Current-induced growth of P-rich phase at electroless nickel/Sn interface
Qiliang Yang, Panju Shang, Jing D. Guo, and Zhiquan LiuShenyang National Laboratory for Materials Science Institute of Metal Research,Chinese Academy of Sciences, Shenyang 110016, China
Jian-Ku Shanga)
Shenyang National Laboratory for Materials Science Institute of Metal Research,Chinese Academy of Sciences, Shenyang 110016, China; and Department of Materials Science andEngineering, University of Illinois at Urbana–Champaign, Urbana, Illinois 61801
(Received 9 December 2008; accepted 20 April 2009)
The role of high current stressing during growth of the P-rich phase at the electrolessNi/Sn interface was examined by transmission electron microscopy. Prior to currentstressing, two layers of Ni12P5, columnar Ni12P5 and noncolumnar Ni12P5, were formedafter soldering. Upon electric stressing, the two layers of P-rich phase showed oppositegrowth patterns at the two opposing electrode interfaces. At the cathode, columnargrowth of the P-rich phase was greatly enhanced while growth of the noncolumnar layerwas inhibited. By contrast, the opposite was found at the anode where the currentstressing promoted the noncolumnar growth but suppressed the growth of the columnarlayer. Such a strong polarity effect resulted from directional electromigration of the keyreaction species, nickel, to and from the interfacial reaction fronts. As a result of thedifference in reaction mechanism, overall growth of the P-rich phase was much faster atthe cathode during current stressing.
I. INTRODUCTION
Electroless Ni–P (EN) is an attractive under bumpmetallization (UBM) for solder interconnects because ofits low cost, easy processing, good selective deposition,and corrosion resistivity.1,2 EN can also act as an effec-tive diffusion barrier between Cu and Sn3–5 to limit thedevelopment of a thick Cu–Sn intermetallic compound(IMC) layer, which may degrade the mechanical proper-ties of a solder joint. For these reasons, a number ofstudies have been conducted on the interfacial reactionbetween solder and the EN layer. One group of thestudies focused on the microstructural evolution and thekinetics of intermetallic compound growth.6–27 Duringsolid-state reaction, the interfacial microstructure wasfound to depend strongly on the diffusion process of Niatoms in the EN layer and Sn in the solder,6,13–15 withthe P atoms in the EN layer playing a less active role inthe solid reaction. As Ni diffuses out of the EN layer,Kirkendall voids may develop in the P-rich layer, whichmay weaken the interface significantly.13,14 The othergroup of the EN studies have emphasized the mechan-ical properties of EN/solder interface after liquid orsolid reaction.28–38 These studies have shown that Ni–Sn IMC played an important role in decreasing the in-
terface strength,29,31 and the stress caused by Ni diffu-sion in the P-rich layer could also affect the interfacestrength.1,2,28,35
The P-rich layer usually forms between the EN layerand Ni–Sn IMC during soldering and aging processes dueto the consumption of Ni in the EN layer. The structure ofthe P-rich layer varies greatly with the concentration ofP in the EN,15,17 as several intermediate phases of Ni andP may form at the interface. This P-rich layer has beenshown to play a critical role in degrading the strength ofsolder joints and reducing the reliability of the solderedinterconnects.1,2,28,35
Although most of the studies on EN have focused onthermal reactions with solders, the effects of high currentloading also need to be considered. As the microelec-tronics industry moves toward further miniaturization ofdevice packages, the current density in solder joints mayrise to 104A/cm2, where electrical current stressing canreduce the strength of a Sn-based solder alloy39 or in-duce brittle fracture at the solder interface.40,41 Severalstudies have examined evolutions of microstructure andmechanical properties of solder/Ni–P structure undercurrent stressing42–46 and found that the current stressingstrongly affects the growth of Ni–Sn IMC layers and thestrength of solder joints. Despite the critical role of theP-rich layer in determining the reliability of the solderjoint, work aimed at understanding the effects of highcurrent density (>104 A/cm2) on the P-rich layer hasseldom been reported.
a)Address all correspondence to this author.e-mail: [email protected]
The objective of this study was to investigate thepotential effects of current stressing on growth of P-richlayer at the reactive interface between Sn and EN. Forthis purpose, Ni–P/Sn/Ni–P interconnects were prepared,and growth of P-rich layer under current stressing wasobserved under scanning electron microscopy (SEM)and transmission electron microscopy (TEM). A strongelectric polarity effect was found at the two opposingelectrodes where the growth rate of the P-rich layer wasmuch faster at the cathode interface than that at the an-ode. Such a polarity effect was shown to result from thedifferent reaction mechanisms that controlled the growthof P-rich layers at the two opposing electrode interfaces.
II. EXPERIMENTAL
The interconnect samples were prepared as Cu/Ni–P/Sn/Ni–P/Cu sandwich structures. Two Cu cubes wereused as the substrates for the electroless-nickel plating.The electroless-Ni plating solution consisted of 30 g/L ofNiCl2�6H2O, 72 g/L of Na3C6H5O7�2H2O, 48 g/L ofNH4Cl, and 22.5 g/L of NaH2PO2�H2O. The pH valueof the solution was adjusted to 9 by NH3�H2O. Pure Snwas used as the solder to bond two plated Cu cubes byreflow soldering.
The electroless plating was carried out in a water bathat a temperature of 385 K. Prior to plating, the surfacesof the Cu cubes were ground with 1200 grit abrasivepaper and polished with 1-mm diamond polishing paste.After polishing, the Cu surfaces were cleaned by ethanolin an ultrasonic bath for 2 min, and dried by the blast air.The prepared Cu surfaces were activated by a piece ofFe wire once the Cu cubes were placed into platingsolution because Cu was inert to the electroless-Niplating solution. The plating process was conducted for30 min to produce a layer of EN about 7.5 mm in thick-ness on the polished Cu surface. Under these conditions,electroless plating produced a Ni–P coating with a Pcontent of 14 at.% (8 wt%).
The two EN-plated Cu substrates were soldered withpure Sn at a temperature of 280 �C, and the soldering wasfinished seconds after the Sn melted. Electric stressingwas performed on the narrow beams cut from the sol-dered Cu/EN/Sn sandwich joints. The beam samples hada cross section of about 450 mm � 450 mm, before theywere ground to be about 230 mm � 230 mm. The sampleswere tested by applying a direct current at a currentdensity of around 4 � 104 A/cm2, while the temperatureof the samples was kept at about 85 � 5 �C. The electricstressing was continued for various times, up to 140 h.
After current stressing, samples were wrapped in anepoxy resin, ground on abrasive papers, and etched by asolution of CH3CH2OH, 95 mL, FeCl3 5 g, HCl 5 mL.The morphology and microstructure of the samples wereobserved under SEM and TEM. TEM samples were
prepared by using a focused ion beam (FIB) techniqueto cut thin, electron-transparent slices from the solderjoint. TEM observations were made on a JEM-2010(JEOL, Tokyo, Japan) and an FEI Tecnai F30 electronmicroscope (Hillsboro, OR) operated at an acceleratingvoltage of 200 and 300 kV, respectively.
III. RESULTS
Figure 1 shows the initial structure of the Cu/Ni–P/Sninterconnect after soldering. In front of the 7.5-mm ENlayer on the Cu substrate there is an IMC layer about2 mm thick. Between the IMC and EN layer, there is adark layer of about 0.4 mm. The SEM line scan from Sn,across this dark layer, to the EN layer shows that the Pcontent increased sharply in this dark layer, which isthus marked as a P-rich layer. Similar P-rich layers werereported previously after soldering Sn or Sn-based solderalloys onto the electroless Ni.6,8,11,16 In addition, smallNi-containing particles of IMCs were also seen in theSn, which were formed during the liquid reaction of Snand Ni as Ni atoms diffused into the liquid Sn.The interfacial microstructures at the two electrodes
are shown in Fig. 2 after the current stressing at a con-stant current density of 4 � 104 A/cm2 for 140 h and asample temperature of 80–90 �C. At the cathode, thethickness of the P-rich layer increased to about 2.3 mmin Fig. 2(a). Next to the continuous P-rich layer, IMCparticles were randomly distributed in a band about10 to 30 mm away from the Sn/P-rich layer interface.On the Sn/P-rich layer interface, almost no IMCs wereadhered to the P-rich layer. Instead, most IMC hadflaked off from the P-rich layer after the current stress-ing. At the anode side, as shown in Fig. 2(b), the IMClayer formed during soldering was still adhered to theEN layer and thickness of the P-rich layer was only
FIG. 1. Interfacial microstructure between Sn and EN in the as-soldered
state, together with the line scan of P content across the interface.
Q. Yang et al.: Current-induced growth of P-rich phase at electroless nickel/Sn interface
J. Mater. Res., Vol. 24, No. 9, Sep 20092768
increased slightly to about 0.5 mm in thickness. There-fore, the growth of the P-rich layer was clearly polarity-dependent under the current stressing.
To further understand this disparity in the growth ofthe P-rich layer between the two electrode interfaces, themicrostructures of the interfaces were observed underTEM before and after current stressing. In the as-sol-dered state, three distinct layers were observed at theinterface, which were marked as layer I, II, and III inFig. 3(a). Energy dispersive x-ray (EDX) analysis [seeFigs. 3(b)–3(d)] indicated that layer II was made of thebinary compound of Ni and P, and layer III a ternarycompound of Ni, Sn, and P. The compositions of thoselayers are listed in Table I. The selected-area electrondiffraction (SAED) patterns of layers I and II, and theconvergent-beam electron diffraction (CBED) pattern oflayer III are also included in Fig. 3(a). On the basis ofboth compositional analysis and diffraction patterns, lay-er I was identified as amorphous EN, layer II as Ni12P5,and the ternary compound layer III as Ni2SnP. TheNi2SnP layer rather than Ni3Sn4 was also previouslyobserved by others,15 but the P-rich structure varied
greatly from different studies.13–17,19,23,47 For clarifica-tion, the EN/Ni12P5 and Ni12P5/Ni2SnP interface areindicated in Fig. 3(a) by the black and white arrows,respectively. The P-rich layer, namely Ni12P5, was about0.4 mm in thickness, which is comparable to the SEM
FIG. 2. SEM images of the interfacial microstructures at the
(a) cathode side and (b) anode side after current stressing at 4 �104 A/cm2 for 140 h.
FIG. 3. (a) Bright-field image of interfacial microstructure in the
as-soldered state and the corresponding diffraction patterns of
amorphous EN layer-I, Ni12P5 layer-II, and Ni2SnP layer-III. In
layer II, there are two sublayers whose interface is indicated with
asteroids. (b)–(d) EDX analysis of columnar Ni12P5 sublayer (right
side in II), noncolumnar Ni12P5 sublayer (left side in II), and
Ni2SnP layer (III).
Q. Yang et al.: Current-induced growth of P-rich phase at electroless nickel/Sn interface
J. Mater. Res., Vol. 24, No. 9, Sep 2009 2769
measurement, but it consisted of two sublayers with dif-ferent grain structures as marked by the asterisks. Thegrains in the sublayer right next to the amorphous EN werecolumnar, extending perpendicular to the Ni–P/Ni12P5 in-terface. The widths of these columnar Ni12P5 grainsranged from 50 to 150 nm. The second sublayer hadroughly equiaxed grains, thus referred to as noncolumnarNi12P5 grains. The grain sizes of this second sublayerranged from 150 to 430 nm, with the grain boundariesdelineated by the white dashed line in Fig. 3(a). Bothsublayers had almost the same thickness, about 0.2 mm.
After the current stressing at 4.0 � 104 A/cm2 for 93 h,the resulting interfacial microstructure at the cathode sideis shown in Fig. 4(a), in which different interfaces areindicated with arrows and asterisks. According to theSAED patterns [insets: Fig. 4(a)], layer II was stillNi12P5 while layer III was confirmed as Ni3Sn4. Com-pared to the as-soldered state, the total thickness ofNi12P5 layer was increased significantly to about 1.6 mm,which is slightly smaller than the SEM measurement.Most of the thickness increase was contributed by the fastgrowth of columnar Ni12P5, which accounted for four-fifths of the thickness of the entire Ni12P5 layer. Thegrowth of the columnar Ni12P5 crystals maintained thesame direction as that in the as-soldered state, with theirlong axes perpendicular to the EN/Ni12P5 interface. Bycomparison, the noncolumnar Ni12P5 layer grew onlyslightly. Next to the noncolumnar Ni12P5 layer, a Ni3Sn4IMC layer replaced the Ni2SnP in the as-soldered state[see diffraction patterns in Fig. 4(a)]. Within the Ni3Sn4IMC layer, most of the areas were free of P [Fig. 4(b)] butisolated P-rich spots were also detected, as Fig. 4(c) indi-cated, which contained 9.8 at.% P.
The effect of the current stressing on the microstructureof the anode interface is shown in Fig. 5(a) after theelectric loading for 142 h. The total thickness of theNi12P5 layer was noticeably smaller compared to that atthe cathode interface. Within the Ni12P5 layer, the colum-nar Ni12P5 sublayer had almost the same thickness as thatin the as-soldered state, but the thickness of the nonco-lumnar Ni12P5 layer was tripled so that the noncolumnarNi12P5 layer at the anode became much thicker than thatat the cathode. The grain size of the noncolumnar Ni12P5had also expanded. As observed at the anode interface,the Ni2SnP layer was replaced by a Ni3Sn4 IMC layer,according to the SADP pattern [inset: Fig. 5(a)]. Withinthe Ni3Sn4 IMC layer, no P was detected, as shown by theEDX in Fig. 5(b).
For comparison, the growth kinetics of P-rich layers isplotted in Fig. 6 as a function of time. At the cathode, thethickness of the noncolumnar Ni12P5 layer only in-creased slightly under the current stressing for 93 h, butthe columnar Ni12P5 layer grew much faster, especiallyafter the current stressing for more than 40 h. As for theanode, an opposite growth behavior was observed inthat the columnar Ni12P5 grew slightly in about 140 h ata slope of 0.003 mm/(h1/2), whereas the noncolum-nar Ni12P5 grew 10 times faster at a growth slope of0.03 mm/(h1/2). Thus, the growth of the P-rich layer atthe cathode was mainly contributed by the columnarNi12P5, but the growth of P-rich at the anode was mostlycontributed by the noncolumnar Ni12P5. For the wholeP-rich layer, the growth rate was much faster at thecathode side. In addition, the growth of the P-rich layerat the anode side scaled linearly with t1/2, as expected fora diffusion-controlled process. However, for the cathodeside, the growth kinetics deviated greatly from a simplelinear relation.
IV. DISCUSSION
The results of this study have shown that the growthof the P-rich layer under the electric stressing was
FIG. 4. (a) Bright-field image and diffraction patterns of Ni12P5 lay-
er-II and Ni3Sn4 layer-III showing the interfacial microstructure at
the cathode side after current stressing at 4.0 � 104 A/cm2 for 93 h.
(b, c) EDX results of P-free and P-rich areas in Ni3Sn4 layer.
TABLE I. Compositions of the Ni12P5 and Ni2SnP layers.
Columnar Noncolumnar Ni2SnP
P (at.%) 25.94 29.87 19.99
Ni (at.%) 74.06 70.13 50.70
Sn (at.%) . . . . . . 29.31
Q. Yang et al.: Current-induced growth of P-rich phase at electroless nickel/Sn interface
J. Mater. Res., Vol. 24, No. 9, Sep 20092770
strongly polarity-dependent. The layer grew faster atthe cathode, where the columnar growth dominated thegrowth process. In contrast, it was the expansion of thenoncolumnar layer that dominated the growth process atthe anode. For the binary compound, the solid-stategrowth is generally determined by the diffusive supplyof the two reactive species. Because either or both reac-tive species could control the Ni12P5 growth, understand-ing such a strong polarity effect from the electricstressing requires the determination of the critical reac-tion step, which must be based on the specific interfacialreaction mechanisms involved. In the case of interfacialreaction between EN and Sn, as this study and othersdemonstrated,13–15,17,19,23,47 the reaction products arevery complicated so that the reaction mechanisms mayvary greatly with experimental conditions. Withoutknowledge of the key reaction mechanisms involved,the following analysis was conducted by assuming thatNi was the controlling reactive species in the interfacialreaction during the growth of the Ni12P5 phase and byanalyzing diffusive fluxes of Ni under electric stressing.
Since sandwich structures were used in this study, thetemperature field from the Joule heating should be sym-
metrical at the anode and the cathode when the currentwas passed through the sample. When the analysis isfocused on the difference between the two electrodeinterfaces, the thermomigration effect does not haveto be considered. Thus, the analysis was focused onlyon the roles of the chemical potential gradient forceand electron wind force in the growth of IMC. Underthose two forces, the atomic drift fluxes of Ni can bewritten as
�!Jchem ¼ ND
@ lnN
@x; ð1Þ
�!Jem ¼ ND
kTZ�e!E ; ð2Þ
where N is the atomic density; D, diffusivity; k, T, Boltz-mann’s constant and the temperature in Kelwin, respec-tively; x, the coordinate in the direction of electronflow; Z�, the effective charge number; e, the electroniccharge; E, the electric field; and E ¼ rj, with r being theresistivity and j the current density.
FIG. 6. The kinetics of Ni–P IMC (columnar and noncolumnar
Ni12P5) during current stressing at the (a) cathode side and (b) anode
side.
FIG. 5. (a) Interfacial microstructure at the anode side after current
stressing at 4.3 � 104 A/cm2 for 142 h. Inset is the corresponding
diffraction pattern of Ni3Sn4 layer. (b) EDX spectrum of Ni3Sn4 layer.
No P was detected in this layer.
Q. Yang et al.: Current-induced growth of P-rich phase at electroless nickel/Sn interface
J. Mater. Res., Vol. 24, No. 9, Sep 2009 2771
Because the effective charge number, Z�, of Ni is nega-tive,48,49 Ni atoms would be forced by the current stressingto migrate from the cathode to the anode. The resultingatomic drift fluxes of Ni at the cathode and anode inter-faces are as follows:
where Jcathode is the Ni flux at the cathode, and Janode isthe Ni flux at the anode. Both Jcathode and Janode are in thedirection of electron flow, from the cathode to the anode.Jchem, the Ni flux due to the chemical potential gradient,and Jem, the Ni flux due to the electron wind force.
The effect of the current stressing on the interfacialreactions in the Cu/Ni–P/Sn/Ni–P/Cu interconnect isschematically illustrated in Fig. 7. At the cathode side,the electron wind force is in the same direction as thedirection of the chemical potential force so that the dif-fusion of Ni from the EN layer to Sn would be enhancedby the electric current, as Eq. (3) indicates. As Ni atomsdiffused out of the EN layer, the P concentration wouldrise quickly, from 14 to �29 at.%, which is the stoichio-metric proportion of Ni12P5 at the columnar Ni12P5/ENinterface. The influx of Ni and P to the interface wouldthen feed the growth of the columnar Ni12P5 crystals,leading to the transformation of the EN to Ni12P5 by thefollowing reaction:
5Ni86P14 ! 14Ni12P5 þ 262Ni : ð5ÞBecause no additional nucleation of Ni12P5 crystals
is required, the columnar Ni12P5 layer should extendrapidly in the direction perpendicular to the Ni–P/Ni12P5 interface. Moreover, Ni2SnP IMC layer that ad-hered to the P-rich layer in the as-soldered state wasflaked off following the action of the electron windforce, as shown in Fig. 2(a). Without the barrier ofNi–Sn IMC layer, Ni atomic drift would be acceleratedeven more, resulting in a much higher growth rate ofP-rich layer in deviation from the linear relation witht1/2, as seen in Fig. 6(a).
Besides the columnar growth of Ni12P5, the transfor-mation of the Ni2SnP phase to Ni3Sn4 may also followfrom Ni migration at the anode interface. Yeh and Hun-tington48 reported previously that current stressingforced Ni atoms to migrate in the direction of the elec-tron flow, resulting in decomposition of Ni–Sn IMC.Similar decomposition might have occurred in thisstudy. In particular, the Ni2SnP phase formed in the as-soldered state could decompose under the current stress-ing by the following reaction:
4Ni2SnP ! Ni3Sn4 þ 5Niþ 4P : ð6ÞUnder the current stressing, the departure of Ni atoms
from the cathode would drive reaction (6) to the right so
that Ni3Sn4 would replace Ni2SnP. The excess P wouldbe left behind in the newly formed Ni3Sn4 layer, asdetected by the EDX analysis during the TEM observa-tion [Fig. 4(c)], from which the P concentration wasestimated to range from 0 to 9.7 at.%.At the anode side, the chemical potential gradient force
for Ni is in the opposite direction to the direction of theelectron wind force. The electric wind force should im-pede the Ni diffusion out of the EN layer, as described byEq. (4). The resulting reduction in the Ni flux at the ENinterface would inhibit the growth of the columnar P-richlayer at the anode side, as observed in Fig. 6(b).On the other hand, the minimal growth of the nonco-
lumnar Ni12P5 layer at the cathode means that growthof the noncolumnar Ni12P5 layer may have proceededby a rather different mechanism. It should be noted thatthe noncolumnar layer is located between the columnarand the Ni2SnP layer after the reflow. Its growth couldoccur by the motion of the interface with the columnarlayer or the motion of the interface with the Ni2SnPphase or both. However, the growth of the noncolumnarlayer to the side of the columnar layer requires the
FIG. 7. Schematics showing interfacial reactions are affected by the
Ni fluxes under current stressing at the (a) cathode side and (b) anode
side.
Q. Yang et al.: Current-induced growth of P-rich phase at electroless nickel/Sn interface
J. Mater. Res., Vol. 24, No. 9, Sep 20092772
recrystallization of the columnar layer, which is energeti-cally challenging. Suppose that the growth of the non-columnar Ni12P5 takes place mostly at the interface withthe Ni2SnP phase. At the anode, Ni atoms would bedriven to flow from Sn through the Ni2SnP phase to theinterface, where the following reaction may emerge
20Ni2SnPþ 23Ni ! 4Ni12P5 þ 5Ni3Sn4 : ð7ÞThe reaction would not only produce Ni12P5 but also
result in the decomposition of Ni2SnP to Ni3Sn4, both ofwhich were the reaction products observed at the anodeafter the current stressing [Fig. 5(a)]. At the cathode,since Ni was diffusing away from the reacting interfaceunder the electric wind force, reaction (7) would not befavored. Consequently, the growth of the noncolumnarlayer would be retarded by the current stressing, as ob-served in Fig. 6(a). Even though the Ni2SnP layer at theanode was also replaced by Ni3Sn4 layer after the currentstressing, no excess P should be produced and the Ni3Ni4layer should be free of P, as confirmed by the TEMobservation.
Although the changes of the microstructures in theP-rich layers observed in this study do not constituteelectromigration failures of the joint, the newly formedP-rich layer could act as a weak link in subsequent elec-trical and/or mechanical stressing of the solder inter-connect and therefore directly impact the reliability ofthe electroless nickel/solder interconnects in flip-chipdevices, because a number of studies have shown thatthe microstructure of the P-rich layers plays a criticalrole in determining the reliability of the electrolessnickel/solder interconnects.1,2,13,14,28,35 The intermetalliccompounds have been a reliability concern for manydevices because they generate stress concentrations andoften break in a brittle manner. The observations inFigs. 3–5 clearly revealed the exact types of the interme-tallic compounds and the previous analysis showed theyformed by reactions (5)–(7). Therefore, this study hasprovided vital new knowledge to understanding electro-migration failures by clarifying the structures of P-richlayer and revealing what and how the intermetallic com-pounds were formed in the P-rich layer.
V. CONCLUSIONS
The effects of current stressing on interfacial reactionshave been examined by cross-section TEM. The electriccurrent induced a strong polarity effect on the interfacialreactions between Sn and electroless Ni. On the basis ofexperimental observations and analysis, the followingconclusions have been reached.
(1) Two separate P-rich sublayers were found at theinterface between Sn and EN. One had a columnar grainstructure and the other equiaxed grain structure. Bothlayers were Ni12P5. The columnar Ni12P5 layer grew from
the EN by the reaction 5Ni86P14 ! 14Ni12P5 þ 262Ni;whereas the noncolumnar Ni12P5 layer came from thereaction 20Ni2SnPþ 23Ni ! 4Ni12P5 þ 5Ni3Sn4 at theinterface with IMC.(2) The current stressing promoted the growth of co-
lumnar Ni12P5 at the cathode side, especially after thedetachment of Ni–Sn IMC from the cathode interface,while impeding the growth of the noncolumnar Ni12P5layer. At the anode, the current stressing enhanced thegrowth of the noncloumunar Ni12P5 layer, but inhibitedthe growth of the columnar Ni12P5 layer.(3) At both electrode interfaces, the current stressing
resulted in the replacement of P-containing Ni2SnP IMCby Ni3Sn4 compound. At the cathode, the replacementreaction benefited from a loss of Ni, whereas at the an-ode it required an infusion of Ni.
ACKNOWLEDGMENTS
This study was supported by the Chinese Natural Sci-ence Foundation under Grant No. 50228101 and theNational Basic Research Program of China, Grant No.2004CB619306. The experimental assistance of J.J. Guoand John Paul Daghfal are greatly appreciated.
REFERENCES
1. M.O. Alam, Y.C. Chan, and K.N. Tu: Effect of reaction time and
P content on mechanical strength of the interface formed between
eutectic Sn–Ag solder and Au electroless Ni–P Cu bond pad.
J. Appl. Phys. 94, 4108 (2003).
2. M.O. Alam, Y.C. Chan, and K.C. Hung: Reliability study of the
electroless Ni–P layer against solder alloy. Microelectron. Reliab.42, 1065 (2002).
3. J.W. Jang, D.R. Frear, T.Y. Lee, and K.N. Tu: Morphology of
interfacial reaction between lead-free solders and electroless Ni–P
under bump metallization. J. Appl. Phys. 88, 6359 (2000).4. C.Y. Lee and K.L. Lin: The interaction kinetics and compound
formation between electroless Ni–P and solder. Thin Solid Films249, 201 (1994).
5. S.J. Wang and C.Y. Liu: Retarding growth of Ni3P crystalline
layer in Ni(P) substrate by reacting with Cu-bearing Sn(Cu)
solders. Scr. Mater. 49, 813 (2003).
6. A. Kumar, M. He, and Z. Chen: Barrier properties of thin Au/Ni–
P under bump metallization for Sn–3.5Ag solder. Surf. Coat.Technol. 198, 283 (2005).
7. A. Kumar, Z. Chen, S.G. Mhaisalkar, C.C. Wong, P.S. Teo, and
V. Kripesh: Effect of Ni–P thickness on solid-state interfacial
reactions between Sn–3.5Ag solder and electroless Ni–P metalli-
zation on Cu substrate. Thin Solid Films 504, 410 (2006).
8. M. He, Z. Chen, G.J. Qi, C.C. Wong, and S.G. Mhaisalkar: Effect
of post-reflow cooling rate on intermetallic compound formation
between Sn–3.5 Ag solder and Ni–P under bump metallization.
Thin Solid Films 462–463, 363 (2004).
9. J.W. Yoon and S.B. Jung: Growth kinetics of Ni3Sn4 and Ni3P
layer between Sn–3.5Ag solder and electroless Ni–P substrate.
J. Alloys Compd. 376, 105 (2004).
10. P. Sun, C. Andersson, X. Wei, Z. Cheng, D. Shangguan, and
J. Liu: High temperature aging study of intermetallic compound
formation of Sn–3.5Ag and Sn–4.0Ag–0.5Cu solders on electro-
less Ni(P) metallization. J. Alloys Compd. 425, 191 (2006).
Q. Yang et al.: Current-induced growth of P-rich phase at electroless nickel/Sn interface
J. Mater. Res., Vol. 24, No. 9, Sep 2009 2773
11. M.O. Alam, Y.C. Chan, and K.C. Hung: Interfacial reaction of
Pb–Sn solder and Sn–Ag solder with electroless Ni deposit during
reflow. J. Electron. Mater. 31, 1117 (2002).
12. Y.C. Lin and J.G. Duh: Optimal phosphorous content selection
for the soldering reaction of Ni–P under bump metallization with
Sn–Ag–Cu Solder. J. Electron. Mater. 35, 1665 (2006).
13. J.W. Jang, P.G. Kim, K.N. Tu, D.R. Frear, and P. Thompson:
Solder reaction-assisted crystallization of electroless Ni–P under
bump metallization in low cost flip chip technology. J. Appl.Phys. 85, 8456 (1999).
14. M. He, Z. Chen, and G.J. Qi: Solid state interfacial reaction of
Sn–37Pb and Sn–3.5Ag solders with Ni–P under bump metalliza-
tion. Acta Mater. 52, 2047 (2004).
15. Y.C. Lin, T.Y. Shih, S.K. Tien, and J.G. Duh: Morphological and
microstructural evolution of phosphorous-rich layer in SnAgCu/
Ni–P UBM solder joint. J. Electron. Mater. 36, 1469 (2007).
16. Y.C. Lin and J.G. Duh: Phase transformation of the phosphorus-rich
