Anal and Preven. (2013)13: 194-201 DOI10.1007/sll668-013-96585 TECHNICAL ARTICLE-PEER-REVIEWED Failure Analysis of Un-Wetting for the Surface Finish on the enig Shi Yan. Fei-Jun Chen. Yue-Yue Ma Zhen-Guo Yang Submitted: 12 November 2012/in revised form: 7 January 2013/Published online: 6 February 2013 C ASM International 2013 Abstract Electroless nickel immersion gold (ENIG) together and provide the interconnection between them of the most prevailing surface finishes for printed The surface finish of PCB is one of the key factors which board. It is widely adopted by manufacturers all over the has a significant influence on its reliability performance. As world for its relative low price and compliance with the trend illustrated in Fig. l, surface is the bonding bridge of die of high density packaging. However, its reliability is always and substrate. Nowadays, electroless nickel immersion of a concern. In this paper, the poor wetting performance of gold(ENIG) is the most popular surface finish adopted by tin solder on the ENG surface, which is one of its reliability nearly all the PCB manufacturers around the world [3]. The issues was addressed. A series of modern facilities were primary process of ENIG is to deposit a layer of nickel and utilized for finding the failure mechanism and chasing down then a layer of gold onto the PCB substrate, as exhibited in the root cause. It was concluded that the inferior quality of formulae (1)and(2). According to IPC-4552 standard, the immersion gold layer during plating, to be more specifically, nickel layer should be between 3- and 6-um thick and the coarse and big grain size and the crystalline interface golden layer be 50-100-nm thick to ensure its reliability cracking were the main causes for the un-wetting failure. The specific thickness of ENIG should be determined by Last but not least, the improper choice of pure tin solder special circumstance in which the product will be applied should also contribute to the final failure The golden layer is employed for maintaining the solder ability during manufacturing process like reflow and wave Keywords ENIG. Un-wetting. Grain size soldering. In some situation, such as Golden Finger, the Failure analysis. Oxidation golden layer is utilized for providing anti-corrosion and abrasion resistance property. In another application, man- ufacturer will bond aluminum or golden wires onto the Introduction golden surface to realize the electric interconnection between the pcb and outer circuit. nickel is used as a Printed circuit board(PCB)is the foundation of all kinds of barrier preventing atoms in the substrate material from electronic devices ranging from cell phone, personal diffusing into the golden layer and forming intermetallic computer to military products. With the booming devel- compound(IMC) opment of integrated circuits (C), there is increasingly Ni2++2H,PO +2H20- Ni+2HPO3-+4H*+H2 t stringent requirements placed on reliability of PCB [1, 2] since it is the platform which connects all the chips 2Au(CN)2+Ni- 2Au+Ni++ 4CN S. Yan(), F-J. Chen.Y.Y Ma. Z.-G. Y Materials Science, Fudan University, Shanghai In this paper, a batch of PCB sample finished with ENIG Departme 200433, was found to display poor wetting performance after hot air e-mail:11210300039@fudan.edu.cn solder leveling. For the purpose of z.一G.Yang mechanism, locating the root cause and finally improving zgyang@fudan.edu.cn reliability, a series of modern analytical instruments such
TECHNICAL ARTICLE—PEER-REVIEWED Failure Analysis of Un-Wetting for the Surface Finish on the ENIG Shi Yan • Fei-Jun Chen • Yue-Yue Ma • Zhen-Guo Yang Submitted: 12 November 2012 / in revised form: 7 January 2013 / Published online: 6 February 2013 ASM International 2013 Abstract Electroless nickel immersion gold (ENIG) is one of the most prevailing surface finishes for printed circuit board. It is widely adopted by manufacturers all over the world for its relative low price and compliance with the trend of high density packaging. However, its reliability is always of a concern. In this paper, the poor wetting performance of tin solder on the ENIG surface, which is one of its reliability issues was addressed. A series of modern facilities were utilized for finding the failure mechanism and chasing down the root cause. It was concluded that the inferior quality of immersion gold layer during plating, to be more specifically, the coarse and big grain size and the crystalline interface cracking were the main causes for the un-wetting failure. Last but not least, the improper choice of pure tin solder should also contribute to the final failure. Keywords ENIG Un-wetting Grain size Failure analysis Oxidation Introduction Printed circuit board (PCB) is the foundation of all kinds of electronic devices ranging from cell phone, personal computer to military products. With the booming development of integrated circuits (IC), there is increasingly stringent requirements placed on reliability of PCB [1, 2] since it is the platform which connects all the chips together and provide the interconnection between them. The surface finish of PCB is one of the key factors which has a significant influence on its reliability performance. As illustrated in Fig. 1, surface is the bonding bridge of die and substrate. Nowadays, electroless nickel immersion gold (ENIG) is the most popular surface finish adopted by nearly all the PCB manufacturers around the world [3]. The primary process of ENIG is to deposit a layer of nickel and then a layer of gold onto the PCB substrate, as exhibited in formulae (1) and (2). According to IPC-4552 standard, the nickel layer should be between 3- and 6-lm thick and golden layer be 50–100-nm thick to ensure its reliability. The specific thickness of ENIG should be determined by special circumstance in which the product will be applied. The golden layer is employed for maintaining the solder ability during manufacturing process like reflow and wave soldering. In some situation, such as Golden Finger, the golden layer is utilized for providing anti-corrosion and abrasion resistance property. In another application, manufacturer will bond aluminum or golden wires onto the golden surface to realize the electric interconnection between the PCB and outer circuit. Nickel is used as a barrier preventing atoms in the substrate material from diffusing into the golden layer and forming intermetallic compound (IMC). Ni2þ þ 2H2PO 2 þ 2H2O ! Ni þ 2HPO2 3 þ 4Hþ þ H2 " ðEq 1Þ 2Au CN ð Þ 2 þNi ! 2Au þ Ni2þ þ 4CN ðEq 2Þ In this paper, a batch of PCB sample finished with ENIG was found to display poor wetting performance after hot air solder leveling. For the purpose of finding the failure mechanism, locating the root cause and finally improving reliability, a series of modern analytical instruments such S. Yan (&) F.-J. Chen Y.-Y. Ma Z.-G. Yang Department of Materials Science, Fudan University, Shanghai 200433, China e-mail: 11210300039@fudan.edu.cn Z.-G. Yang e-mail: zgyang@fudan.edu.cn 123 J Fail. Anal. and Preven. (2013) 13:194–201 DOI 10.1007/s11668-013-9658-5
J Fail. Anal. and Preven. (2013)13: 194-201 Fig. 1 The role of surface finish in PCB Resin Substrate sooderobum(pes Copp SoodeOOooo( Surface finish Resin Substrate as 3D stereo microscopy, scanning electron micros pe One well-wetted pad was also analyzed by EDs for con- (SEM), and energy dispersive spectroscopy(EDS) trast and comparison adopted. The un-wetting mechanism and the root cause of he failure were illustrated. Counter strategies and sug- gestions were given in the end Results and discussion Sample and Experiment A section of failed pad indicated by the red square Fig. 2a was chosen for SEM inspection. The reason for Figure 2a is the plane configuration of the failed PCB choosing this section is that it contains both the well-wetted which has a dimension of 1.2 cm x 2.4 cm. The sample and de-wetted pads, thus the contrast is distinct was treated with ENIG surface finish, then desmeared, and As Fig 3a shows, although the normal and failed pads subsequently went through hot air solder leveling. The little all appear gray under SEM, there is still observable dif- squares on sample are pads where tin solder should be ference between the two parts. The normal(well-wetted) attached on. Under optical inspection, as Fig. 2a shows, pad displays a bulgy surface topography, while the failed some pads are emitting shiny silver light and these are the ( de-wetted) pad displays completely fatness. This is ones soldered appropriately. While some of the pads just because tin solder will form a bump on pad under the effect appear golden, and these are the ones not properly wetted of gravity and capillary force, rendering the normal pad a by tin solder, thus letting the underlying golden layer bulgy profile. In a similar way, since the de-wetted pad has our naked eye. Figure 2b is the corresponding schematic the final finish of ENIG which accounts for its flatness exposed to the environment, displaying a color of golden to no solder attached on it, its outermost surface is actuall diagram of the failed sample which clearly shows that there The failed pad which is indicated by the red circle are 8 out of 129 pad un-wetted which is not endurable. Fig 3a was further magnified. As it can be seen in Fig. 3b. In order to dig out the primary cause of the failure, the grains of the immersion gold on the failed pad are cle modern analytic instruments and characterization methods and it appears coarse under magnification of 2,000x. This have been adopted. The de-wetting pad was observed by feature became more obvious when the spot was magnified SEM, and its chemical composition was analyzed by EDS. to 4,000X(Fig. 3c). When magnified to 16,000X, besides Spring
as 3D stereo microscopy, scanning electron microscope (SEM), and energy dispersive spectroscopy (EDS) were adopted. The un-wetting mechanism and the root cause of the failure were illustrated. Counter strategies and suggestions were given in the end. Sample and Experiment Figure 2a is the plane configuration of the failed PCB which has a dimension of 1.2 cm 9 2.4 cm. The sample was treated with ENIG surface finish, then desmeared, and subsequently went through hot air solder leveling. The little squares on sample are pads where tin solder should be attached on. Under optical inspection, as Fig. 2a shows, some pads are emitting shiny silver light and these are the ones soldered appropriately. While some of the pads just appear golden, and these are the ones not properly wetted by tin solder, thus letting the underlying golden layer exposed to the environment, displaying a color of golden to our naked eye. Figure 2b is the corresponding schematic diagram of the failed sample which clearly shows that there are 8 out of 129 pad un-wetted which is not endurable. In order to dig out the primary cause of the failure, modern analytic instruments and characterization methods have been adopted. The de-wetting pad was observed by SEM, and its chemical composition was analyzed by EDS. One well-wetted pad was also analyzed by EDS for contrast and comparison. Results and Discussion A section of failed pad indicated by the red square in Fig. 2a was chosen for SEM inspection. The reason for choosing this section is that it contains both the well-wetted and de-wetted pads, thus the contrast is distinct. As Fig. 3a shows, although the normal and failed pads all appear gray under SEM, there is still observable difference between the two parts. The normal (well-wetted) pad displays a bulgy surface topography, while the failed (de-wetted) pad displays completely flatness. This is because tin solder will form a bump on pad under the effect of gravity and capillary force, rendering the normal pad a bulgy profile. In a similar way, since the de-wetted pad has no solder attached on it, its outermost surface is actually the final finish of ENIG which accounts for its flatness. The failed pad which is indicated by the red circle in Fig. 3a was further magnified. As it can be seen in Fig. 3b, the grains of the immersion gold on the failed pad are clear and it appears coarse under magnification of 2,0009. This feature became more obvious when the spot was magnified to 4,0009 (Fig. 3c). When magnified to 16,0009, besides Fig. 1 The role of surface finish in PCB J Fail. Anal. and Preven. (2013) 13:194–201 195 123
J Fail. Anal and Preven.(2013)13: 194-201 appearance of the failed (a) Plane 口口口口口口口口 口口口口口口口口 schematic diagram of sample 口口口口口口口口口 口口口口口口口口口 口口口口口口口口口 口口口口口口口口 口口口口口口口口口 口口口口口口口口 Wetting 口口口口口口 口口口口 EM 口口口口口口口口 口口口口口口口 口口口口口口 口口口口口 (b) -40 um (c d) Fig. 3 SEM results of the un-wetted pad (a) Un-wetting pad displays flatness, (b)coarse grains under 2,000x, (e)coarse grain under 4,000x and(d) crystalline interface crac
Fig. 2 Macroscopic appearance of the failed sample. (a) Plane configuration of sample, (b) schematic diagram of sample Fig. 3 SEM results of the un-wetted pad. (a) Un-wetting pad displays flatness, (b) coarse grains under 2,0009, (c) coarse grain under 4,0009, and (d) crystalline interface cracking 196 J Fail. Anal. and Preven. (2013) 13:194–201 123
J Fail. Anal. and Preven. (2013)13: 194-201 the coarse topography, we can also indentify crystalline oxide(Fig. 7)whose stoichiometry is Ni2gO7 according to interface cracking as labeled in Fig. 3d. It can also be Chong [4], or as Lee [5] proposed in his work, the nickel estimated that the cracking length is about 3-4 um oxide displayed stoichiometry of NiO2 in the outermost The region marked by a red square on failed pad was part and Nio in the inner part. No matter how different the chosen to be analyzed by EDS. The chemical composition stoichiometry of nickel oxide is, it ultimately acted as an and corresponding weight percentage is given in Fig. 4. As obstacle and blocked off the tin solder from contacting the it illustrates, four detected elements are Ni, Au, P, and Sn. gold layer under it during hot air solder leveling. Since And the weight percentage is 82.33, 13.60, 2.3, and 1.69%0, nickel oxide is a relatively stable substance, the desmear respectively. It is reasonable to find Ni and Au, since the before soldering could hardly play any role. This explains surface finish is ENG. The existence of P is acceptable why de-wetting still could not be avoided even after careful because phosphorus was introduced in during chemical desmear had been implemented before soldering plating of nickel. The little quantity of Sn on failed pad As revealed by the SEM result(Fig. 3d), the existence ould be explained as follows: although the pad was not of crystalline interface cracking acted as a"Grand well-wetted, there was still some tin solder residue on it Canyon", thus aggravated this mechanism by exposing through hot air solder leveling more underlying nickel atoms to oxygen in the environ- In order for comparison, EDS analysis was also utilized ment. Moreover, through"Grand Canyon", the nickel on a section of a normal (well-wetted) pad of the same atoms lied below diffused upward through the canyon to sample as show in Fig. 5. Likewise, the information of the outermost surface, covering the gold layer and forming chemical composition and weight percentage were a sandwich structure finally. Then the external nickel layer revealed. From Fig. 5, we mainly found Sn with little was oxidized and formed a barrier between tin solder and quantity of C and O. This result confirmed us that solder gold layer during hot air solder leveling later. The process used for joint was pure tin and C and o were organic described above was demonstrated in Fig 8. contamination from the surrounding environment which Besides that, organic substance which is highly volatile accidentally adhered to tin bump during storage from the depositary environment was absorbed into the According to the formula, one nickel atom will be canyon and covered the nethermost nickel layer, acting as replaced by two gold atoms. The radius of nickel and gold an isolating layer between solder and nickel and led the atom is 1. 24 and 1.44 A, respectively, thus there is an un-wetting during soldering. Since canyon had a consid- atomic radiuses difference of 16% between Ni and Au. The erable absorption capacity and as proved by Fig. 5 that the synergy of the two factors above left the ENIG finish a quantity of organic contamination in storage circumstance rough bumpy surface full of pits and holes as illustrated in could not be neglected, this mechanism was equally pos- Fig. 6. These pits and holes then exposed the nickel sible for the poor wetting performance. Figure 9 schemed straight to oxygen in the air and caused it to form nickel out the process. However, it is necessary to be put out that c:edax32genesis genmaps spc 08-Jul-2011 15: 33:35 71 KEnt 002.004006.008.0010.001200140016.001800 Energy.kev Element Ni Au P Wt%82.3313.602.381.69 Fig. 4 EDS result of the un-wetting pad (a) The un-wetting pad, (b) chemical composition, and (e) weight percentage of main elements Spring
the coarse topography, we can also indentify crystalline interface cracking as labeled in Fig. 3d. It can also be estimated that the cracking length is about 3–4 lm. The region marked by a red square on failed pad was chosen to be analyzed by EDS. The chemical composition and corresponding weight percentage is given in Fig. 4. As it illustrates, four detected elements are Ni, Au, P, and Sn. And the weight percentage is 82.33, 13.60, 2.3, and 1.69%, respectively. It is reasonable to find Ni and Au, since the surface finish is ENIG. The existence of P is acceptable because phosphorus was introduced in during chemical plating of nickel. The little quantity of Sn on failed pad could be explained as follows: although the pad was not well-wetted, there was still some tin solder residue on it after been through hot air solder leveling. In order for comparison, EDS analysis was also utilized on a section of a normal (well-wetted) pad of the same sample as show in Fig. 5. Likewise, the information of chemical composition and weight percentage were revealed. From Fig. 5, we mainly found Sn with little quantity of C and O. This result confirmed us that solder used for joint was pure tin and C and O were organic contamination from the surrounding environment which accidentally adhered to tin bump during storage. According to the formula, one nickel atom will be replaced by two gold atoms. The radius of nickel and gold atom is 1.24 and 1.44 A˚ , respectively, thus there is an atomic radiuses difference of 16% between Ni and Au. The synergy of the two factors above left the ENIG finish a rough bumpy surface full of pits and holes as illustrated in Fig. 6. These pits and holes then exposed the nickel straight to oxygen in the air and caused it to form nickel oxide (Fig. 7) whose stoichiometry is Ni29O71 according to Chong [4], or as Lee [5] proposed in his work, the nickel oxide displayed stoichiometry of NiO2 in the outermost part and NiO in the inner part. No matter how different the stoichiometry of nickel oxide is, it ultimately acted as an obstacle and blocked off the tin solder from contacting the gold layer under it during hot air solder leveling. Since nickel oxide is a relatively stable substance, the desmear before soldering could hardly play any role. This explains why de-wetting still could not be avoided even after careful desmear had been implemented before soldering. As revealed by the SEM result (Fig. 3d), the existence of crystalline interface cracking acted as a ‘‘Grand Canyon’’, thus aggravated this mechanism by exposing more underlying nickel atoms to oxygen in the environment. Moreover, through ‘‘Grand Canyon’’, the nickel atoms lied below diffused upward through the canyon to the outermost surface, covering the gold layer and forming a sandwich structure finally. Then the external nickel layer was oxidized and formed a barrier between tin solder and gold layer during hot air solder leveling later. The process described above was demonstrated in Fig. 8. Besides that, organic substance which is highly volatile from the depositary environment was absorbed into the canyon and covered the nethermost nickel layer, acting as an isolating layer between solder and nickel and led the un-wetting during soldering. Since canyon had a considerable absorption capacity and as proved by Fig. 5 that the quantity of organic contamination in storage circumstance could not be neglected, this mechanism was equally possible for the poor wetting performance. Figure 9 schemed out the process. However, it is necessary to be put out that Fig. 4 EDS result of the un-wetting pad. (a) The un-wetting pad, (b) chemical composition, and (c) weight percentage of main elements J Fail. Anal. and Preven. (2013) 13:194–201 197 123
J Fail. Anal and Preven.(2013)13: 194-201 c:edax32genesis'genmaps spc 08.Jul-2011 15: 34: 36 002.003.004005.006.007.008.00900 (b) Energy.kev ment Sn C Wt%93.512.783.72 Fig 5 EDS result of the wetting pad. The chemical reaction formula of depositing Au on nickel layer is as follows: 2Au(CN)2+Ni- 2Au Ni-+4CN.(a)The wetting pad, (b) chemical composition, and(c)weight percentage of main elem 8887 surface Incompletely protected nickel atom Fig 6 Formation of bumpy surface during immersion gold. (a)The deposition of Au on Ni, (b) bumpy surface formed Oxygen atom Bumpy surface oxygen atom bumpy surface Fig. 7 Formation of nickel oxide layer(a)Oxygen contacts the unprotected Ni(b) The unprotected nickel were oxidized within our reach there is no device could directly detect or total grain boundary area as illustrated in Fig. 10. It is the observe such little amount of organic contamination fact that energy along the grain boundary is higher than As mentioned before(Fig 3), the grain size of the that in bulk [6]. Since the atoms alone the grain bound- failed pad under inspection of SEM was relatively coarse aries were not perfectly bonded, they must have a higher and big (3-4 um). The big grain size led to the smaller tendency to engage in a chemical reaction compared with
within our reach there is no device could directly detect or observe such little amount of organic contamination. As mentioned before (Fig. 3), the grain size of the failed pad under inspection of SEM was relatively coarse and big (3–4 lm). The big grain size led to the smaller total grain boundary area as illustrated in Fig. 10. It is the fact that energy along the grain boundary is higher than that in bulk [6]. Since the atoms alone the grain boundaries were not perfectly bonded, they must have a higher tendency to engage in a chemical reaction compared with Fig. 5 EDS result of the wetting pad. The chemical reaction formula of depositing Au on nickel layer is as follows: 2Au(CN)2 ? Ni ? 2Au ? Ni2??4CN. (a) The wetting pad, (b) chemical composition, and (c) weight percentage of main elements Fig. 6 Formation of bumpy surface during immersion gold. (a) The deposition of Au on Ni, (b) bumpy surface formed Fig. 7 Formation of nickel oxide layer. (a) Oxygen contacts the unprotected Ni. (b) The unprotected nickel were oxidized 198 J Fail. Anal. and Preven. (2013) 13:194–201 123