Online proceedings - EDA Publishing Association

24-26 September 2008, Rome, Italyhigh temperature stability of the bond. The bonding Cu/Ag/Cu detect the temperature gradient along the Agbond.Due to the impossibility to measure the gradienttemperature is similar to bonding temperatures of solders,but the bond withstands much higher working temperature in directly by sensors, FE-Simulation is used to make up forcomparison to standard solders. The joining process is based that impossibility. So the only unknown is the thermalon a solid state reaction without liquid phase as the melting conductivity of the Ag layer.point of silver is 961°C. The sintered material shows also aHeaterL1good electrical conductivity. Both thermal and electricalj_thPrim. Circuit Sec. Circuitconductivity should be clearly related to the porosity.Ideally chip and substrate have a closed silvermetallisation of any thickness. Gold metallisations are alsoused but requires adopted parameters. Silver powder isavailable in different shapes like flakes and spheres andvarious particle sizes. Flakes powders are available withmost particles smaller than 10 µm. Spherical powders are onhand with even lower particle sizes down to the sub-micronscale.Fig. 6: FIB analysis: Interfaces bonded with differentpressures (different scales)For a better applicability the powder is often embedded ina liquid matrix, for example solvents which vaporize duringthe bonding process. Various methods for the application ofthe sintering material are possible like screen or stencilprinting, dispensing or spaying. The powder is for exampleapplied on the substrate. To obtain the final bond, substrateand chip are bonded by use of pressure and temperature.Figure 5 shows a sketch of the bonding. By the variation ofpressure the porosity of the final joint can be adjusted (seefigure 6).Azu (area)TIM #2VacChamberWater CoolerR_thL2DBondPI / EPOXYFig. 7: Test specimen for nano-Ag measurement. Forprocess reasons, the specimen needs to be flat. The layerto be tested uses up nearly all the heat flow (primarycirtuit).To come up with a suitable design we made 10 FE-Simulations changing the cross-sectional-area (Azu), thedistance between heater and Ag-bond (L1), the distancebetween the Ag-bond and mechanical stabilization (L2) asdesign paramters. We also took the influence of air asthermal conductor between the copper bars into account.Evaluation criterions are heat flux through the Ag-bond(PAG), heat flux through the mechanical stabilizer (Pmech),heat flux through air (Pair) and the temperature gradientalong the Ag bond (∆TAG).Variante 8: L1 = 30mm;Azu = 40mmB. Simulation & Specimen DesignDue to the technological boundary conditions, we decidedto design a completely new specimen (figure 7). The testspecimenconsists of seven main parts, the electric heaterwitch is connected to the upper copper bar, the small sinteredAg bond as thermal connector between the upper and thelower copper bar, the polymer bond at the end to obtainmechanical stability, the temperature sensors and the heatsinkfor cooling which is connected to the lower copper bar.The working principle is analogous to the one in figure 1,using a steady state technique: A constant temperature set bythe electric heater as well as the heat-sink generates aconstant thermal flow through the specimen. To determinethe thermal conductivity of the Ag we use the knowledge ofthe geometry and the thermal conductivity of the copper bar.The heat flow and the temperature gradient will be measuredwith temperature sensors. The first three sensors integratedin the upper copper bar will be used to detect the total heatflow though the bar. Two more sensors near the interfaceFig. 8: FE-Simulation of test-specimenFigure 8 exemplifies the temperature distribution inthe test specimen. Table 2 shows the parametric changesmade during simulations. The results of the FE-simulationsare shown table 3. Material Properties used in thesimulations are λCU = 401 W/mK, λAG = 200 W/mK, λAIR= 0,026 W/mK, λPI/EPO = 0,3 W/mK and λVAC ≈ 0W/mK. The thickness of the gap is 100µm and themechanical stabilizer has an area of 20mm².Variation V0 is the standard test specimen without theinfluence of air and mechanical stabilizer. In this case we geta high temperature gradient and no failure heat flux causedby air or mechanical stabilization. The heat flux through airinvestigated in V1 and V3 reach approximately one third ofthe total heat flux. For the experiment this will cause a highfailure in determination of thermal conductivity.©EDA Publishing/THERMINIC 2008 115ISBN: 978-2-35500-008-9