity of the bridges. The complex architecture of the i-TBC is obtained by a succession of different fabrication steps. During each step, modifications occur in terms of the bonds quality between Cu and FeNi36 and of the microstructure. These different elaboration steps define the final i-TBC performances. Therefore, the objective of the present study is to investigate the modifications associated with each elaboration step to be able to identify the key parameters playing an important role in process leading to the final i-TBC properties.

1 2

i-TBC is a registered trademark of TG-GRISET. Invar™ is a registered trademark of Aperam alloys Imphy.

Fig. 2. Formation of thermal bridges by plastic flow of Cu in FeNi36 perforations: (a) interrupted first cold rolling; (b) top view after the second cold rolling and after polishing showing the final geometry of the thermal bridge as well as the cross-section plan for different characterizations and (c) different areas of the same thermal bridge used for characterizations carried out in the present study (see Part 3).

2. Experimental procedure 2.1. Elaboration steps of the i-TBC A first level of optimization of i-TBC architecture was reached in a previous study [5]. Combination of 20% of surface area of thermal bridges in the FeNi36 internal layer and relative final thicknesses for two Cu and one FeNi36 sheets of 33/33/33 (%) were chosen. The final i-TBC geometry is obtained through different steps including two cold rolling that should lead to a complete filling of the FeNi36 perforations by Cu and ensure a good adherence of Cu\\Cu and Cu-FeNi36 interfaces. The quality of the final i-TBC structure obtained after all elaboration steps will be further investigated and it is one of the main objectives of the present work. It is worth noting that Cu was alloyed by small additions of phosphorous and iron to form iron phosphides that should limit the Cu grain size. Before the first cold rolling step, Cu and FeNi36 sheets are subjected to a chemical degreasing by acetone followed by wire-brushing to remove surface contamination like oxide films or dust. Further, the elaboration of the i-TBC is carried out in three steps: 1. First cold rolling step results in about 60% of thickness reduction of initial two Cu sheets and perforated FeNi36. During this step the thermal bridges are being formed by plastic flow of Cu in FeNi36 perforations (Fig. 2a); 2. Heat treatment at 450 °C for 6 h in an inert atmosphere aims to improve the ductility of i-TBC required for the second cold rolling; 3. Second cold rolling step leads to a thickness reduction of about 20% to achieve final i-TBC thickness of 1.2 mm and final geometry of thermal bridge (Fig. 2b). All elaboration steps were carried out using industrial facilities at TG-GRISET on an existing rolling mill with 400 mm wide cylinders (while the laminates do not exceed 140 mm in width). The dimensional accuracy of the roll bonded composite can be maintained within 10 μm for the thickness (typically 1 to 1.5 mm) with short (b400 mm) and thick (N400 mm) rolls for a duo rolling mill and better accuracies are expected using a quarto. There are no serious concerns of end users with overall width (between 90 and 150 mm) as the edges can be trimmed before cutting the bands in individual substrates. Internal dimensions of both the deformed and perforated FeNi36 and of the Cu bands cannot be guaranteed with the same level of accuracy as the external dimensions of the composite band. It is worth noting, that the cold rolling steps are performed without lubrication and with a rolling rate of 10 m min−1.

70


The main issues for the i-TBC elaboration is, first, to ensure good contact between Cu-Cu and Cu-FeNi36 interfaces to guarantee the reliable final properties (coefficient of thermal expansion and thermal conductivity), and, second, to have homogeneous Cu grain structure (at least at the centre of the thermal bridge) to guarantee homogeneous and stable properties of i-TBC. The state of the interfaces as well as Cu grain structure is investigated using different methods as described further. 2.2. Microstructure characterization The objective of the present study was an accurate characterization of interfaces and microstructure of the i-TBC after each elaboration step. Combined observations using optical microscopy (OM) and scanning electron microscopy (SEM) with electron backscatter diffraction (EBSD) feature were carried out. All observations were performed on the sample cross-sections corresponding to longitudinal plane (i.e. ND-RD (Fig. 2a and c)). After each elaboration step, the samples were cut in the middle of the thermal bridge parallel to RD (Fig. 2b) using a low speed diamond wire saw, in order to limit interface damage during cross section preparation. Then, the samples were mechanically polished for the observations in OM using SiC abrasive with a sequence of grit size from P1200 to P4000 and polishing time around 5 min for each grit size. The final polishing was carried out using diamond abrasive with a particle size of 1 μm; the polishing was finished when any scratch was not visible in OM at the magnification of 1000. For EBSD analysis, additional electrolytical polishing was done at room temperature for 5 s applying the voltage ranging from 8 to 10 V and using an electrolyte based mainly on phosphoric acid. It is worth noting, that the conditions for the electrolytical polishing were optimized to get a reliable surface quality for Cu part of the i-TBC, since FeNi36 part was highly strain hardened and did not allowed EBSD analysis. The EBSD data acquisition was performed using SEM S-FEG FEI Nova NanoSEM 450 operating at acceleration voltage of 20 kV, the step and map size were optimized, taking into account the observed microstructure heterogeneities. In order to investigate the influence of elaboration on the microstructure for different parts of i-TBC, the EBSD maps were acquired systematically at the centre of the thermal bridge (Cu-Cu interfaces, zone 2 in Fig. 2c) and close to Cu-FeNi36 junctions (called hereafter “triple junctions”, zones 1 and 3 in Fig. 2c). 2.3. Analysis of strain hardening During the cold rolling, the Cu and FeNi36 are strain hardened due to intensive plastic deformation. The strain hardening can be analysed using EBSD data applying different criteria [6]. In the present study

two different local misorientation criteria were used to reveal the strain distribution: Grain Orientation Spread (GOS) and Kernel Average Misorientation (KAM). GOS represents the average misorientation for all pixels of the same grain relative to the average orientation of the grain [6]. A low average misorientation angle (b 2°) was used as the criterion of a recrystallized state. The GOS maps were obtained from EBSD data by applying a misorientation of 5° and higher between two different pixels to be considered to be in two different grains and a minimum grain size (in terms of constituting pixels) of 9 pixels. KAM consists in the calculation of the average misorientation angle of a given pixel with all its neighbours located at a given fixed distance [7,8]. To avoid the misorientation due to the sub-grain boundaries, only second nearest neighbours with a misorientation angle lower than 6° have been taken into account. Moreover, such a misorientation angle of 6° led to results consistent with those obtained using GOS maps. Finally, the grain boundaries (GBs) were analysed in the frame of Coincidence Site Lattice (CSL) model. It was used as an indicator of the recrystallization in copper, since the formation of twin boundaries was widely observed during the recrystallization in fcc metals with low to medium stacking fault energy [9]. Moreover, the CSL model was extensively used in the evaluation and characterization of copper recrystallization [10,11]. For example, it was shown that the fraction of Ʃ3 GBs for fully-recrystallized copper was above 0.2. 3. Results It is worth noting that only Cu microstructure could be analysed by EBSD, since it was not possible to get reliable diffraction signal from FeNi36 due to its increased strain hardening even after the heat treatment at 450 °C. Therefore, only the Cu microstructure evolution in different parts of the thermal bridge during different elaboration steps will be discussed further together with interface adherence. 3.1. Initial state of Cu To be able to follow Cu microstructure evolution during the i-TBC elaboration, first, the initial Cu sheets were characterized by EBSD. Special attention was paid to grain structure as well as initial strain hardening. Inverse pole figure (IPF), GOS and KAM maps are presented in Fig. 3. It was observed that initial Cu had a microstructure with elongated grains characteristic for metals after rolling (Fig. 3a). Moreover, GOS and KAM maps show almost fully deformed structure (Fig. 3b and c) with some small non-deformed areas due, probably, to the beginning of recrystallization (blue areas in GOS map (Fig. 3b)).

Fig. 3. Microstructure of Cu in initial state: (a) inverse pole figure, (b) GOS and (c) KAM maps. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).


Fig. 4. Thermal bridge after the first cold rolling: poor adherence of Cu-Cu interface can be seen.

3.2. First cold rolling After the first cold rolling step, the observations using OM showed a poor adherence of Cu-Cu interface throughout the thermal bridge (Fig. 4). Two different Cu sheets were certainly in contact during the rolling taking into account their morphology in the thermal bridge close to the interface and the contact was lost once the rolling stress had been suppressed.

71

Cu grain structure after the first cold rolling was studied by EBSD. IPF are shown in Fig. 5 for three different regions of thermal bridge: two regions are located close to triple junctions between two Cu sheets and FeNi36 sheet (Fig. 5a and c); the region of Fig. 5b corresponds to the centre of the thermal bridge. It is worth noting, that the region shown in Fig. 5c was subjected to deformation prior those represented in Fig. 5b and a, according to the rolling direction (RD). Different non-indexed areas (represented by black pixels) can be seen in Fig. 5. These areas corresponded mostly to the zones of poor interface adherence (for example, in Fig. 5b). However, few non-indexed regions can be observed as well in Cu sheets (Fig. 5c) due to the difficulties in sample preparation for EBSD analysis. Meanwhile, it is not disturbing the interpretation of observed results. For the sake of simplicity, the FeNi36 zones were highlighted. EBSD data in Fig. 5 confirmed the observations made using OM in Fig. 4 that Cu-Cu interface had a poor adherence throughout the thermal bridge. Moreover, one can notice a very heterogeneous Cu microstructure along the thermal bridge. On one hand, the structure of Cu grains exhibits a basic rolling microstructure with elongated grains parallel to RD at the centre of the thermal bridge (Fig. 5b) as well as for the upper Cu sheet in the regions close to triple junctions (Fig. 5a and c).

Fig. 5. Inverse pole figure maps for Cu in the thermal bridge after the first cold rolling. Analysis was carried out in different zones, according to rolling direction as schematized above: (a) left triple junction of Cu and FeNi36 interfaces, (b) centre of the thermal bridge and (c) right triple junction of Cu and FeNi36 interfaces. Highlighted areas are enlarged in Fig. 6. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

Fig. 6. Enlarged inverse pole figure maps for Cu in the thermal bridge after the first cold rolling for different areas highlighted in Fig. 5: (a) left triple junction of Cu and FeNi36 interfaces, (b) centre of the thermal bridge and (c) right triple junction of Cu and FeNi36 interfaces. The heterogeneity of Cu grain structure can be seen with the presence of UFG. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

72


On the other hand, lower Cu sheet at triple junction with FeNi36 has a heterogeneous microstructure with combination of elongated and very fine grains in the vicinity of the Cu-FeNi36 interfaces (Fig. 5a and c). The areas containing fine grains are highlighted in Fig. 5a and c (yellow squares) and enlarged in Fig. 6a and c; an enlarged area of the centre of the thermal bridge is given as well in Fig. 6b for comparison. It can be seen that the first cold rolling leads to the appearance of grains with diameter significantly lower than 1 μm, called usually ultra-fined grains (UFG) [12,13], close to Cu-FeNi36 interfaces (Fig. 6a and c). At the same time, the centre of the thermal bridge represents a standard microstructure formed during cold rolling. Such a heterogeneous grain structure in the same thermal bridge emphasizes the complexity of strain and stress state during the first cold rolling. It is worth noting that intensive plastic deformation coupled with important strain gradient can lead to the crack formation in the material. Nevertheless, the microstructure investigation using optical microscope as well as scanning electron microscope did not reveal any crack in the i-TBC. Further, the local misorientation of Cu in regions shown in Fig. 6 was analysed using GOS and KAM criteria (Fig. 7). The UFG observed close to triple junctions of Cu and FeNi36 interfaces do not exhibit strong misorientation (Fig. 7a, c, d and f) while higher misorientation has been evidenced for the centre of the thermal bridge (Fig. 7b and e). 3.3. Heat treatment The heat treatment after the first cold rolling step was carried out at 450 °C during 6 h to reduce the strain hardening of Cu and to increase thus its ductility and to reinforce welding of Cu-Cu interfaces. However, as it was shown in the previous section, Cu-Cu interfaces had poor adherence in thermal bridges after the first cold rolling and therefore a subsequent heat treatment was not able to improve its adherence due to a poor contact. Moreover, an adherence degradation of Cu-Cu interfaces in thermal bridge was observed using OM (Fig. 8). The latter was more pronounced in the triple junction area rolled last. On the other hand, it was shown as well that Cu had a heterogeneous grain structure in the same thermal bridge before the heat treatment with UFG observed close to Cu-FeNi36 interfaces (Fig. 6) and a difference in Cu grain structure was stated between upper and lower copper sheets. The influence of the heat treatment on copper grain structure was analysed by EBSD. IPF maps for copper for different regions of the

Fig. 8. Thermal bridge after the heat treatment at 450 °C for 6 h: degradation of Cu-Cu interface adherence can be seen as compared with that after the first cold rolling (Fig. 4).

same thermal bridge are shown in Fig. 9. A mainly equiaxed grain microstructure is observed throughout the thermal bridge suggesting a complete recrystallization of the microstructure obtained during the first cold rolling step. A slight difference in Cu grain size is observed between upper and lower Cu sheets close to Cu-FeNi36 interfaces. It should be related to different strain gradient induced by the first cold rolling resulting in different driving force for the recrystallization. GOS and KAM maps presented in Fig. 10 confirm the copper recrystallization, since most regions have a GOS value b 2°. Moreover, the copper microstructure exhibits an important fraction of the Ʃ3 twin type GBs and the appearance of the second and third twin generations (Ʃ9, Ʃ27) is observed indicating the recrystallization process (Table 1). Any difference in strain hardening is not observed between upper and lower Cu sheets after the heat treatment (Fig. 10). 3.4. Second cold rolling The aim of the second cold rolling step was to enhance the bonding strength of the interfaces, particularly that of Cu\\Cu in the thermal bridge, and to achieve a suitable thickness of the i-TBC and a hardness of the Cu sheet conform to expectations of end users in the electronic industry. Indeed, the observations using OM showed a significant improvement of the adherence of Cu-Cu interface in the centre of thermal bridge (Fig. 11). However, a poor Cu-Cu and Cu-FeNi36 interfaces adherence was still observed close to triple junctions of Cu and FeNi36. Since the second cold rolling is the final step of the i-TBC elaboration, a detailed examination of the microstructure in the centre of the

Fig. 7. GOS and KAM maps for Cu in the thermal bridge after the first cold rolling in areas shown in Fig. 6: (a)–(c) GOS maps, (d)–(f) KAM maps. Highlighted areas are enlarged in Fig. 15. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).


73

Fig. 9. Inverse pole figure maps for Cu in the thermal bridge after the heat treatment at 450 °C for 6 h. Analysis was carried out in different zones, according to rolling direction as schematized above: (a) left triple junction of Cu and FeNi36 interfaces, (b) centre of the thermal bridge and (c) right triple junction of Cu and FeNi36 interfaces. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

thermal bridge was carried out using four stitched together EBSD maps covering a large zone (Fig. 12a). A continuous Cu microstructure can be observed across the Cu-Cu interface indicating a complete welding during the second cold rolling (Fig. 12a). Enlarged IPF maps also show the presence of fine grains close to Cu\\Cu interface (Fig. 12b and c). It can be seen that the surface fraction of these fine grains is variable along the Cu-Cu interface with areas having increased and decreased number of fine grains (compare Fig. 12b and c). The level of strain hardening in regions shown in Fig. 12b and c is less than that after the first cold rolling step (compare GOS maps in Fig. 13a and b and those in Fig. 7b; and KAM maps in Fig. 13c and d and those in Fig. 7e), since a lower strain was applied during the second cold rolling step. Moreover, the fine grains seem to be little strained (Fig. 13). Besides, the EBSD analysis close to triple junctions of Cu and FeNi36 confirmed a poor welding of Cu-Cu interfaces (Fig. 14a and b). GOS and KAM maps show a significant difference in the local misorientation for the two opposite triple junctions of Cu and FeNi36 interfaces in the same thermal bridge (Fig. 14c–f) This difference is more pronounced in KAM maps (Fig. 14e and f). These analyses suggest that the strain hardening seems to be a function of the rolling direction: the triple junction of Cu and FeNi36 submitted lastly to rolling is less strain-hardened than the triple junction of Cu and FeNi36 deformed firstly. Upper and lower Cu sheets seem to have the same grain structure (Fig. 14a and b). However, a slight difference in strain hardening is observed: upper Cu sheet represents a higher misorientation gradient along the rolling direction than the lower one. Such a difference between upper and lower Cu sheets observed for grain structure after the first cold rolling as well as for strain hardening after the second cold rolling should be related to the elaboration parameters (geometry of rolling mill and rolling rate). 4. Discussion The observation of thermal bridges of the i-TBC carried out after each of the three different elaborations steps showed a difficulty in obtaining a perfect filling of FeNi36 perforations by plastic flow of Cu. Except the

state after the heat treatment, Cu-FeNi36 interfaces were always well welded while several not welded regions of Cu-Cu interfaces were still observed. The junction of Cu-FeNi36 interfaces should be facilitated by a direct contact of Cu and FeNi36 sheets from the beginning of the elaboration process. Such a direct contact coupled with applied rolling stress results in a good welding usually referred as cold roll bonding. On the other hand, the situation is more complex for the welding of Cu-Cu interfaces in the thermal bridges. Indeed, the Cu sheets are not in contact before the elaboration and the intensive plastic flow of Cu through FeNi36 perforations is required for Cu-Cu bonding. Such a Cu plastic flow can be limited due to insufficient applied rolling stress or to air trapped during the rolling process between Cu sheets as it will be discussed further. EBSD analysis of copper after the first cold rolling revealed a heterogeneous microstructure with elongated grains in the centre of the thermal bridge for both upper and lower Cu sheets as well as for upper Cu sheet close to the triple junctions of Cu and FeNi36. UFG microstructure was found for lower Cu sheet in the vicinity of the triple junctions of Cu and FeNi36 (Figs. 5 and 6). Such a fine-grained equiaxed microstructure was widely observed in the literature after severe plastic deformation (SPD) or large strain rolling (up to 90% in thickness reduction) leading to the formation of grains with size ranging from 10 nm to 2 μm [14]. Therefore, the formation of UFG microstructure during the i-TBC elaboration observed in this work indicated a high strain level during the first cold rolling close to triple junctions of Cu and FeNi36. The mechanisms leading to the formation of UFG under plastic deformation have been investigated in different studies [12,14,15]. It was concluded that the final microstructure depended on the SPD procedure and on the material nature itself. It is generally admitted that the formation of UFG structure can be divided into several stages. First, the plastic deformation generates a high density of dislocations that progressively reorganize into a substructure of cell blocks (sub-grains) leading thus to an important stored energy. Further, the stored energy might be reduced by a dynamic recrystallization. It is well-known that there exist two types of dynamic recrystallization: discontinuous dynamic recrystallization (dDR) and continuous dynamic recrystallization (cDR) [16].

74


Fig. 10. GOS and KAM maps for Cu in the thermal bridge after the heat treatment at 450 °C during 6 h in areas corresponding to Fig. 9: (a)–(c) GOS maps, (d)–(f) KAM maps. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

dDR is characterized by the migration of high-angle GBs through the deformed microstructure, resulting in the formation of free strain regions. cDR occurs via a rotation of sub-grains (low-angle GBs) leading to the formation of high-angle GBs. The analysis of GB character coupled with GOS maps for UFG microstructure observed after the first cold rolling step (Fig. 15) suggests that Cu close to FeNi36 undergoes the cDR [17]. The strained grains (with average misorientation higher than 2°) contain mostly low-angle GBs (in grey) forming cells (with size corresponding to non-deformed grains) while the recrystallized

grains (with average misorientation lower than 2°) do not contain low-angle GBs and are separated by high-angle GBs. It is worth noting that in the present study the formation of UFG structure occurs during the first cold rolling with the macrostrain level significantly lower (thickness reduction of 60%) than that applied in the SPD processes. Therefore, the formation of UFG microstructure close to triple junctions of Cu and FeNi36 (Figs. 5 and 6) and elongated

Table 1 Number fraction of CSL GBs for Cu in the thermal bridge after the heat treatment at 450 °C during 6 h in regions shown in Fig. 9. Regions

Left triple junction (Fig. 9a) Centre (Fig. 9b) Right triple junction (Fig. 9c)

Number fraction of CSL GBs Ʃ3

Ʃ9

Ʃ27a

Ʃ27b

0.29 0.55 0.17

0.045 0.063 0.029

0.008 0.011 0.005

0.010 0.014 0.006

Fig. 11. Thermal bridge after the second cold rolling: improvement of the Cu-Cu interfaces adherence in the centre of thermal bridge can be seen, as compared with that after the first cold rolling and heat treatment (Figs. 4 and 8); a poor Cu-Cu and Cu-FeNi36 interfaces adherence is still observed close to the triple junctions of Cu and FeNi36.


75

Fig. 12. Inverse pole figure maps for Cu in the centre of the thermal bridge after the second cold rolling: (a) four IPF stitched together to show a continuous Cu microstructure across Cu-Cu interface, (b) enlarged zone showing the presence of fine grains close to Cu-Cu interface and (c) enlarged zone emphasizing continuous Cu microstructure. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

grains in the centre of the thermal bridges (Fig. 7b) suggest an important heterogeneity in strain throughout the thermal bridge during the first cold rolling step and a very high local strain value due to the filling of FeNi36 perforations by Cu. After the heat treatment at 450 °C during 6 h, the observation of thermal bridges showed a degradation of Cu-Cu interfaces adherence (compare Figs. 4 and 8). Different explanations can be proposed. First, the plastic flow of Cu in FeNi36 perforations could lead to air trapping (since the cold rolling is carried out in air) in the thermal bridge. The air should be blocked mostly in the second triple junction of Cu and FeNi36, according to the rolling direction. This air trapped during the first cold rolling should have an important internal pressure corresponding to the compressive stress under the cylinders. This applied

stress could be estimated in Ref. [6] by a static planar compression test and was found to be about 550 MPa. Further, an increase in temperature during the heat treatment leads to increase in trapped air pressure (Boyles's law) and can lead to the decoupling of Cu-Cu interfaces if the bonding strength induced by the first cold rolling was not sufficiently strong. The second explanation for degradation of Cu-Cu interfaces adherence during the heat treatment bases on the large CTE mismatch between Cu and FeNi36. The decoupling of the interfaces should occur during the cooling to ambient temperature. The heating of the i-TBC to 450 °C leads to a complete relaxation of the residual stresses induced by the first cold rolling since FeNi36 has a Curie temperature of 280 °C and above this temperature the CTE of FeNi36 approaches that of Cu.

Fig. 13. GOS and KAM maps for Cu in the centre of the thermal bridge after the second cold rolling for the areas corresponding to Fig. 12b and c: (a)–(b) GOS maps, (c)–(d) KAM maps. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

76


Fig. 14. EBSD maps for Cu for opposite junctions of Cu and FeNi36 in the same thermal bridge after the second cold rolling: (a)–(b) IPF maps, (c)–(d) GOS maps showing a heterogeneous strain hardening as a function of position in the thermal bridge along the rolling direction; (e)–(f) KAM maps showing a heterogeneous strain hardening as a function of rolling direction. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

During further cooling, the thermal deformation of the Cu is hindered by the FeNi36 in the temperature range from 280 °C to ambient temperature generating thus residual stresses. In order to assess the stress distribution during the cooling, a finite element simulation was carried out on a representative volume of a thermal bridge. The results revealed that the level of the stress in the direction perpendicular to the Cu-Cu interface (commonly referred to tensile stresses) close to Cu-FeNi36 interfaces can reach the value of 100 MPa. Such a stress generation during the cooling from the temperature of the heat treatment should lead to the decoupling of Cu-Cu interfaces. In addition, the heat treatment leads to a complete recrystallization of Cu in the thermal bridge, as shown by EBSD (Figs. 9 and 10); the fraction of the Ʃ3 GBs suggests as well a fully recrystallized state. The second cold rolling step (with a thickness reduction of about 20%) enhances the filling of FeNi36 perforations and usually allows bonding of Cu-Cu interfaces in the centre of the thermal bridge. The observations using OM reveal a good adherence in the centre of the thermal bridge (Fig. 11) and a detailed analysis by EBSD confirms a continuous Cu microstructure through the interface suggesting a good

welding of Cu-Cu interfaces (Fig. 12). The cold roll welding mechanism has been described in numerous studies [18–27]. It was reported that the roll bonding of metals is affected by various factors that can be classified in two groups: rolling parameters (thickness reduction, rolling speed, friction coefficient) and initial surface state. Different studies emphasize the crucial role of the surface state and work hardened layers on the achievement of cold welding during the rolling [28–30]. For the iTBC elaboration a scratch brushing was used before the first cold rolling step to ensure the appropriate initial surface quality of Cu and FeNi36 sheets. The microstructure of the surface layer resulting from the scratch brushing was studied in Refs. [20,31]. It was reported that wire brushing resulted in near surface severe plastic deformation (NS-SPD), which induced intense strain to the surface layer leading to the nanocrystalline structure formation. However, it is difficult to assign the welding of the Cu-Cu interfaces after the second cold rolling to the conventional cold-rolling model (fragmentation of the work hardened layers and extrusion of the virgin metal) because the work hardened layers formed during the scratch brushing undergo a first rolling (which does not give rise to any welding of the copper


77

Fig. 15. Enlarged GOS maps highlighted in Fig. 7c superposed with low-angle (grey) and high-angle (black) GBs after the first cold rolling: (a) strain hardened region (strained grains (with average misorientation higher than 2° highlighted in red, green and yellow) contain mostly low-angle GBs forming a substructure) and (b) regions after continuous dynamic recrystallization (recrystallized grains (in blue) a separated by high angle GBs). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

interfaces) and an intermediate heat treatment leading to the copper recrystallization. The cold welding may be mainly due to a high rolling rate (10 m min−1). This high rolling rate should lead to a significant increase in local temperature at the interfaces. The temperature increase combined with the surface roughness can lead to the welding of the interfaces, according to the scheme proposed in Fig. 16. At the beginning of the second cold rolling several areas of Cu sheets are in contact (necking areas) whereas the others are not due to the surface roughness (Fig. 16a). Therefore, the deformation of surface layers of Cu sheets would be heterogeneous: the necking areas deform first. Then high rolling deformation rate leads to high local stress and relative movement of surfaces of Cu sheets and subsequent friction, thus to an increase in temperature facilitating the atomic diffusion and therefore the rearrangements of atoms with formation of metallic bonds resulting thus in a continuous microstructure through the interface. Moreover, for high strained (necking) areas the increase in temperature leads to the recrystallization and formation of small grains (Figs. 12b, c and 16b). Unlike heterogeneity of Cu microstructure observed close to the junctions of Cu and FeNi36 after the first cold rolling, the Cu microstructure in those regions after the second cold rolling is rather homogeneous (Fig. 14). Nevertheless, EBSD analysis still revealed a poor bonding strength of Cu-Cu interfaces close to triple junctions of Cu-Cu and FeNi36, certainly because the FeNi36 sheet limited local contact. Moreover, the analysis of strain hardening in the same thermal bridge revealed that the strain hardening of Cu is not uniform. It seems that

such heterogeneity in strain hardening depends on the rolling direction. It can be seen that the triple junction of Cu and FeNi36 first submitted to deformation is more work hardened than the second one (Fig. 14). This heterogeneity was not observed after the first cold rolling, probably, due to important thickness reduction (three times higher than that during the second cold rolling). It can be suggested thus that during the cold rolling (first and second) the strain is not uniform in the same thermal bridge with more strained Cu in the regions first submitted to rolling. Therefore, the adherence of Cu-Cu interfaces should be different as well. Such a conclusion could be one of the explanations of decoupling of Cu-Cu interfaces observed during the heat treatment (Fig. 8). Indeed, the decoupling is more pronounced for the area of the thermal bridge which was last subjected to the rolling. The heterogeneities of strain hardening in Cu should play an important role on the damage endured during service life and should be taken into account. Further work is in progress to avoid the occurrence of defects like those shown in Figs. 8 and 11 which are usually referred to as “kissing bonds” in welding. The elements given in Section 3.3 suggest reconsidering the heat treatment parameters between the first and second rolling passes to reduce the effect of the CTE mismatch at this stage of the fabrication process. Currently solutions are sought in optimizing roll-bonding parameters and the initial geometry of the perforation. 5. Conclusion The innovative Thermal Bridge Composite (i-TBC) was proposed as a substrate material for power electronic systems. i-TBC is an architectured material based on Cu and Invar™ FeNi36. Due to the simultaneous use of Cu and FeNi36, i-TBC possesses both a good thermal transverse conductivity and a limited longitudinal CTE. Therefore, it can provide an efficient solution for electronic packaging issues for the high temperature applications (up to 250 °C). The main innovation of the iTBC consists in the formation of a bond area between the Cu sheets through the perforations in FeNi36 sheet. These bond areas, called thermal bridges, provide a good transverse thermal conductivity to the iTBC. The formation of the thermal bridges was investigated through three different elaboration steps: first cold rolling, heat treatment and second cold rolling. The obtained results can be summarized as follows:

Fig. 16. Schematic illustration of Cu\ \Cu bonding in the centre of the thermal bridge during the second cold rolling: (a) Cu-Cu interface before the second cold rolling, (b) Cu-Cu interface after the second rolling.

• The first cold rolling did not lead to a perfect adherence of the Cu-Cu interface in the thermal bridge. The Cu grain structure was found to be very heterogeneous with elongated grains at the centre of the thermal bridge, typical for cold rolling, and ultra-fine grains close to the Cu-FeNi36 junctions.

78


• The heat treatment carried out at 450 °C for 6 h resulted in a complete recrystallization of Cu grains in the thermal bridge. Moreover, the adherence of Cu-Cu interfaces was further deteriorated. • The second cold rolling led to a perfect bonding of Cu-Cu interfaces at the centre of the thermal bridge while several unbound zones were observed close to Cu-FeNi36 junctions. A gradient of strain hardening for Cu grains depending on the rolling direction was observed. • The Cu-Cu welding by cold rolling is very heterogeneous when carried out through the perforations in other material with different properties (FeNi36 in our case). However, a good welding quality can be obtained through appropriate several elaboration steps, as it is done for i-TBC. Acknowledgements TG-GRISET (B. Pierre and D. Guinet) is gratefully acknowledged for the samples providing and for technical discussions. The work is carried out in the framework of MEGaN project of poles de compétitivité Mov'eo and Minalogic and supported by BPI France and ARMINES (contract 40288 PSPC MEGAN). References [1] C. Buttay, D. Planson, B. Allard, D. Bergone, State of the art of high temperature power electronics, Mater Sci Eng B 176 (2011) 283–288. [2] S. Gao, S. Yuki, H. Osanai, W. Sun, K.D. Ngo, G.-Q. Lu, Thermo-mechanical reliability of high-temperature power modules with metal-ceramic substrates and sintered silver joints, Proceedings of International Conference on Electronics Packaging (ICEP) 2016, pp. 395–399. [3] Akihisa Fukumoto, D. Berry, Khai D.T. Nago, Effects of extreme temperature swings (−55 °C to 250 °C) on silicon nitride active metal brazing substrates, IEEE Trans Device Mater Reliab 14 (2014) 751–756. [4] A. Kaabi, Y. Bienvenu, D. Ryckelynck, B. Pierre, Architectured materials to improve the reliability of power electronics modules: substrate and lead-free solder, J Electron Mater 43 (2014) 648–657. [5] A. Kaabi, Y. Bienvenu, D. Ryckelynck, L. Prévond, B. Pierre, Architectured bimetallic laminates by roll bonding: bonding mechanisms and applications, Mater Sci Technol 30 (2014) 782–790. [6] Stuart I. Wright, Matthew M. Nowell, David P. Field, A review of strain analysis using electron backscatter diffraction, Microsc Microanal 17 (2011) 316–329. [7] A.J. Schwartz, M. Kumar, B.L. Adams, D.P. Field, Electron Backscatter Diffraction in Materials Science, Springer, US, Boston, MA, 2009. [8] C. Moussa, M. Bernacki, R. Besnard, N. Bozzolo, About quantitative EBSD analysis of deformation and recovery substructures in pure tantalum, IOP Conf Ser Mater Sci Eng 89 (2015) 012038. [9] M. Tikhonova, Y. Kuzminova, X. Fang, W. Wang, R. Kaibyshev, A. Belyakov, Σ3 CSL boundary distributions in an austenitic stainless steel subjected to multidirectional forging followed by annealing, Philos Mag 94 (2014) 4181–4196.

[10] S. Jakani, T. Baudin, C.-H. de Novion, M.-H. Mathon, Effect of impurities on the recrystallization texture in commercially pure copper-ETP wires, Mater Sci Eng A 456 (2007) 261–269. [11] G. Benchabane, Z. Boumerzoug, T. Gloriant, I. Thibon, Microstructural characterization and recrystallization kinetics of cold rolled copper, Phys B Condens Matter 406 (2011) 1973–1976. [12] R.Z. Valiev, R.K. Islamgaliev, I.V. Alexandrov, Bulk nanostructured materials from severe plastic deformation, Prog Mater Sci 45 (2000) 103–189. [13] L. Kunz, L. Collini, Mechanical properties of copper processed by Equal Channel Angular Pressing-a review, Frat. Ed. Integritá Strutt. (2012) 19. [14] E.I. Teitel', L.S. Metlov, D.V. Gunderov, A.V. Korznikov, On the structural and phase transformations in solids induced by severe plastic deformation, Phys Met Metallogr 113 (2012) 1162–1168. [15] A.M. Glezer, A new approach to a description of structural-phase transformations under a very severe plastic deformation, Russ Phys J 51 (2008) 480–491. [16] F.J. Humphreys, M. Hatherly, Recrystallization and Related Annealing Phenomena, second ed. Elsevier, Oxford, 1995. [17] S.R. Ahl, H. Simons, Y.B. Zhang, C. Detlefs, F. Stöhr, A.C. Jakobsen, D. Juul Jensen, H.F. Poulsen, Ultra-low-angle boundary networks within recrystallizing grains, Scr Mater 139 (2017) 87–91. [18] W. Zhang et, N. Bay, Cold welding-theoretical modeling of the weld formation, Weld J- Weld Res Suppl 76 (1997) 477s. [19] L. Li, K. Nagai, F. Yin, Progress in cold roll bonding of metals, Sci Technol Adv Mater 9 (2008) 023001. [20] M. Hosseini, H. Danesh Manesh, Bond strength optimization of Ti/Cu/Ti clad composites produced by roll-bonding, Mater Des 81 (2015) 122–132. [21] R. Jamaati, M.R. Toroghinejad, Cold roll bonding bond strengths: review, Mater Sci Technol 27 (2011) 1101–1108. [22] S.A. Hosseini, M. Hosseini, H. Danesh Manesh, Bond strength evaluation of roll bonded bi-layer copper alloy strips in different rolling conditions, Mater Des 32 (2011) 76–81. [23] R. Jamaati, M.R. Toroghinejad, Effect of Al2O3 nano-particles on the bond strength in CRB process, Mater Sci Eng A 527 (2010) 4858–4863. [24] H. Yan, J.G. Lenard, A study of warm and cold roll-bonding of an aluminium alloy, Mater Sci Eng A 385 (2004) 419–428. [25] Y. Jiang, D. Peng, D. Lu, L. Li, Analysis of clad sheet bonding by cold rolling, J Mater Process Technol 105 (2000) 32–37. [26] C. Barlow, P. Nielsen, N. Hansen, Multilayer roll bonded aluminium foil: processing, microstructure and flow stress, Acta Mater 52 (2004) 3967–3972. [27] C. Wang, Y. Jiang, J. Xie, D. Zhou, X. Zhang, Interface formation and bonding mechanism of embedded aluminum-steel composite sheet during cold roll bonding, Mater Sci Eng A (2017) https://doi.org/10.1016/j.msea.2017.09.111. [28] C. Wang, Y. Jiang, J. Xie, D. Zhou, X. Zhang, Effect of the steel sheet surface hardening state on interfacial bonding strength of embedded aluminum-steel composite sheet produced by cold roll bonding process, Mater Sci Eng A 652 (2016) 51–58. [29] B. Wu, L. Li, C. Xia, X. Guo, D. Zhou, Effect of surface nitriding treatment in a steel plate on the interfacial bonding strength of the aluminum/steel clad sheets by the cold roll bonding process, Mater Sci Eng A 682 (2017) 270–278. [30] N. El Mahallawy, A. Fathy, W. Abdelaziem, M. Hassan, Microstructure evolution and mechanical properties of Al/Al–12%Si multilayer processed by accumulative roll bonding (ARB), Mater Sci Eng A 647 (2015) 127–135. [31] M. Sato, N. Tsuji, Y. Minaminob, Y. Koizumi, Formation of nanocrystalline surface layers in various metallic materials by near surface severe plastic deformation, Sci Technol Adv Mater 5 (2004) 145–152.