INCREASING PACKAGE ROBUSTNESS WITH PALLADIUM COATED COPPER WIRE

Transcription

1 INCREASING PACKAGE ROBUSTNESS WITH PALLADIUM COATED COPPER WIRE Rodan A. Melanio Regine B. Cervantes Sonny E. Dipasupil New Package Development ON Semiconductor Philippines Incorporated Golden Mile Business Park - Special Economic Zone Governor's Drive, Carmona, Cavite, 4116 Philippines rodan.melanio@onsemi.com, reg.cervantes@onsemi.com, sonny.dipasupil@onsemi.com ABSTRACT Alternative solutions to gold wire are being sought as a for lower cost packages. Palladium coated copper (PCC) wires serve as a potential alternative to increase package robustness and lower package cost particularly in the automotive industry. Palladium on the surface of the wire promotes better adhesion of the wire to the surface where it is bonded. In effect, this may result to elimination of broken wire at heel and lifted ball. This paper discusses the material considerations involved in shifting to PCC wire. In this paper, the benefits of coating bare Cu wire with Pd are discussed. These improvements include enhancements in material properties, wirebond responses, intermetallic compound (IMC) formation and reliability that Pd coating imparts on Cu wire. Palladium coating on Cu wire results to better first and second bond reliability hence, improving package robustness. 1.0 INTRODUCTION For many years, the use of copper (Cu) wire has been adapted to reduce the packaging cost of wire bonded consumer, communication, and computing devices. [1] Since then, the use of Cu wire has matured such that it has been introduced for more demanding applications, including automotive. As the automotive technology advances, the industry requirements also become more stringent. Automotive industries require that incidences of lifted ball and broken wire at heel be eliminated. In IC manufacturing, the emerging approach of both original equipment manufacturers (OEMs) and subcontractors to eliminate these is to use palladium (Pd) coated Cu wire. To make the surface of Pd coated Cu wire softer some manufacturers add gold (Au) flash on the coating. The use of Pd coated Cu wire with Au flash is currently being developed in order to enhance the package robustness of automotive devices. Since this technology is relatively new, responses of the wire are closely being studied. Among the points of interest in the study of Pd coated Cu wires with Au flash is the free air ball (FAB) formed during electronic flame-off (EFO), first and second bond responses before and after reliability testing. Since most suppliers claim to have some special knowledge or manufacturing process that differentiates their Pd coated wire, this study will also determine if such claims are relevant in wire selection Wire Manufacturing Flow The manufacturing flow for Pd coated Cu wire with Au flash differs from the manufacturing flow of bare Cu wire. For wires coated with Pd and Au, an additional step for coating is taken after drawing the wire. This implies that in general, Pd and Au are confined on the Cu wire surface before wire bonding. Regardless of wire manufacturer, the general process flow for Pd coated Cu wire manufacture is the same. The general wire manufacturing sequence is shown below. Fig. 1. General wire manufacturing sequence for Pd coated Cu wire with Au flash. [2] The Role of Pd in Copper Wires

2 Being noble metals, addition of Pd on the surface of Cu wire generally provide resistance to corrosion and oxidation. Addition of Pd and Au increases the floor life of the wire. The introduction of Pd on the coating has been shown to improve process window and reliability. [3] Gold is added on the surface of the wire serves as additional protection for oxidation and in general, it provides additional robustness to the wire. The most important attribute of Au on the Pd coated wire is that Au gives softness on the surface of the wire which aids in the faster bonding to the bond pad surface. [3] Since Au on Pd coated Cu wires is very minute (~<=9 nm), emphasis will be given on Pd that is added as a coating to the Cu wire surface. Cu Au Fig 2. Schematic representation of Pd coated Cu wire with Au flash In terms of wire bond performance, Pd coating improves second bond performance of Cu wire. Better second bond performance may allow for faster second bond processes in some applications and in effect, provide greater throughput relative to bare copper. Better performance is attributed to the confinement of Pd to the surface of the stitch. [3] Even Pd distribution on the surface of the stitch appears to promote better adhesion. Pd on the surface improves reliability because it is less reactive than Cu. This means that when the wire is exposed to ionic contaminants found on the molding compound such as Cl and Br; Pd acts as a barrier that protects Cu from forming salts with these halogens which causes corrosion. The effects of Pd on the first bond are less certain and at present, studies are being undertaken in order to understand it. The methods that were used in this study are presented in the experimental section and the proposed mechanism on how Pd coating can improve first bond reliability is discussed in the next subsection. 2.0 RELATED WORK Evaluation of Pd-Doped Cu Wire Another way of adding Pd to Cu wire is through doping. Previous evaluations have been undertaken for Pd doped Cu wires. Evaluation was carried out using wires from two suppliers, X and Y. Both wires were used for the same Pd package that experienced failure with bare Cu wire. The wires went through the same reliability tests and though X and Y wires failed at different points of the reliability test, both wires experienced failure because of heel crack. Failure due to heel crack implies that since Pd in doped wires is distributed within the wire instead of being on the surface, adhesion of the stitch to the leadframe is most likely similar to that of bare Cu wire. Following the same analogy, Pd coated wire appears to be a better alternative because Pd on the surface of the wire provides better adhesion to the surface to which the wire is bonded Chemical Model Simulation In 2012 Abe and Horie, conducted a chemical model simulation to explore the effect of Pd existence and distribution of Pd in Cu/Al IMC. [4] Since IMC is too thin during wire bonding and analysis of cross section images through scanning electron microscopy (SEM) would yield very little results, chemical modeling aided in predicting the type of IMC formed and how elements diffuse. Modeling showed that dispersed Pd contributed to create new IMC of Cu/Al/Pd instead of easily corroded Cu rich Cu/Al IMC. Cu, Al, and also Cl ion diffusion were inhibited by Pd at surface. Fig 3. Chemical Modelling by Abe and Horie [4] Results of the chemical modeling simulation conducted by Abe and Horie will be related to experimental builds. The next section explain the methods used in this study. 3.0 EXPERIMENTAL SECTION For this study, three Pd coated Cu wires were initially considered. The wires are from three wire manufacturers, A, B, and C. The wires are all 0.8 mil in diameter with a purity of 4N. Their material properties will be compared with each other and then compared with the material properties of the bare Cu wire qualified at ON Semiconductor Philippines, Inc. (OSPI).

3 3.1.0 Material Selection and Free Air Ball (FAB) Evaluation Studies FAB evaluation was done for the three wires. The FABs were formed using the same EFO current settings. For this paper, three EFO current settings will be considered, 108 ma, 70 ma, and 40 ma. The bonding machine automatically calculates the EFO time for a given current setting to achieve the targeted FAB size for a given wire diameter. The summary of parameter setting combinations is found on the succeeding table. For this study, the FAB were formed using the same forming gas (95%N 2-5%H 2) Table 1. Parameter Setting Combinations Expt EFO Current (ma) EFO Time (µs) The FABs formed were evaluated in several ways. Visual inspection was done on 80 balls per parameter. Visual inspection was done using low magnification to detect deformed FABs. Scanning electron microscope (SEM) evaluation was done on the deformed FABs in order to identify the cause for the deformations found. Pd coverage on the surface of the FAB will be evaluated. Evaluation of Pd coverage on the FAB surface was done using SEM/EDX. Pd coverage on the FAB surface was done on several points in order to determine how evenly Pd is distributed on the surface of each FAB Wire Bond Responses extreme conditions that the IC may be subjected to during operation Corrosion Simulation Studies It is necessary to test how PCC wires will compare to bare Cu wire. In conducting this part of the study, units were bonded and encapsulated using epoxy molding compound (EMC) especially designed to simulate extreme conditions. Formulation of current qualified EMC material at OSPI was altered such that ppm Cl was higher and ph was lower than the specifications limit. Molded units were studied and subjected to reliability tests. Temperature cycling (TC), unbiased HAST (uhast), HAST and high temperature storage life (HTSL) were carried out until samples failed. Failing units will be subjected to analysis. 4.0 RESULTS AND DISCUSSION Material Selection and Free Air Ball (FAB) Evaluation Studies Table 2 shows the important material properties of each wire. These material properties were taken from manufacturer data. In terms of Pd and Au content on the Cu wire, it can be seen that the regardless of manufacturer, Pd and Au content on the wire is almost similar. This implies that the wire manufacturing process for all suppliers is similar. The column highlighted in blue shows the existing qualified Cu wire at OSPI. Table 2. Material Properties of Cu Wire After evaluation of FABs formed using the three different PCC wires, one wire material will be considered in this study and used for the qualification builds and then subjected to full reliability testing. Wire material selection was based on which material had the best results in terms of FAB formation and material properties provided by each supplier. Wire bonding responses particularly, ball shear, wire pull, stitch pull, pad metal displacement (PMD) and cratering results were compared with those of bare Cu wire Reliability Responses Reliability testing was done on the qualification builds. The samples were observed using transmission electron microscope (TEM) imaging. In this paper, the intermetallic formed by both PCC and bare Cu wire after highly accelerated stress test (HAST) 192h, temperature cycling (TC) 2000 tests were observed. These tests simulate the Comparing the wires with one another, it can be seen that some of the properties of the wires are similar. Density and melting points of all wires being studied are similar to each other. Their similarities are due to the dominance of Cu in

4 the wire. Because Cu is the dominant element on the wire, the intrinsic properties of the wires will most likely be close to those of Cu. Some of the properties of bare Cu wire have been altered by the presence of Pd. Among them are FAB hardness, wire hardness, electrical resistivity and floor life. The floor life of Pd coated Cu wire is longer than that of bare Cu wire except for the wire from supplier A. This means that in general, Pd on the surface of the wire provides better resistance against oxidation. The presence of Pd on the coated wires increased the resistivity of bare Cu wire. It is not yet certain whether the increased resistivity will have a significant effect on the flow of current from the pad to the lead. Wire from supplier B has the hardest wire among all Pd coated wires. When incorporated into materials, Pd contributes to the hardness of the material. The other Pd wires may not be as hard because of the presence of Au on the surface. Au when integrated into materials provides softness which appears to outweigh the hardness given by Pd for the other two Pd coated Cu wires. The relationship of the presence of Pd on the FAB deformation can be seen in the succeeding paragraph. Eighty samples were inspected for FAB deformation. Inspection of any visual defect such as abnormal FAB shape, dimple, or oxidation was performed and results are in Figure 4. As shown below, no deformation was found for supplier A wire for all parameter sets. Deformation on supplier B wire was found however, when inspected under SEM, it was found that the deformations were scratches from the interaction of the FAB with the leadframe and not during FAB formation. For all parameter sets, deformation during EFO was found on supplier C wires. of bare Cu wire. This may be due to a couple of factors. One may be that the parameter settings used in this study are not applicable for their wire. Another possibility is that supplier C material may perform better with a different forming gas ratio. In terms of FAB visual examination, both wires from suppliers A and B were free from deformation during FAB formation. Although the FAB of supplier C has the most number of deformations, possible problems may arise due to the hardness of FAB from using supplier B PCC wire. The harder the wire, the more likely cratering will occur. [5] Supplier B has the hardest FAB ( Hv) among the three wires and it is even harder than the existing qualified bare Cu wire in OSPI (85-95Hv) Wire Bond Responses From initial validation of material properties, and FAB responses, supplier B PCC material was considered for the rest of this study. Figure 5 below shows that visually, flower shaped balls were encountered for bare Cu wire but under the same first bond parameters, the ball bonds of PCC wire had a consistent, circular shape. For the same parameter sets, the first bond of PCC wire had a better adhesion on the bond pad. This correlates with the ball shear responses on the succeeding figure. More force was required to shear the ball formed w PCC wire than the force required for bare Cu. Fig. 4 FAB Visual Inspection per Parameter Set Deformation on FAB of supplier C wires occurred because the FAB hardness of the wire (40-70Hv) is very much less than the FAB hardness of the other Pd coated wires and that

5 Fig.7 Stitch Comparison between Wire Types Fig.5. First Bond Responses of Cu Wire vs PCC Wire At time zero, staining was done on cross sectioned first bond to determine where Pd is initially concentrated. Units were stained using a combination of buffered oxide etch (BOE) and aluminum etchant. This combination was used in order to etch copper at a controlled rate such that the Pd layer will remain. Based on secondary electron (SE) and backscatter electron (BSE) images, it can be seen that at time zero, Pd is distributed inside the ball bond. This can be seen from the bright areas of the stained samples. Fig.8. Wire Pull Readings of Cu Wire vs PCC Wire Fig.6 Bonded Ball Pd Distribution of PCC Wire Similarly, second bond responses were also studied. In the succeeding figures it can be seen that wire pull (WPT) and stitch pull (SPT) readings show that the second bond of PCC wires are significantly better than Cu wire. This means that Pd coating improves the adhesion between the wire and the leadframe. Improved adhesion means better reliability because it prevents broken wire at heel. Further discussion and SEM images to support this can be seen on the next section. Fig.9 Stitch Pull Readings of Cu Wire vs PCC Wire At time zero, it can be seen that PCC wire is more robust than bare Cu. Responses after stress are discussed in the next section Reliability Responses Full reliability testing was conducted for PCC qual builds as well as bare Cu wire. Some of the tests that the units underwent were highly accelerated stress test (HAST) which

6 evaluates reliability of the package in humid environments, unbiased HAST (uhast) which evaluates moisture resistance of the package, temperature cycling (TC) which determines the resistance of the package to high/ low temperature changes and high temperature storage life (HTSL), which tests for charge gain/ loss and metallurgical reactions. [6] Both PCC and bare Cu qual builds passed all qualification points based on automotive requirements. To determine what sets PCC apart from bare Cu, transmission electron microscope (TEM) images and EDS mapping were carried out to determine the IMC that formed after HAST 384. Analysis of the IMC reveals that PCC wire suppressed the formation of Cu-rich IMC (Cu 3Al 2). Corrosion studies were conducted to further differentiate PCC from bare Cu wire. At the time this paper was written, both PCC and bare Cu still passed extended reliability testing. Even though the units were electrically good, potential corrosion sites were observed on bare Cu units at high ppm Cl EMC that were observed. SEM images showed that PCC wire did not have any IMC cracks until ppm level of Cl was beyond automotive electronics council (AEC) standards of 20 ppm. This finding is consistent with other data presented in this paper that PCC is more robust than bare Cu. Fig. 11 SEM Images of Bare Cu and PCC IMC after HTSL 1008 using variable ppm Cl EMC, taken at 40000x magnification Fig. 10 TEM Images of Bare Cu (top) wire vs PCC wire (bottom) after HAST Pd coating on Cu wire suppressed the formation of Cu-rich IMC by forming a solid solution with Cu. This agrees with the chemical modeling study conducted by Abe and Horie [4] discussed above. The TEM images revealed lattice imperfections on Cu 3Al 2 that are potential corrosion sites. These sites can be attacked by Cl ions as well as Br, F and S found on EMC. Because Pd coating forms a solid solution with Cu, Cu 3Al 2 formation was hindered. In this manner, Pd coating increases the robustness of Cu wire. Since both PCC and bare Cu wires have passed automotive reliability requirements, corrosion simulation studies have been carried out until such time that further differentiation can be made between PCC and bare Cu Corrosion Simulation Studies 5.0 CONCLUSION Palladium coating improves both first and second bond reliability of Cu wire thereby, increasing package robustness. On the first bond, Pd coating suppressed the formation of Cu-rich IMC that can potentially be attacked by ions found in EMC. Palladium coating also improves the second bond by increasing the adhesion between the wire and leadframe. 6.0 RECOMMENDATION Palladium coated Cu wire should be tested with sensitive die technologies. Also, extended HAST should be carried out to further study how PCC would be able to withstand corrosion as compared with bare Cu wire. 7.0 ACKNOWLEDGMENT The authors would like to thank:

7 Wentao Qin and Denisse Barrientos For providing TEM Analysis for this study 8.0 REFERENCES [1] Carson, F., Copper Wire Interconnect Has Arrived, ChipScale Review, Vol 15, No. 1 (Jan-Feb 2011), pp [2] Tanaka. (2013). Copper Wire Data Sheet. [3] Clauberg, H. et al, Wire Bonding with Pd-coated Copper Wire, CMPT Symposium Japan, 2010 IEEE, Tokyo, Japan, Aug. 2010, pp [4] Abe, H. and Horei, T. Cu Wire and Pd-Cu Wire Package Reliability and Molding Compounds, IEEE, [5] Harman, G. Wire Bonding in Microelectronics. McGraw-Hill, USA, pp [6] Kittle, C. Introduction to Solid State Physics: 8 th Edition. 9.0 ABOUT THE AUTHORS Rodan A. Melanio is a Principal Engineer of ON Semiconductor New Package Development Group. He has more than 19 years extensive experience both Equipment and Copper Wirebond Engineering gained from Semiconductor and Electronics Manufacturing Firm. He is a graduate of Electronics and Communication Engineering from Mapua Institute of Technology. Regine B. Cervantes is a Package Development Engineer at ON Semiconductor. She is currently doing materials analysis with focus on new materials. She has a Bachelor of Science in Chemical Engineering from Ateneo de Davao University. She is currently obtaining her Master s Degree in Materials Science and Engineering at Mapua Institute of Technology. Sonny E Dipasupil is a Principal Engineer of ON Semiconductor New Package Development Group. He has more than 15 years of extensive experience for both Equipment and Process Engineering wherein he gains his expertise in Wirebonding. He is a graduate of Electrical Engineering from Technological University of the Philippines-Visayas.