[IEEE 2014 15th International Conference on Thermal, Mechanical and Multi-Physics Simulation and Experiments in Microelectronics and Microsystems (EuroSimE) - Belgium (2014.04.7-2014.04.9)] 2014 15th International Conference on Thermal, Mechanical and Mulit-Physics Simulation and Experiments in Microelectronics and Microsystems (EuroSimE) - CMOS stress sensor for 3D integrated circuits: Thermo-mechanical effects of Through Silicon Via (TSV) on surrounding silicon

CMOS STRESS SENSOR FOR 3D INTEGRATED CIRCUITS: THERMO-MECHANICAL EFFECTS OF THROUGH SILICON VIA (TSV) ON SURROUNDING SILICON Komi Atchou EWUAME1,2,., Vincent FIORI1,b, Karim INAL2,", Pierre-Olivier BOUCHARD2,d, Sebastien GALLOIS­ GARREIGNOT1,", Sylvain LIONTIlJ, Clement TAVERNIER1,g, Herve JAOUEN1,h lSTMicroelectronics, 850 rue Jean Monnet, 38926 Crolles Cedex, France 2CEMEF, Mines ParisTech, rue Claude Daunesse CS 10207,06904 Sophia Antipolis Cedex, France 'komiatchou,ewuame@stcom, bvincentfiori@stcom, ckarim,inal@mines-paristech,fr, dpierre­ olivier, bouchard@mines-paristech,fr, ese bastien, gallois-garreignot@stcom, fsy I vainJionti@stcom, gclementtavemier@stcom, hherve.jaouen@stcom Abstract This work aims at determining thermo­ mechanical stresses induced by annealed copper filled Through Silicon Via (TSV) in single crystalline silicon by using MOS (Metal Oxide Semiconductor) rosette sensors. These sensors were specifically designed and embedded. Through the piezoresistive relations, the stress tensor was evaluated by carrying out electrical measurements on test vehicle. The MOS stress sensors would have been needed to be calibrated: first results of the calibration were obtained however, since they were still partial, they were not used to make the bridge from electric to mechanic quantities. Experimental findings were based on the direct calculation of stresses from electrical measurements data and literature piezoresistive coefficients. In order to get only the TSV contribution and to suppress the manufacturing process variability contribution, an optimization calculation was needed. A finite element approach was also adopted to evaluate numerically the stresses induced by TSV. The stress values obtained from the optimization are in the range of the ones obtained by simulation in the sensor area. Thus, it can be stated that the methodology is relevant, and the results will be confirmed by extracting the true piezoresistive coefficients for the embedded MOS. Once calibration performed, the piezoreslstlve coefficients should enable getting more accurate stress values. At this stage, the quite good agreement between numerical and experimental results seems promising. 1. Introduction The ever increasing improvement of semiconductor technologies related to 3D stacking processes, brings thermo-mechanical and residual stress issues to the forefront [1]. Numerous studies were published regarding stress fields distribution within 3D structures and the stress consequences on electrical devices were explored [2]. Among the different measurement methods allowing stress evaluation within devices, several in-situ MOS based stress sensors have been widely developed [3,4]. In this paper, combined numerical and experimental investigations will be performed: calibration step will be carried out in order to determine the true relation between electrical current measurement and stress evaluation for these sensors. Then, a four point bending tool, which was specifically designed for sensors calibration will be presented. Finally, thanks to the evaluation of the piezoresistive coefficients and through the piezoresistivity relation [5], the link between drain current and stress will be established in order to assess specifically the TSV (Through Silicon Via) effects. TSV s are structures connecting multiple layers of stacked integrated circuits or 3D chips to deliver better performance, more functionalities and smaller packages of chips while consuming less power. On the other hand, finite element models (FEM), focusing on the impact of TSV on the surrounding silicon will be presented. Numerical results will be compared to experimental measurements for correlation purposes. 978-1-4799-4790-4/14/$31.00 ©20 14 IEEE - 1 / 8 - 2014 15th International Conference on Thermal, Mechanical and Multi-Physics Simulation and Experiments in Microelectronics and Microsystems, EuroSimE 2014 Comprehensive assessment of the induced stresses allows avoiding the priming of the first failure within products, the propagation of the failure in sub-layers. This also allows drawing design guidelines by preventing detrimental thermo-mechanical effects of TSV s on device performances, controlling the TSV process, optimlzmg products manufacturing and ensuring their quality and reliability. 2. Piezoresistive relations Piezoresistivity describes the change in the electrical resistivity when mechanical stress is applied. This effect particularly occurs in semiconductor materials, making piezoresistive devices an attractive mean for stress sensoring in the microelectronic industry. In this part, the relations used for describing piezoresistivity are defined and its application is presented. The MOS stress sensor (Fig. 1) is made up of four transistors in which channels are oriented in different crystallographic directions. As is always the case, MOS are made into the silicon, stress field is hence evaluated in a thin region of the chip, near the rosette. A tensor relation comes from the material properties and the "theory of piezoresistivity". According to this theory, the resistivity variation !1p, as well as the resistance (!1R) and the drain current (!1Id) variations, is related to the applied stress by the relation (Eq.1) [5]: !1R = _ Md = I1p = n!1cr (1) R Id P With n the tensor of piezoresistive coefficients and �cr the applied mechanical stress tensor. The tensor I1p (respectively the tensors !1R and p R Md) is built from scalar measurements of Id resistivity (respectively of resistance and drain current) in different directions. (110) S : Source Di : Drain G: Gate B: Bulk [1(0) Figure 1- MOS rosette stress sensor This relation is the same as the one used in microelectronics for evaluating stress in devices. Note that the resistance variation is often preferred for passive serpentines or doped active sensors ([3, 4]). In our transistor based ones, the drain current variation is rather used. For the silicon cubic crystal, the three independent coefficients (7T11' 7T12 and 7T44) are written in a shortened matrix (6x6) form, in ([100], [010], [001]) coordinate system (Eq. 2): 1[11 1[12 1[12 0 0 0 1[12 1[11 1[12 0 0 0 [n] = 1[12 1[12 1[11 0 0 0 (2) 0 0 0 1[44 0 0 0 0 0 0 1[44 0 0 0 0 0 0 1[44 The knowledge of these coefficients IS mandatory and their extraction process is called calibration. 3. Calibration In the microelectronic industry, (001) substrate and two types of wafers are commonly used for chips manufacturing: the so-called rotated and not rotated wafers. For a rotated or wafer, the x axis (respectively y axis) corresponds to the [100] (respectively [010]) direction (Fig. 2). For a not rotated or wafer, the x axis (respectively y axis) corresponds to the [11 0] (respectively C[ 1 0]) direction (Fig. 3). (010) y Figure 2- Schematic representation of wafer 0. -I-__ *",,�'---7-->'(110J MOS (100J Figure 3- Schematic representation of wafer - 2 / 8 - 2014 15th International Conference on Thermal, Mechanical and Multi-Physics Simulation and Experiments in Microelectronics and Microsystems, EuroSimE 2014 a is the angle between the MOS channel and the x aXiS. So, for a wafer, the coordinate system transformation (i.e. rotation of 45° around the z axis) of [11:] leads to the following piezoresistive coefficients matrix (Eq. 3): [rc'] = [T] [rc] [Tr1 (3) With [T] (Eq. 4), the transformation matrix [6] and [T]"l its inverse; [rc'] the new piezoresistive coefficients matrix (Eq. 4). The calibration will be performed on the aforedescribed MOS rosette (Fig. 1). Note that, in the following, Idi represents the drain current of the MOS i and 12i its initial value; and the index of the drain current is related to the MOS index inside the rosette. 1 1 0 0 0 "2 (7T11 + 7T12 + 7T44) "2 (7T11 + 7T12 - 7T44) 7T12 1 1 0 0 0 - (7T11 + 7T12 - 7T44) - (7T11 + 7T12 + 7T44) 7T12 2 2 [7T�5°]= 7T12 7T12 0 0 0 0 0 0 To perform the calibration, a four-point bending machine (Fig. 4) is designed to apply a known uniform uniaxial stress into the sample. Figure 4- Pictures of the four point bending tool wafer First, a sample is cut along the [100] direction. A uniform uniaxial stress (LlaLl) is applied in the [100] direction. Electrical measurements are performed on two MOS of the sensor (MOS 1 and MOS2). The combination of Eq.l and Eq.2 allows obtaining: Mdz 1 Mdl 1 n12 = - -0--- ; n11 = - -0--- (5) I dz L1CTLl I dl L1CTLl Then, another sample is cut at 45° (i.e. in the [110] direction). If a uniform uniaxial stress (Lla'L1) is applied in that direction, an electrical measurement on MOS3 allows obtaining the last piezoresistive coefficient n44 from relations Eq. 1 and Eq. 4: L1Id3 1 , - -0- = - (n11 +n12+n44) Lla L1 Id3 2 7T11 0 0 0 (4) 0 7T44 0 0 0 0 7T44 0 0 0 0 7T11 - 7T12 7T = _ _ 2_ Md3 + _1_ (Mdl + MdZ) => 1044 0 0 0 L1CT'Ll Id3 L1CTLl Idl Idz (6) wafer The procedure is the same as the one described above. The difference here, is that the sample is first cut along the [110] direction, and then in the [100] direction. It is so needed to take care about the index of the MOS. This calibration methodology is the same for n-MOS and p-MOS. The calibration was started and the preliminary results obtained are plotted in Fig. 5 . 10,0% 8,0% g 6,0% .. 0 4,0% '., ... ';: ... 2,0% ,. � .. !!! 0,0% 1i .. ·2,0% . � Q -4,0% ·6,0% -8,0% • nMOS [1001 . nMOS[OlOI -Trendline nMOS [1001 -Trendline nMOS lQ!QI __ -140-120 -100 -so -60 -40 -20 0 20 40 60 80 100 120 140 Stress (MPa) Figure 5- Preliminary results of calibration One can notice that all the dots in Fig. 5 for each MOS type are almost aligned and the trend lines pass close to the axes origin. The slopes of these lines correspond to the piezoresistive - 3 / 8 - 2014 15th International Conference on Thermal, Mechanical and Multi-Physics Simulation and Experiments in Microelectronics and Microsystems, EuroSimE 2014 coefficients. However, these results are still partial and the whole coefficients have not been determined yet. Until the complete measurement of the piezoresistive tensor, the coefficients extracted from previous works and related to similar MOS types are used in the following sections for stress evaluation. aiO"xx + biO"yy + CiO"zz + diO"xy = a20"xx + b20"yy + C20"zz + d20"xy = a30"xx + b30"yy + C30"zz + d30"xy = a40"xx + b40"yy + C40"zz + d40"xy = With ab bi, Ci and di the coefficients defined in appendix for each wafer type. Again, these relations are unchanged for n-MOS and p-MOS. 5. Application to TSV induced stress Test vehicle, description and sensors location In this part, MOS rosette sensors, of both n and p types were used to evaluate the stress induced by the TSV. Two types of CMOS65 samples were manufactured: a wafer with TSVs and a one without TSVs. Note that designs are similar, and the presence of TSV s is managed by process routes: one sample followed the whole flow, whereas the other skipped the TSV s related steps. As shown in Fig. 6, the sensors were embedded into the wafers near the TSV s (i.e. at 121-Im away from the center of the TSVs). Their size is given in Fig. 1. Electrical measurements from the two wafers were subtracted to evaluate the sole effect of the TSV. Electrical measurements Drain currents variations were measured on the aforementioned two wafers and the difference should correspond to the impact of TSV. The evaluation of stresses components requires four equations. As, there are four n-MOS and four p­ MOS, two configurations were considered: two 4. Stress evaluation It is then possible to evaluate stresses within a device thanks to electrical measurements. In the literature ([3, 4] for example), vertical ( a zz) and shear stresses (a xy) are sometimes not evaluated. In this section, the relations that allow evaluating these stresses are presented. By combining Eq. 1, Eq. 2 and Eq. 5, one can write: Mdl (a = 0°) I�l Md2 (a = 90°) I�2 Md3 (7) I�3 (a = 45°) Md4 (a = -45°) I�4 p-MOS oriented at 0° and 90° and two n-MOS at 45° and -45° (configl ), or two n-MOS at 0° and 90° and two p-MOS at 45° and -45° (config2). As an alternative to our own four points bending calibration, which is planned to be completed sooner, piezoresistive coefficients taken from the literature [5] were used for this study. Results of the direct calculation of stresses from Eq. 4, for the two configurations are presented in Table I-b. However, preliminary internal studies showed that, independently to any stress variation, the current values of transistors are not the same between two similar wafers: this is related to the intrinsic variability of the CMOS process (Fig. 7-a). [110J [010J [110J Figure 6- MOS rosette stress sensor and TSV location - 4 / 8 - 2014 15th International Conference on Thermal, Mechanical and Multi-Physics Simulation and Experiments in Microelectronics and Microsystems, EuroSimE 2014 0,000013 0,0000125 0,000012 0,0000115 0,000011 0,0000105 0,00001 0,0000095 a) Wafer I Wafer 1 2 Wafer I Wafer 1 2 Wafer I Wafer 1 2 o· 45' -45' Wafers and orientations b) Wafers and 1\'108 Variation ofId orientation from wafer1 (0/0) 0° 45° -45° 90° Wafer! Wafer2 Wafer! Wafer2 Wafer! Wafer2 Wafer! Vlafer2 r- C- - I- I- wafer1 Wafer 1 2 ". Figure 7- Representative diagram of process variability Stress identification So, in order to get only the TSV contribution, measurements were averaged and an optimization calculation was performed. Since the sensors provide a limited amount of equations and considering the whole unknowns of the system to solve, some assumptions had to be made: the drain currents for the process variability in [100] (0°) (respectively in [110] (45°)) direction are found to be the same as in [010] (90°) (respectively in CL IO] (-45°)) direction (Fig. 7-b). The simplified system is then obtained by adding the term L1�dVi to each 'dvi line of Eq. 7. With .1Idvi the drain current variation related to process variability in the transistor number i of the rosette (see Fig. 1) and Igvi its nominal value. For this work, the so-called evolutionary optimization excel solver was used. This class of optimization method is particularly well adapted for avoiding local minima [7]. Based on the system of equations Eq. 7, a global cost function is built as the sum of the cost functions associated with each equation. By minimizing this global cost function, a set of stress values is found. The optimization initial conditions, boundaries and the results are summarized in the following table (Table I-a): Table 1- a) Table summarizing the initial conditions, boundaries and results from optimization, b) direct calculation Initial Lower Upper Obtained a) values limits limits values [MPa] [MPa] [MPa] [MPa] ox-x 75 -100 100 66,08 Oyy -75 -100 100 23,05 Ozz -25 -100 100 13,04 Oxv 25 -50 50 5,19 Cost function 0,01306 Stress values Stress values b) [MPa] [MPa] Configl Config2 Oxx 19 95 Oyy -51 6 Ozz -51 -109 Oxy 0,3 -7 The comparison of these results shows that the whole methodologies do not give similar stress fields, which is not suitable. Considering the direct approach (b), differences between configurations 1 and 2 would be attributed to piezoresistive coefficients inaccuracy and also to the fact that process variability is neglected. Hence, this method cannot be used as it and four point bending calibration is mandatory. As a consequence, it is needed to include the process variability and the measurements from the eight MOS in the equation system: the optimization method (a) should then provide a reliable estimation of the stress fields, which are: axx = 66,08 MPa ; ayy = -23,05 MPa ; azz = -13,04 MPa; axy = -5,19 MPa. TSV stress 3D simulation This section deals with the numerical evaluation of stress fields induced by an isolated TSV in the silicon. Finite element simulations were performed with and without TSV using the ANSYS 14.5.R finite elements software. The copper TSV has a diameter of 10/lm and a height of 80/lm. The silicon oxide liner surrounding TSV has a thickness of 0.2/lm and a height of 80/lm. The model is 48/lm long and 40/lm wide. The PMD (Pre-Metal Dielectric), the low-k and the Si02 based interconnects have respectively thicknesses of OA/lm, 1.67/lm and 4.74/lm. - 5 / 8 - 2014 15th International Conference on Thermal, Mechanical and Multi-Physics Simulation and Experiments in Microelectronics and Microsystems, EuroSimE 2014 Due to geometrical symmetries, only one fourth of the stack was modeled (Fig. 8). Hexahedral quadratic elements (20 nodes) were used. Regarding boundary conditions, the node (x=O, y=O, z=O) was locked and the normal displacements of the areas A, B and the bottom base (z=O) were set to zero. The loading condition was a uniform thermal cooling from copper stress free temperature (i.e. 260°C) [8] to room temperature. No material property variation during cooling was also assumed. Materials were considered as isotropic and perfect adhesion was assumed at interfaces. Figure 8- One fourth of the meshed model Hence, simulations were performed under linear elastic hypothesis. The interconnects being patterned layers, analysis of the designed metal densities of the homogenized BEoL showed composition of 70% copper and 30% dielectrics. The stresses induced by TSV were obtained by subtracting stress fields from the model with TSV (Fig. 9) and those from the model without TSV. Stress variations were extracted at the top, middle and at the bottom of the models (Fig. 10). �"'-+TSV Liner Silicon Stress [MPa I ·350 o 1+----' Symmetry +350 planes Figure 9- Stress Uxx for one fourth of the simulated model with TSV 250 • • ••• BOTTOM . . 200 ��-r---�-�--� . .. . . . � . .. • • •• lvITDDLE . � , 150 +-. --.:;.,-.. -+__--f----'C.-'.-•• '-'.'-'I--'-'TO�P_---1 � . .. . . . . g 100 +-�� �.�. --" •• r-----f----�-------1 f.I'l • • • � .. . . . . . � 50 +---� • .r.�.�-r----�--� . . . .. . .. . ..... :: : : : ! i� · O +-----+__------'--'�!..1..L�:U"O'�___1 5 10 15 a) Distance from TSV [�ml 5 10 15 20 25 20 25 0 -,-------,----------,---------,----,--:--,-------, . ... . ; ��� ')�)))))� . . . . . - . " , ' -100 +-_"'-, •• -y}+.-'--: .'---. --+_----+_--_____i .. .. : . . . . "CO • • •• ,§§ , -200 -t::�::."_ • .=--+-----+-----+--------i - . . C> -. : 5' -300 -fL,;'------+-------+-----+----_i BOTTOM � : MIDDLE � -400 -t::�--_+_----+_----'-' •• ....".-,-. T=�""'--I . • ••• TOP -500 -'---____ L--__ ----'-____ ----'-____ --' b) Distance from TSV [�ml 25 . . . o +-i-. --•• -=.---. �.'-"-"-'�-f.��._.jo_��_i . . . . .. . � . . . . . .. . . . . ::2 : ••• �-25 �.���.�. �������� � -. . . ..... � . . . b ••• � . ; -50 �.'---- -- +- ---- +__�.�.� •• � .�B�OT �TO�M_____i tt:! .: ••••• :NIIDDLE ••••• TOP -75 -'--------'--------'--------'--------" 5 10 15 c) Distance from TSV [�ml 20 25 Figure 10- Stresses a)- O"xx, b)- O"yy, c)- O"zz mduced by TSV at the top, middle and at the bottom of the model One can note that, in the silicon material, TSV induces tensile stresses (O"xx) in the radial (x axis) direction, whereas in tangential (y axis) and vertical (z axis) directions, the stresses (O"m O"zz) are compreSSive. Indeed, during the cooling, the model is contracted in the tangential direction ( compressive stress), which by Poisson effect, generates a tensile stress in the perpendicular direction (radial direction). The surfaces of the model, except those to which are applied boundary conditions are free, so that from the center of the TSV to these surfaces the stress values decrease and tend to zero. This also occurs from the bottom to the top of the model as shown in Fig. 10 a), b). - 6 / 8 - 2014 15th International Conference on Thermal, Mechanical and Multi-PhYSiCS Simulation and Experiments in Microelectronics and Microsystems, EuroSimE 2014 A companson of simulation results with the literature was limited to a qualitative analysis, since configurations used (material properties, model dimensions) are not the same. One can mention the works of De Wolf and at. [2], and Li Yu and al. [9] in which, the radial 200 150 '" � 100 � 50 00 " b 0 rn -50 0 0 -25 '" -50 � t:l 0 -75 � -100 rn -125 0 TSV- --- - -- - -- TSV 5 5 , S ensor , ocatl on , '. - - " " - - - ---- -- 10 15 20 Distance from TS V (fJlll] - - , - .-. .. - .. ... -, - " l ocati on • 10 15 20 Distance from TSV [f.I1ll1 25 .. - 25 (respectively tangential and vertical) stresses obtained by finite element simulation and micro Raman are tensile (respectively compressive). These stresses fall back to zero as one moves from the center of the TSV. 50 0 ·50 r;;' � -100 � -150 [ -200 " b -250 rn -300 - - - - - - • • - --- - - -;.:-� .. - ---- --- TSV '--. 0 - , , , I I 5 ... , 3ensor ocati on 10 15 Distance from TSV [f.I1ll] Simul : results Simul : minimum value in sensor area Simul : maximum v alue in sen sor area Exp I-a: direct c.alculaion (configl ) Exp I-b: direct calculation (confi g2) Exp2: optimization 20 25 Figure 11- Stress induced by TSV at the top of the model, experimental and numerical results According to the location of the sensor at the top of the model (from x=7.5j1m to x=16.5j1m), the stress variation, i.e. the difference between maximum and minimum values, is quite large (L�O"xx=-80MPa, �O"yy=+80MPa and �O"zz=+30MPa). Hence, the comparison between sensor size and simulated stress fields shows that the accuracy of such sensor is likely to be weak. Averaged simulated stresses values and their variations in the sensor region are: (jxx�30MPa±40, (jyy=-55MPa±40 and (jzz�-20MPa±15. According to Fig. 11, the stress values (except O"zz) obtained from direct calculation for the first configuration are in the range of stresses obtained by simulation in the sensor area: O"xx E:[-10MPa, 70MPa], O"yy E:[-95MPa, -15MPa] and O"zz E:[-35MPa, -5MPa]. But the stress values are out of the range for the second configuration. Furthermore, as mentioned before, this method cannot be used and these results must be dismissed. The technique employing the optimization procedure seems to be relevant by taking into account the process variability. However, it is difficult at this stage to confirm stress values calculated from experiments since the exact piezoresistive coefficients for the MOS sensors have not yet been calibrated. 6. Conclusion In order to evaluate the stress induced by TSV in microelectronic devices, MOS rosette stress sensors were used, and the results are reported in this work. The calibration of the sensors was started and preliminary results were obtained. These results are partial and the piezoresistive coefficients were not fully determined, therefore they were not used for stress evaluation. In this work, a combined approach of experiment and simulation was carried out. Experimental investigations were based on the direct - 7 / 8 - 2014 15th International Conference on Thermal, Mechanical and Multi-Physics Simulation and Experiments in Microelectronics and Microsystems, EuroSimE 2014 calculation of stresses from electrical measurements data and literature piezoresistive coefficients. This is in part related to process variability which leads to the need for using an optimization calculation in order to get only the TSV contribution. A finite element approach was also adopted to evaluate numerically the stresses induced by TSV. The stress values obtained from the optimization are in the range of the ones provided by simulation in the sensor area. Thus, it can be stated that the methodology is relevant, and at first, the results will be confirmed by extracting the true piezoresistive coefficients for the embedded MOS. At this stage however, the quite good agreement between numerical and experimental results seems promising. Pros and cons of such sensors allow determining a reliable work flow for stress probing. This should enable to provide further design and integration recommendations. 7. Appendix wafer References 1. International Technology Roadmap for Semiconductors (ITRS), 2010. 2. 1. De Wolf, V. Simons, V. Cherman, R. Labie, B. Vandevelde, E. Beyne and IMEC, In­ depth Raman Spectroscopy Analysis of Various Parameters Affecting the Mechanical Stress near the Surface and Bulk of Cu-TSV s, IEEE, 978-1-4673-1965-2,2012. 3. K. Aditya, X. Zhang, Q. X. Zhang, M. C. Jong, G. Huang, L. W. S. Vincent, C. Lee, J. H. Lau, D. L. Kwong, R. R. Tummula and G. Meyer-Berg, Residual Stress Analysis in Thin Device Wafer Using Piezoresistive Stress Sensor, IEEE Transactions on Components Packaging and Manufacturing Technology, vol. 1, p. 6,2011. 4. X. Zhang, K. Aditya, Q. Zhanga, Y. Onga, S. Hoa, C. Khonga, V. Kripesha, J. Laua, D.-L. Kwonga, V. Sundaramb, R. Tummulab and G. Meyer-Berg, Application of Piezoresistive Stress Sensors in Ultra Thin Device Handling and Characterization, Sensors and Actuators A156, pp. 2-7, 2009. 5. C. S. Smith, Piezoresistance effect in germanium and silicon, Phys. Rev., vol. 94, pp. 42-49, 1954. 6. J. C. Suhling, R. C. Jaeger, Silicon Piezoresistive Stress Sensors and Their Application in Electronic Packaging, IEEE Sensors, vol. 1, No. 1, pp. 14-30,2001. 7. E. Roux, P-O. Bouchard, Kriging metamodel global optimization of clinching Jommg processes accounting for ductile damage, Journal of Materials Processsing Technology, 213(7):1038-1047,2013. 8. M. Gregoire, Properties of Thin Film and Copper Interconnects. Thesis in Science and Engineering of Materials. Grenoble : National Polytechnic Institute of Grenoble, 248p, 2006. 9. L. Yu, W. Chang, K. Zuo, J. Wang, D. Yu, D. Boning, Methodology for Analysis of TSV Stress Induced Transistor Variation and Circuit Performance, IEEE Quality Electronic Design, pp. 216-222, 2012. - 8 / 8 - 2014 15th International Conference on Thermal, Mechanical and Multi-Physics Simulation and Experiments in Microelectronics and Microsystems, EuroSimE 2014