The development of Pb-free solutions for the highreliability electronics community necessitates the consideration of hybrid microcircuit (HMC) products. This study used a test vehicle that included both plastic and ceramic packages as well as leaded and area-array solder joints on an alumina substrate. The conductor was a Ag-Pd thick film layer. The shear strength was measured for interconnections made with 63Sn-37Pb (wt.%, abbreviated Sn-Pb) and 95.5Sn-3.0Ag-0.5Cu (Sn-Ag-Cu) solders as a function of isothermal aging, thermal cycling, and thermal shock environments. The area-array packages indicated that solder joint fatigue was not altered significantly in a forward compatibility situation (i.e., Sn-Pb balls and a Sn-Ag-Cu assembly process). Local CTE mismatch fatigue strains are important for solder joints connecting ceramic area array packages to ceramic substrates. The gull-wing lead, SOT plastic package solder joints assembled with the Sn-Ag-Cu solder exhibit a greater strength loss under temperature cycling than did the corresponding Sn-Pb interconnections. Thermal shock is more detrimental to Sn-Pb HMC solder joints than are the equivalent number of thermal cycles. Copyright 2012 ASM International® All rights reserved.
The assembly of the BDYE detector requires the attachment of sixteen silicon (Si) processor dice (eight on the top side; eight on the bottom side) onto a low-temperature, co-fired ceramic (LTCC) substrate using 63Sn-37Pb (wt.%, Sn-Pb) in a double-reflow soldering process (nitrogen). There are 132 solder joints per die. The bond pads were gold-platinum-palladium (71Au-26Pt-3Pd, wt.%) thick film layers fired onto the LTCC in a post-process sequence. The pull strength and failure modes provided the quality metrics for the Sn-Pb solder joints. Pull strengths were measured in both the as-fabricated condition and after exposure to thermal cycling (-55/125 C; 15 min hold times; 20 cycles). Extremely low pull strengths--referred to as the low pull strength phenomenon--were observed intermittently throughout the product build, resulting in added program costs, schedule delays, and a long-term reliability concern for the detector. There was no statistically significant correlation between the low pull strength phenomenon and (1) the LTCC 'sub-floor' lot; (2) grit blasting the LTCC surfaces prior to the post-process steps; (3) the post-process parameters; (4) the conductor pad height (thickness); (5) the dice soldering assembly sequence; or (5) the dice pull test sequence. Formation of an intermetallic compound (IMC)/LTCC interface caused by thick film consumption during either the soldering process or by solid-state IMC formation was not directly responsible for the low-strength phenomenon. Metallographic cross sections of solder joints from dice that exhibited the low pull strength behavior, revealed the presence of a reaction layer resulting from an interaction between Sn from the molten Sn-Pb and the glassy phase at the TKN/LTCC interface. The thick film porosity did not contribute, explicitly, to the occurrence of reaction layer. Rather, the process of printing the very thin conductor pads was too sensitive to minor thixotropic changes to ink, which resulted in inconsistent proportions of metal and glassy phase particles present during the subsequent firing process. The consequences were subtle, intermittent changes to the thick film microstructure that gave rise to the reaction layer and, thus, the low pull strength phenomenon. A mitigation strategy would be the use of physical vapor deposition (PVD) techniques to create thin film bond pads; this is multi-chip module, deposited (MCM-D) technology.
The SA1358-10 and SA2052-4 circular JT Type plug connectors are used on a number of nuclear weapons and Joint Test Assembly (JTA) systems. Prototype units were evaluated for the following specific defects associated with the 95Sn-5Sb (Sn-Sb, wt.%) solder joint used to attach the beryllium-copper (BeCu) spring fingers to the aluminum (Al) connector shell: (1) extended cracking within the fillet; (2) remelting of the solder joint during the follow-on, soldering step that attached the EMR adapter ring to the connector shell (and/or soldering the EMR shell to the adapter ring) that used the lower melting temperature 63Sn-37Pb (Sn-Pb) alloy; and (3) spalling of the Cd (Cr) layer overplating layer from the fillet surface. Several pedigrees of connectors were evaluated, which represented older fielded units as well as those assemblies that were recently constructed at Kansas City Plant. The solder joints were evaluated that were in place on connectors made with the current soldering process as well as an alternative induction soldering process for attaching the EMR adapter ring to the shell. Very similar observations were made, which crossed the different pedigrees of parts and processes. The extent of cracking in the top side fillets varied between the different connector samples and likely the EMR adapter ring to the shell. Very similar observations were made, which crossed the different pedigrees of parts and processes. The extent of cracking in the top side fillets varied between the different connector samples and likely reflected the different extents to which the connector was mated to its counterpart assembly. In all cases, the spring finger solder joints on the SA1358-10 connectors were remelted as a result of the subsequent EMR adapter ring attachment process. Spalling of the Cd (Cr) overplating layer was also observed for these connectors, which was a consequence of the remelting activity. On the other hand, the SA2052-4 connector did not exhibit evidence of remelting of the spring finger solder joint. The Cd (Cr) layer did not show signs of spalling. These results suggested that, due to the size of the SA1358-10 connector, any of the former or current soldering processes used to attach the EMR adapter ring and/or EMR shell to the connector shell, requires a level of heat energy that will always result in the remelting of the spring finger solder joint attached with either the Sn-Ag or the Sn-Sb alloy. Lastly, it was construed that the induction soldering process, which is used to attach the EMR adapter ring onto the shell, was more likely to have caused the remelting event rather than the more localized heat source of the hand soldering iron used to attach the EMR shell to the adapter ring.
A study was performed that examined the microstructure and mechanical properties of 63Sn-37Pb (wt.%, Sn-Pb) solder joints made to thick film layers on low-temperature co-fired (LTCC) substrates. The thick film layers were combinations of the Dupont{trademark} 4596 (Au-Pt-Pd) conductor and Dupont{trademark} 5742 (Au) conductor, the latter having been deposited between the 4596 layer and LTCC substrate. Single (1x) and triple (3x) thicknesses of the 4596 layer were evaluated. Three footprint sizes were evaluated of the 5742 thick film. The solder joints exhibited excellent solderability of both the copper (Cu) lead and thick film surface. In all test sample configurations, the 5742 thick film prevented side wall cracking of the vias. The pull strengths were in the range of 3.4-4.0 lbs, which were only slightly lower than historical values for alumina (Al{sub 2}O{sub 3}) substrates. General (qualitative) observations: (a) The pull strength was maximized when the total number of thick film layers was between two and three. Fewer that two layers did not develop as strong of a bond at the thick film/LTCC interface; more than three layers and of increased footprint area, developed higher residual stresses at the thick film/LTCC interface and in the underlying LTCC material that weakened the joint. (b) Minimizing the area of the weaker 4596/LTCC interface (e.g., larger 5742 area) improved pull strength. Specific observations: (a) In the presence of vias and the need for the 3x 4596 thick film, the preferred 4596:5742 ratio was 1.0:0.5. (b) For those LTCC components that require the 3x 4596 layer, but do not have vias, it is preferred to refrain from using the 5742 layer. (c) In the absence of vias, the highest strength was realized with a 1x thick 5742 layer, a 1x thick 4596 layer, and a footprint ratio of 1.0:1.0.