Microelectronics Reliability
Volume 47, Issue 12,
December 2007
, Pages 2152-2160
Author links open overlay panel, , , ,
Abstract
This paper presents a study of the optimization of the gold plating thickness for the use of both wire bonding and soldered interconnects on a flexible printed circuit board sample module. Wire bondability is typically better, when the gold plating thickness is greater than 30μin.; however, the risk of problems with solder joint embrittlement becomes a concern with thick gold plating. In order to better understand the effect of the gold plating thickness on wire bondability and solder joint embrittlement, an evaluation was performed on samples with three ranges of gold plating thicknesses (10–20μin., 20–30μin., and 30–45μin.), on flexible printed circuit board (PCB), substrates. Mechanical shear testing and metallurgical analyses were conducted on chip component solder joints in this three thickness gold study. Thermal shock and drop testing were conducted to evaluate the reliability of the sample modules. Drop testing is especially critical for determining the reliability of the sample modules, which are used in portable consumer electronics products. Reliability testing and metallurgical analyses have been performed to characterize the effect of gold embrittlement on the mechanical integrity of the solder joints with a gold content ranging from 1 to 4wt.%.
Introduction
Gold has a long history of applications in the electronics industry and is commonly used as a mechanical contact finish, a wire bondable surface finish, and a solderable coating. The application of gold as a solderable coating has been a concern because of the impact of gold on solderability and the effects of gold on the mechanical and metallurgical properties of the solder joints. Much of the concern is based on a history of problems with gold embrittlement that were experienced in the past, when the thicknesses of the gold surface finishes were much greater than in more current applications. More recently, gold embrittlement has caused missing ball failures in ball grid array packages (BGA), and premature fatigue failure problems in newer technology BGA packages [1], [2], [3], [4], [2].
Gold plating is very versatile and its properties vary greatly depending on the plating process conditions. When gold is used as a mechanical contact surface finish, harder, fine-grained deposits are formed by the use of hardening elements, grain refining elements and brighteners, which produce smooth, shiny, and generally thinner coatings. Gold deposits with few or no additives produce rougher, duller, matte finishes, which are desirable for wire bonding and soldering because they are metallurgically soft and pure.
Solder forms intermetallic compounds with the various metals in the component and printed circuit board, (PCB) solder pads, which is a requirement for a good solder joint. The metallurgy of the solder and the surface finish of the solder pads play an important part in the reliability and the failure modes of the solder joints. In this paper, electrolytic gold–nickel PCB surface finish, tin–lead eutectic solder, and tin–silver–copper lead free solder are evaluated in solder joints to chip components and gull wing style connector leads. When solder is reflowed in contact with a gold–nickel surface finish, the gold quickly dissolves into the solder and forms a distribution of gold, nickel, and tin bearing intermetallic phases which are brittle structures in the solder. A loss of ductility of the solder joint, or so-called gold embrittlement can occur if the gold concentration in the solder joints is too high. Gold embrittlement can lead to a significant reduction in the fatigue life and the mechanical integrity of the solder joints.
It is perceived that gold embrittlement is a concern in solder joints with greater than 3% gold [5]. Zhong gives a range of gold content greater than 3–5% [2]. Although there are several reports that gold embrittlement can occur in at a much lower percentage in a ball grid array (BGA), packages [6], [7], [8]. A rule of thumb is that gold embrittlement is not a problem as long as the thickness is kept below 15μin. Fortunately, most current applications have very thin gold plating in the range 3–15μin. The emergence of improved plating processes that produce dense, low porosity plating along with the reduction in storage time between plating and assembly have allowed for adequate solderability with relatively thin gold surface finishes. The key to reducing the extent of gold embrittlement in solder joints is keeping the gold thickness to a minimum.
There is a trade-off between maintaining a low enough gold thickness to prevent gold embrittlement, and a high enough gold thickness for robust wire bondability, because wire bondability is typically better with thicker gold pads for either gold or aluminum bonding wire. Some electronics assembly manufacturers prefer to use thinner gold to minimize the risk of gold embrittlement problems, while others prefer thicker gold plating, which affords a longer shelf life and more robust wire bonding processes.
In this study with wire bonding to electrolytically plated flex circuits, the first bond is made by ball bonding gold wire to an aluminum pad on the die and then stitch bonding to the flex circuit substrate pad. The process is generally robust, yet if a problem occurs, it is usually related to either a problem with the plating metallurgy or surface contamination. Cross-contamination from a previous soldering process or a die attach process can cause problems with the wire bond process. Wire bonding requires clean and metallurgically soft wire bond pads. The use of thicker gold allows for the use of more aggressive wire bonding conditions for scrubbing through surface contaminants. Wire bond pull testing is commonly used to test wire bonds; however, the effectiveness of wire pull testing for detecting wire bonding problems in the manufacturing process and screening good and bad lots of printed circuit boards has been debated [9], [3].
Gold plating thickness requirements vary widely from customer to customer, yet many customers require a minimum of 30–40μin. There is little agreement on the optimum gold thickness for either wire bonding or for making solder joints in microelectronics packages. Holcomb concluded that 24μin. is a reasonable lower limit for a well-controlled electrolytic gold process on organic substrates in BGA packages [9]. Plasma cleaning removes adventitious carbon and inter diffused nickel oxides from the surface of the gold, and allows for good wire bondability with thinner gold plating. The success of a wire bonding process is very dependent on the quality of the plating and the manufacturing process conditions. For example, wire bonding to a flexible printed circuit board on our sample modules is much more difficult than to a die on a lead frame, because the lead frame can be heated to a much higher temperature during wire bonding.
Section snippets
Reliability testing
Air to air thermal shocks tests were performed from −40 to 85°C, with 30min dwell times at each temperature. The sample modules were functionally tested at intervals of 250 thermal shocks. Hot storage testing might have been a better indicator of problems with gold embrittlement than thermal shock; however, the optics in our sample modules become damaged at the temperatures typically used for hot storage tests, so hot storage was impractical [2], [3], [4], [2], [10]. Hot storage testing is also
Three thickness gold study with tin–lead solder
This study was conducted on sample modules with flexible printed circuit boards (PCB’s), with electrolytically plated gold–nickel pads and FR4 dielectric, Fig. 1.
The sample modules were assembled using standard assembly and wire bonding practices using a 63% tin–37% lead solder alloy. The gold thicknesses were 10–20μin., 20–30μin. and 30–45μin., and were measured by X-ray fluorescence. The thickness of the nickel barrier layer was approximately 80–120μin. Theoretical values for the gold content
Conclusions
In the study with tin–lead solder, the results of the theoretical percentage gold calculations, capacitor shear testing, metallurgical studies, and reliability testing did not allow for a conclusive determination of whether gold embrittlement was a reliability problem in sample modules with thick gold plating (30–45μin.). The modules passed thermal shock testing, and exhibited capacitor shear test failure modes that consisted of mostly PCB pad lifts and end-terminal metallization failures
References (25)
- D.T. Rooney et al.
Evaluation of Wire bonding Performance, process conditions, and metallurgical integrity of chip on board wire bonds
Microelectron Reliab
(2005)
- M. Mita et al.
Effect of Ni on reactive diffusion between Au and Sn at solid state temperatures
Mater Sci Eng B – Solid State Mater Adv Technol
(2006)
- Zhong CH, Yi S, Mui YC, Howe CP, Olsen D, Chen WT. Missing solder ball failure mechanisms in plastic ball grid array...
- Zhong CH, Sung Yi. Effects of ball pad metallurgy and ball composition on solder ball integrity of plastic ball grid...
- Glazer J, Kramer PA, Morris JW. Effect of Au on the reliability of fine pitch surface mount solder joints. In:...
- Abbott D, Romm D, Lange B. A nickel palladium gold integrated circuit lead finish and its potential for solder joint...
- W. Peng et al.
Microstructural and performance implications of gold in Sn–Ag–Cu–Sb interconnections
- J. Zribi
IEEE Trans Compon Pack Technol
(2000)
Cited by (4)
Multi-objective optimization on laser solder jet bonding process in head gimbal assembly using the response surface methodology
2018, Optics and Laser Technology
Citation Excerpt :
Infrared heater is commonly used as the heat source in reflow oven. A literature on the mechanical properties of solder joints in the component assembly on flexible printed circuit board (FPCB) was proposed by [1]. Reflow process, mechanical shear test and metallurgical analysis of solder joint were reported in their work.
This paper aims to investigate the effect of laser solder jet bonding parameters to the solder joints in Head Gimbal Assembly. Laser solder jet bonding utilizes the fiber laser to melt solder ball in capillary. The molten solder is transferred to two bonding pads by nitrogen gas. The response surface methodology have been used to investigate the effects of laser energy, wait time, nitrogen gas pressure, and focal position on the shear strength of solder joints and the change of pitch static attitude (PSA). The response surface methodology is employed to establish the reliable mathematical relationships between the laser soldering parameters and desired responses. Then, multi-objective optimization is conducted to determine the optimal process parameters that can enhance the joint shear strength and minimize the change of PSA. The validation test confirms that the predicted value has good agreement with the actual value.
Effect of nickel metallization thickness on microstructure evolution and mechanical properties in Sn3.0Ag0.5Cu/Au/Ni/Cu solder joints
2020, Journal of Materials Science: Materials in Electronics
A preliminary study of COB assembly technology for SOI pixel detector
2015, Hedianzixue Yu Tance Jishu/Nuclear Electronics and Detection Technology
Failure analysis on electrolytic Ni/Au surface finish of PCB used for wire bonding and soldering
2014, Soldering and Surface Mount Technology
Recommended articles (6)
Research article
The effect of the hydrophilic spacer length on the functionality of a mercury-supported tethered bilayer lipid membrane
Bioelectrochemistry, Volume 101, 2015, pp. 92-96
A biomimetic membrane consisting of a thiolipid monolayer tethered to a mercury electrode, with a dioleoylphosphatidylcholine (DOPC) monolayer on top of it, was fabricated. The thiolipid, referred to as DPOL, consisted of an octaethyleneoxy (OEO) chain terminated at one end with a lipoic acid residue and covalently linked at the other end to two phytanyl chains. The functionality of this biomimetic membrane, referred to as a tethered bilayer lipid membrane (tBLM), was tested by incorporating gramicidin and alamethicin and verifying their ion channel activity. Advantages and drawbacks with respect to a tBLM using a thiolipid, referred to as DPTL, with a tetraethyleneoxy (TEO) chain were examined by using electrochemical impedance spectroscopy, potential-step chronocoulometry and cyclic voltammetry. The maximum charge surface density of potassium ions stored in the OEO spacer amounts to 70μCcm−2, as compared to a charge surface density of 45μCcm−2 in the TEO spacer. The lipid bilayer moiety of the DPOL/DOPC tBLM is somewhat leakier than that of the DPTL/DOPC tBLM at potentials negative of about −0.65V vs. the saturated calomel electrode. The estimated value of the surface dipole potential of the OEO spacer amounts to −0.180V and is, therefore, smaller than that, −0.230V, of the TEO spacer.
Research article
Preparation of hybrid films containing gold nanoparticles and cobalt porphyrin with flexible electrochemical properties
Thin Solid Films, Volume 545, 2013, pp. 327-331
We report on the fabrication of hybrid films of gold nanoparticles (AuNPs) and [tetrakis (N-methylpyridyl)porphyrinato] cobalt (CoTMPyP) based on electrostatic layer-by-layer (LBL) self-assembly technique. A three-dimensional nanostructured film composed of positively charged CoTMPyP and negatively charged poly(acrylic acid)-stabilized AuNPs can be formed onto various substrates, including 4-aminobenzoic acid modified glassy carbon electrode, 11-mercaptotroponic acid modified gold film, and quartz plate with negatively charged surface. The regular growth of hybrid films is investigated by UV–vis spectroscopy, surface plasma resonance, and impedance spectroscopy. The prepared multilayer films possess flexible electro-catalytic activity toward O2 reduction reaction. It is noted that the electro-catalytic capability of the prepared multilayer film decreases as bilayer number increases. The unexpected observation of catalytic activity with bilayer number is investigated in detail.
Research article
Indentation size effect of stress exponent and hardness in hom*ogeneous duplex eutectic 80Au/20Sn
Materials Letters, Volume 120, 2014, pp. 151-154
The indentation size effect (ISE) of creep stress exponent and hardness in hom*ogeneous duplex eutectic 80Au/20Sn was studied by nanoindentation, using constant load/holding (CL) and continuous multi-cycle (CMC) loading modes, respectively. The stress exponent of each fixed load decreases first and then increases as the indentation depth increases. The stress exponent for the same strain rate range increases with the indentation depth increasing. The result of CMC indicates that Nix-Gao's overestimates the hardness of 80Au/20Sn for small indentations. A revised model incorporating enlarged effective plastic zone and the maximum geometrically necessary dislocation density was achieved to describe the relationship between hardness and indentation depth.
Research article
Measurements of surface concentration and charge number per adsorbed molecule for a thiolipid monolayer tethered to the Au(111) surface by a long hydrophilic chain
Journal of Electroanalytical Chemistry, Volume 793, 2017, pp. 203-208
The charge number per adsorbed molecule and surface concentration of 2,3-di-O-phytanyl-sn-glycero-1-octaethyleneglycol-D,L-α-lipoic acid ester (DPOL) at a Au(111) electrode surface was investigated by chronocoulometry using two different film deposition methods. First, a known amount of the thiolipid was transferred from the air-water interface of a Langmuir trough onto the gold electrode surface via a Langmuir-Blodgett (LB) deposition. The charge density measurements for this monolayer system were used to determine the charge number per adsorbed DPOL molecule (electrosorption valency). The observed electrosorption valency values of the adsorbed DPOL film were much lower than the expected number of transferred electrons for a simple reductive desorption process. In the second deposition method, charge densities were measured for the electrode covered by a self-assembled DPOL monolayer. The electrosorption valency values determined from the LB DPOL film were used to calculate the packing density of the DPOL molecules within the self-assembled monolayer. The surface concentration of the molecules within the thiol monolayer with octaethylene glycol chains gave similar results to a related thiolipid with tetraethylene glycol chains (DPTL). This new finding indicates that both of these molecules (DPTL and DPOL) assume a brush conformation in densely packed self-assembled monolayers.
Research article
Patterning Gradients
Methods in Cell Biology, Volume 119, 2014, pp. 91-121
Surface-chemical and -morphological gradients can be extremely useful in cell-biological research as high-throughput screening tools—for example, exposing a given set of cells to many different surface conditions at once, under identical ambient conditions, in order to monitor cell behavior such as proliferation or specific gene expression. They can also be used to investigate the effects of gradients themselves on cell behavior, such as migration. A number of simple, reliable techniques for both chemical- and morphological-gradient fabrication have been developed in our laboratories and are described in detail in the following.
Research article
Electro-elastic fields around two arbitrarily-shaped holes in a finite electrostrictive solid
Applied Mathematical Modelling, Volume 40, Issues 7–8, 2016, pp. 4625-4639
Stress and electric field concentration around two arbitrarily-shaped holes in a finite electrostrictive solid subjected to uniform remote electric field are studied. The edge of the solid and the shapes of the holes are defined via several conformal mappings. The electro-elastic field of the solid as well as the electric fields both outside the solid and inside the holes are obtained on using conformal mapping techniques, Faber series and Fourier expansion method. Extensive numerical results are shown for elliptical, triangular, square, oval and rectangular holes in a square electrostrictive plane. The exact external electric field near the edge of the solid is found to be essentially non-uniform and its maximum appears to be much larger than the uniform remote electric field, while it is shown that the hoop stress nearby the point of maximum curvature on the hole's boundary increases rapidly as the point of maximum curvature approaches either the edge of the solid or the other hole. On the other hand, for a single centrally-located hole, the finite electrostrictive plane can be approximately treated as an infinite plane if the size of the finite plane is not less than six times the size of the hole, while for two interacting holes, the interaction between them is negligible when the distance between the two holes is larger than three times the size of the holes.
Copyright © 2007 Published by Elsevier Ltd.