Online proceedings - EDA Publishing Association

24-26 September 2008, Rome, Italydissimilar surfaces). It is therefore of importance to assure same materials and assembly technology as they used in thecharacterisation under real technological boundary real assembly applications and that it is suitable for allconditions.classes of TIM (incl. solder and adhesive).New accuracy challenges for TIM characterisation arisewhen high conductivity samples are to be measured with lowbond line thickness, as to be seen in figure 1 to the right.Main sources of error will be the bond line thickness, tilt ofcontact surfaces and temperature measurement.We have developed various test-stands and measuredthermal interface resistance and thermal conductivity ofTIMs for the most important market-going products rangingfrom thermal adhesives (1-3 W/mK) over greases (2-6W/mK), pads (3-10 W/mK), eutectic solders (30-45 W/mK)to mono-metal (200-400 W/mK) die attach materials.Hereby, thickness and pressure are monitored in-situ. Matingsurfaces are used as in real application (silicon and copper oraluminium). Structure analysis has been performed for monometal sintered silver TIM for porosity. Later, this will alsobe done for best-in-class TIM materials to obtain filler size,shape, material and modality as well as conformity to theinterfaces by FIB, SEM and EDX. For adhesives, greasesand pads, interface resistance has been obtained byextrapolation to zero BLT. Here, resistance measurementswere performed while systematically decreasing the gapwidth down to a few microns.In some cases, for greases it was found that theextrapolation scheme breaks down at very thin BLTs, asfiller size effects become apparent. As this regime displays alower thermal resistance, this effect is beneficial for e.g.greases. It is also shown that thinner BLT show conflictingtendencies with respect to reliability due to highermechanical stress.Further results show the development of thermalmeasurement equipment to deal with very high-lambdaTIMs as e.g. sintered silver die attach materials. Here, theprocess development is discussed with respect to structure.As this paper reports from a running project, further resultswill be published in another paper.II.GREASES & ADHESIVESAs investigated and discussed in [2] for Ag particle filledadhesives the shape of the filler particles and its positions inthe epoxy matrix are responsible for the thermal behavior,bulk conductivity and interface resistance, which weredetermined by cross sectioning.Further work had to be investigated on the particlebehavior of thermal greases below 50 µm BLT. For that amore accurate investigation determining the mechanism acton the thermal performance, the test stand described in [2]was developed further on, introducing an online thicknessdetermination, a more accurate temperature measurementand a protection cover to reduce influencing convection.A. Design of Test-Stand for Steady-State MeasurementThe philosophy of the steady state test stand is to havesurface conditions as they occur in real assemblies, using theTherefore a thermal test chip, assembled in flip-chiptechnology, is used and as hot plate, measuring thetemperature on top of the TIM as well. Beneath the TIM atest socket (Al or Cu) is used as cold plate. A NTC sensormeasures the temperature close to the bottom side of theTIM. The accuracy of the NTC’s is up to ~ 0,2 K.A = 1.4 cm 2Fig. 1: TIM-Tester and BLT-dependent resolution.The real heat flux through the TIM is estimated byusing the sensor beneath the TIM and another NTC sensor inthe test socket. Because the test socket is not fixed on a headsink, specimens can be easily changed.Fig. 2: Schematic of steady state test stand.To obtain the thermal bulk conductivity and the sumof both thermal resistances (R th0, Si-TIM and R th0, TIM-Al ) theeffective thermal resistance of minimum three different BLThas to be investigated. Knowing the exact thickness isessential to determine the thermal properties. Therefore adisplacement transducer (LVDT), going through the heat sinkand the test socket (Figure 3), was added to measure onlinethe BLT of the soft TIM (e.g. greases or pads) under theactual temperature and pressure conditions.Fig. 3: In situ thickness gauge by integrated LVDT.Further work optimizing the accuracy of the test-standhad been done by developing and introducing a calibration©EDA Publishing/THERMINIC 2008 113ISBN: 978-2-35500-008-9