+ All Categories
Home > Documents > IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND...

IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND...

Date post: 30-Jun-2020
Category:
Upload: others
View: 7 times
Download: 0 times
Share this document with a friend
9
IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL. 2, NO. 4, APRIL 2012 625 Studies on the Thermal Cycling Reliability of BGA System-in-Package (SiP) with an Embedded Die Seon Young Yu, Yong-Min Kwon, Jinsu Kim, Taesung Jeong, Seogmoon Choi, and Kyung-Wook Paik, Member, IEEE Abstract—Nowadays, major trends in the design of electronic products are toward multifunction and miniaturization. To meet these trends, system-in-package (SiP) has been adapted as one of the core packaging technologies for many product applications. Among the various types of SiPs, SiP with embedded dies has become important due to the smaller size achieved through embedded dies and better electrical performance by the shorter interconnection length. However, reliability data of the SiP with embedded dies have not been reported yet. Therefore, it is necessary to investigate the reliability of the SiPs with embedded dies and the effect of the embedded die on the SiP reliability. Of the several reliability tests, a detailed thermal cycling test (T/C test) was performed on board-level packaged samples. A finite element method (FEM) simulation was also performed to find out the stress and strain distribution of ball grid array (BGA) solder positions and to predict the potential failure sites under the T/C test. Through this paper, it was found that the failure position of the BGA changed from the corner solder ball position of the conventional BGA package, where the largest distance from neutral point was, to the inner BGA solder ball positions, where the edge of embedded die was located, due to the complicated structure of embedded die SiPs. Furthermore, FEM results showed that the inner site of the BGA is more vulnerable than the corner BGAs. This was well matched with experimental results. Index Terms— Embedded die, finite element method, solder joint reliability, system-in-package, thermal cycling reliability. I. I NTRODUCTION A S THE requirements for small-size electronic prod- ucts have increased, electronic components have been developed to downsize and integrate more functional blocks. System-in-package (SiP), which is one of the solutions of this need for providing multiple functions, is a combination of multiple active and passive electronic components that have different functionalities in a single package. Therefore, the Manuscript received December 22, 2010; revised July 26, 2011; accepted July 27, 2011. Date of publication January 31, 2012; date of current version March 30, 2012. Recommended for publication by Associate Editor A. Chandra upon evaluation of reviewers’ comments. S. Y. Yu, Y.-M. Kwon, and K.-W. Paik are with the Department of Materials Science and Engineering, Korea Advanced Institute of Science and Technol- ogy, Daejeon 305-701, Korea (e-mail: [email protected]; [email protected]; [email protected]). J. Kim and S. Choi are with the Corporate Research and Development Institute, Samsung Electro-Mechanics, Gyunggi-do 443-743, Korea (e-mail: [email protected]; [email protected]). T. Jeong is with the Advanced Circuit Interconnection Division, Sam- sung Electro-Mechanics, Gyunggi-do 443-743, Korea (e-mail: ts1010.jung@ samsung.com). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TCPMT.2011.2166555 application fields of SiP are getting broader because of several advantages: high performance, ultralight weight, microscopic size, shorter time-to-market, and performance optimization. Also, the SiP with an embedded die is one of the solutions to make the package size even smaller and obtain higher electronic performance [1]. Although several reliability tests of the SiP packages have been reported [2], reliability data of the SiP with an embedded die are insufficient and more studies are needed to investigate the effects of embedded dies on board-level reliability data of these SiPs. Among the various reliability tests, thermal cycling (T/C) tests were performed, and the results are analyzed in this paper. Also, the difference between the SiP with an embedded die and similar ball grid array (BGA) packages was also investigated using a finite element method (FEM) simulation during the T/C test. The reliability data and failure analysis of the SiP package with an embedded die and the information about effects of the embedded die on SiP under T/C test will provide a guideline for more reliable SiPs with embedded dies. II. EXPERIMENTAL A. Test Samples Preparation The SiP was packaged as a BGA format and consisted of two parts: molding with epoxy molding compounds (EMC) and substrate layers. Fig. 1 shows a schematic of the cross- sectional view of the SiP with an embedded die assembled on a printed circuit board (PCB) test board. First, in the substrate layers, there was an embedded die with 1.9 × 1.9 mm size and 0.18 mm thickness at the center of the substrate. The core layer had a cavity to accommodate the embedded die, and had the same thickness as the embed- ded die. Second, the EMC molding was applied just on the substrate layers. On the substrate layers, there were several passive components, saw filters, and a flip-chip (FC) die. The FC die was placed at the center of the substrate. The SiP was basically assembled on another PCB test board by using BGA solder balls. Two solder compositions were used in this paper: Sn–Ag3–Cu0.5 (wt.%) (SAC 305) and Sn–Ag1–Cu0.5 (wt.%) (SAC 105). Two kinds of metal pad finishes were used at the BGA PCB substrate: electroplated Ni/Au (Au-EP) and Cu-organic solderability preservative (Cu- OSP). The PCB board finishes were: 1) electroless Ni/Au (ENIG), and 2) Cu-OSP, as shown in Fig. 2. 2156–3950/$31.00 © 2012 IEEE
Transcript
Page 1: IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND ...npil.kaist.ac.kr/pdf/foreign_journal/FJ_112.pdf · IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL.

IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL. 2, NO. 4, APRIL 2012 625

Studies on the Thermal Cycling Reliability of BGASystem-in-Package (SiP) with an Embedded Die

Seon Young Yu, Yong-Min Kwon, Jinsu Kim, Taesung Jeong, Seogmoon Choi,and Kyung-Wook Paik, Member, IEEE

Abstract— Nowadays, major trends in the design of electronicproducts are toward multifunction and miniaturization. To meetthese trends, system-in-package (SiP) has been adapted as one ofthe core packaging technologies for many product applications.Among the various types of SiPs, SiP with embedded dies hasbecome important due to the smaller size achieved throughembedded dies and better electrical performance by the shorterinterconnection length. However, reliability data of the SiP withembedded dies have not been reported yet. Therefore, it isnecessary to investigate the reliability of the SiPs with embeddeddies and the effect of the embedded die on the SiP reliability.Of the several reliability tests, a detailed thermal cycling test(T/C test) was performed on board-level packaged samples. Afinite element method (FEM) simulation was also performed tofind out the stress and strain distribution of ball grid array(BGA) solder positions and to predict the potential failure sitesunder the T/C test. Through this paper, it was found that thefailure position of the BGA changed from the corner solder ballposition of the conventional BGA package, where the largestdistance from neutral point was, to the inner BGA solder ballpositions, where the edge of embedded die was located, due to thecomplicated structure of embedded die SiPs. Furthermore, FEMresults showed that the inner site of the BGA is more vulnerablethan the corner BGAs. This was well matched with experimentalresults.

Index Terms— Embedded die, finite element method, solderjoint reliability, system-in-package, thermal cycling reliability.

I. INTRODUCTION

AS THE requirements for small-size electronic prod-ucts have increased, electronic components have been

developed to downsize and integrate more functional blocks.System-in-package (SiP), which is one of the solutions of thisneed for providing multiple functions, is a combination ofmultiple active and passive electronic components that havedifferent functionalities in a single package. Therefore, the

Manuscript received December 22, 2010; revised July 26, 2011; acceptedJuly 27, 2011. Date of publication January 31, 2012; date of current versionMarch 30, 2012. Recommended for publication by Associate Editor A.Chandra upon evaluation of reviewers’ comments.

S. Y. Yu, Y.-M. Kwon, and K.-W. Paik are with the Department of MaterialsScience and Engineering, Korea Advanced Institute of Science and Technol-ogy, Daejeon 305-701, Korea (e-mail: [email protected]; [email protected];[email protected]).

J. Kim and S. Choi are with the Corporate Research and DevelopmentInstitute, Samsung Electro-Mechanics, Gyunggi-do 443-743, Korea (e-mail:[email protected]; [email protected]).

T. Jeong is with the Advanced Circuit Interconnection Division, Sam-sung Electro-Mechanics, Gyunggi-do 443-743, Korea (e-mail: [email protected]).

Color versions of one or more of the figures in this paper are availableonline at http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TCPMT.2011.2166555

application fields of SiP are getting broader because of severaladvantages: high performance, ultralight weight, microscopicsize, shorter time-to-market, and performance optimization.Also, the SiP with an embedded die is one of the solutionsto make the package size even smaller and obtain higherelectronic performance [1].

Although several reliability tests of the SiP packages havebeen reported [2], reliability data of the SiP with an embeddeddie are insufficient and more studies are needed to investigatethe effects of embedded dies on board-level reliability data ofthese SiPs. Among the various reliability tests, thermal cycling(T/C) tests were performed, and the results are analyzed in thispaper. Also, the difference between the SiP with an embeddeddie and similar ball grid array (BGA) packages was alsoinvestigated using a finite element method (FEM) simulationduring the T/C test.

The reliability data and failure analysis of the SiP packagewith an embedded die and the information about effects of theembedded die on SiP under T/C test will provide a guidelinefor more reliable SiPs with embedded dies.

II. EXPERIMENTAL

A. Test Samples Preparation

The SiP was packaged as a BGA format and consisted oftwo parts: molding with epoxy molding compounds (EMC)and substrate layers. Fig. 1 shows a schematic of the cross-sectional view of the SiP with an embedded die assembled ona printed circuit board (PCB) test board.

First, in the substrate layers, there was an embedded diewith 1.9 × 1.9 mm size and 0.18 mm thickness at the centerof the substrate. The core layer had a cavity to accommodatethe embedded die, and had the same thickness as the embed-ded die.

Second, the EMC molding was applied just on the substratelayers. On the substrate layers, there were several passivecomponents, saw filters, and a flip-chip (FC) die. The FC diewas placed at the center of the substrate.

The SiP was basically assembled on another PCB test boardby using BGA solder balls. Two solder compositions wereused in this paper: Sn–Ag3–Cu0.5 (wt.%) (SAC 305) andSn–Ag1–Cu0.5 (wt.%) (SAC 105). Two kinds of metal padfinishes were used at the BGA PCB substrate: electroplatedNi/Au (Au-EP) and Cu-organic solderability preservative (Cu-OSP). The PCB board finishes were: 1) electroless Ni/Au(ENIG), and 2) Cu-OSP, as shown in Fig. 2.

2156–3950/$31.00 © 2012 IEEE

Page 2: IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND ...npil.kaist.ac.kr/pdf/foreign_journal/FJ_112.pdf · IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL.

626 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL. 2, NO. 4, APRIL 2012

SiP Substratelayers

FC die

Embedded die

BGA

EMC molding

PCB test board

Failuremeasurement

Fig. 1. Cross-sectional view of the test sample.

SiP Au-EP

ENIG sample Cu-OSP sample

Cu-OSPENIGSiP

BoardBoard

Fig. 2. Pad finish combinations of the test samples.

(a) (b)

Fig. 3. Two kinds of resistance measurement positions.(a) BGA solder. (b) Embedded die bump.

(a)

(b)

PassivesSaw filter

FC die

EmbeddeddiePCB

Fig. 4. 3-D quarter modeling of SiP assembled board using ABAQUSsimulation program. (a) SiP modeling. (b) Inside structure of SiP.

On the PCB test board, five test SiPs were mounted,and electrical failures at either the BGA or the embeddeddie bumps could be identified. There were two resistancemeasurement sites, i.e., corner and inner sites, to find outat what position the bumps were failing among 156 BGAand 151 embedded die bump positions. In Fig. 3, red colorrepresents BGA balls and embedded die bumps that are placedat a corner site, and yellow color represents those placed atan inner position.

(a)

(b)

(c)

(d)

Fig. 5. Four kinds of 3-D quarter modeling of SiP. (a) SiP modeling withoutembedded die. (b) SiP modeling without FC die. (c) SiP modeling withembedded die in SiP substrate filled only with PCB core material. (d) SiPmodeling with only EMC molding and SiP substrate filled only with PCBcore material.

B. T/C Test

T/C test was performed in the −40 to 125 °C temperaturerange per JEDEC standard. During the T/C test, the resistanceof the BGA solders and embedded die bumps were measuredevery 200 cycles up to 3200 cycles using a resistance meter.The failure criterion was twice the initial resistance and, thelifetime of each sample was obtained from a Weibull plot.

C. FEM Simulation

The FEM was used to simulate the T/C test condition ofthe SiP assembled at the PCB board level using the ABAQUS6.9 program. A quarter of the SiP assembled board was usedto make the simulation model, as shown in Fig. 4, because ofthe bilaterally symmetric structure of the SiP assembled board.

Page 3: IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND ...npil.kaist.ac.kr/pdf/foreign_journal/FJ_112.pdf · IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL.

YU et al.: THERMAL CYCLING RELIABILITY OF BGA SYSTEM-IN-PACKAGE (SiP) WITH AN EMBEDDED DIE 627

0

20

40

60

80

100

BGA-innerBGA-corner

Embedded die-inner

Embedded die-corner

TC

fai

lure

rat

e(%

)

0 400 800 1200 1600 2000 2400 2800 32000

20

40

60

80

100

TC

fai

lure

rat

e(%

)

Test cycles(a)

0 400 800 1200 1600 2000 2400 2800 3200Test cycles

(b)

BGA-inner

BGA-corner

Embedded die-inner

Embedded die-corner

Fig. 6. Failure rate of SAC 305 samples during T/C test. (a) SAC 305 jointedsamples with ENIG to Au-EP pad finish samples. (b) SAC 305 jointed sampleswith Cu-OSP to Cu-OSP pad finish samples.

Therefore, X-symmetric and Y-symmetric boundary conditionswere given.

The thermal load was applied from −40 to 125 °C, andthe element type was fully coupled temperature–displacementto analyze the problem where the simultaneous solutionof the temperature and stress/displacement fields is neces-sary [3]. A fully coupled thermal stress analysis was per-formed since the mechanical and thermal solutions affecteach other strongly and, therefore, they must be obtainedsimultaneously. Using FEM simulation, the strain distributionof the BGA solders resulting from the coefficient of thermalexpansion (CTE) mismatch was observed during temperaturecycling.

In addition, component effects such as from the FC die,embedded die, and SiP substrate were also observed usingthe ABAQUS 6.9 program. Four kinds of modeling werecompleted SiP modeling: 1) without embedded die; 2) withoutthe FC die; 3) with embedded die in SiP substrate filledonly with a core material; and 4) with only EMC moldingand SiP substrate filled only with core material, as shownin Fig. 5.

III. RESULTS AND DISCUSSION

A. T/C Reliability Test

In the case of SAC 305 jointed samples with ENIG (PCBboard side pad) to Au-EP (SiP PCB side pad) and SAC 305jointed samples with Cu-OSP to Cu-OSP (both PCB side

SAC 305

Au-EP pad

ENIG pad

SAC 305

Fig. 7. Cross-sectional scanning electron microscope (SEM) images of crackpropagation at SAC 305 BGA solder jointed samples with ENIG to Au-EPpad finish samples during a T/C test.

SAC 305

OSP pad

OSP pad

Fig. 8. Cross-sectional SEM images of crack propagation at SAC 305 BGAsolder jointed samples with Cu-OSP to Cu-OSP pad finish samples during aT/C test.

pads), the ENIG pad finish sample showed higher failurerate than Cu-OSP pad finish, and all failures were detectedon the BGA solders. Fig. 6 shows the failure rate of SAC305 with ENIG to Au-EP, and SAC 305 with Cu-OSP toCu-OSP samples. From these graphs, it was found that theinner position of BGA solders of the SiP package was morevulnerable than the corner of BGA solders, and this is due tothe position of embedded die which is located at the centerof the SiP. This result is quite different from the conventionalBGA failure case under a T/C test where the outmost cornersolder balls are the most vulnerable ones because of the largestdistance from neutral point (DNP).

Figs. 7 and 8 show the cross-sectional SEM images ofpad finish samples of SAC 305 jointed samples with ENIGto Au-EP, and SAC 305 jointed samples with Cu-OSP toCu-OSP during a T/C test. The cracks near Au-EP and ENIGpad finish propagated along intermetallic compound (IMC),but the cracks near Cu-OSP and Cu-OSP pad finish propagatedmore into the solder bulk than the IMCs.

In the case of SAC 105, BGA solder failures were alsodetected. The failure rate of these samples is shown in Fig. 9.

Page 4: IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND ...npil.kaist.ac.kr/pdf/foreign_journal/FJ_112.pdf · IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL.

628 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL. 2, NO. 4, APRIL 2012

0 500 1000 1500 2000 2500 30000

20

40

60

80

100

TC

fai

lure

rat

e(%

)

Test cycles

BGA-inner

BGA-corner

Embedded die-inner

Embedded die-corner

(a)

0 500 1000 1500 2000 2500 3000Test cycles

(b)

0

20

40

60

80

100BGA-inner

BGA-corner

Embedded die-inner

Embedded die-corner

TC

fai

lure

rat

e(%

)

Fig. 9. Failure rate of SAC 105 samples during T/C test. (a) SAC 105 jointedsamples with ENIG to Au-EP pad finish samples. (b) SAC 105 jointed sampleswith Cu-OSP to Cu-OSP pad finish samples.

SAC 305

ENIG pad

Au-EP pad

Fig. 10. Cross-sectional SEM images of BGA solder failures on SAC 105jointed samples with ENIG to Au-EP pad finish samples during T/C test.

The ENIG pad finish sample showed lower failure rate thanCu-OSP pad finish and the failures were detected on the BGAsolders. Through these graphs, it was noticed that inner andcorner BGA solder failures were more than embedded diefailures. The embedded die bump showed lower failure ratesthan BGAs in Figs. 6 and 9. This is due to the fact that theembedded bumps were surrounded by the substrate layers andtherefore were safer than BGAs.

Figs. 10 and 11 show the cross-sectional SEM imagesof SAC 105 samples during a T/C test. The ENIG pad

SAC 305

SAC 305

OSP pad

OSP pad

Fig. 11. Cross-sectional SEM images of BGA solder failures on SAC 105jointed samples with Cu-OSP to Cu-OSP pad finish samples during T/C test.

Probability Plot of ENIG105, OSP105, ENIG305, OSP305Weibull

99

9080706050403020

10

5

32

1

700

800

90010

0015

0020

0030

0040

00

Test cycle

Variable Life time2544 cycle2610 cycle2612 cycle2612 cycle

Shape3.5767.0738.27314.65

Scale2544261026122612

AD0.3810.6310.6960.253

N5

1185

P>0.250

0.0870.055

>0.250

ENIG105OSP105ENIG305OSP305

Mean time to failure

Early failure

Fig. 12. Weibull plot of all T/C test samples SAC 305 and 105 jointedsamples with ENIG to Au-EP and Cu-OSP to Cu-OSP pad finish samples.

finish samples showed failures on both ENIG and Au-EP padfinish. Also, Cu-OSP pad finish samples showed the samephenomenon.

The cracks near Au-EP and ENIG pad finish propagatedbetween IMC and the solder bulk, while the cracks nearCu-OSP to Cu-OSP pad finish propagated more into the solderbulk than in Au-EP and ENIG pad finish samples. All the BGAinner and corner solder failures showed these failure modes.

The Weibull plot of SAC 105 samples is shown in Fig. 12.To compare with SAC 305 solder balls, the Weibull plotsof both solders are combined. As seen from Fig. 12, themean times to failures (MTTFs) of the sample were compared,MTTF gives the mean time until a failure occurs after whichthe product cannot be repaired. At 63.2% failure, all samplesshow the same lifetime. Therefore, the lifetimes were com-pared with early failure, which is given by the x-intercept. Interms of early failure, SAC 305 jointed samples with Cu-OSPpad finish sample showed the best T/C reliability.

To summarize the T/C test result, SAC 305 solder withCu-OSP to Cu-OSP pad finish sample showed the best T/Creliability.

In addition, the creep property of solder bulk may affectthe T/C reliability. Fig. 13 shows the creep strain results of

Page 5: IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND ...npil.kaist.ac.kr/pdf/foreign_journal/FJ_112.pdf · IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL.

YU et al.: THERMAL CYCLING RELIABILITY OF BGA SYSTEM-IN-PACKAGE (SiP) WITH AN EMBEDDED DIE 629

TABLE I

MATERIAL PROPERTIES FOR FEM SIMULATION OF SiP WITH AN EMBEDDED DIE

Thermal conductivity(W/mmK)

Density (g/mm3) Temperature(°C)

Young’s modulus(N/mm2)

Poisson’sratio

Cu 0.393 0.0089 - 110000 0.343

EMC (Tg = 150 °C) 9.4e-04 -20 26000 0.35

200 900 0.35

FR4 1.7e-03 0.0019 - 17000 0.28

Si 0.148 0.0023 - 112400 0.28

solder 0.022 0.00736 - 45000 0.35

Core (Tg = 200 °C) 4.4e-4 0.002120 29000 0.17

250 18000 0.17

Passive 2.61e-3 0.00584 - 370000 0.22

PPG (Tg = 200 °C) 4.4e-4 0.002120 4000 0.3

250 400 0.3

Underfill (Tg = 110 °C) 1.7e-3 0.001620 10000

0.35150 200

Solder resist (Tg = 114 °C) 2.4e-4 -20 3400 0.29

150 320 0.29

Die bump layer (Tg = 200)20 4000 0.3

250 400 0.3

Saw filter die (TaO3) 4.6e-3 0.00745 80700 0.3

CTE (ppm/°C) Specific heat (J/gK) Yield strength (N/mm2) Strain

Cu 16.4 0.385 69 0/0.01/0.1 At 20 °C

EMC (Tg = 150 °C)11

- - -49

FR4 17 0.800 - -

Si 2.5 0.700 - -

20 0.22630 0

solder 45 0.01180 0.1

Core (Tg = 200 °C)14

1 - -5

Passive 7.4 0.434 - -

PPG (Tg = 200 °C)46

1 - -120

Underfill (Tg = 110 °C)25

0.8 - -93

Solder resist (Tg = 114 °C)60

- - -130

Die bump layer (Tg = 200)46120

Saw filter die (TaO3) 3

SAC 305 and 105 solders. Comparing SAC 305 and 105, thecreep strain of SAC 105 is larger than that of SAC 305. Thismeans that SAC 305 solder can be deformed less than SAC105 resulting in less crack propagation when they are exposedto 125 °C (higher temperature of the T/C test). As a result,SAC 305 solder and Cu-OSP pad finish combination showedthe highest T/C reliability among others.

B. FEM Simulation of SiP with an Embedded Die During aT/C Test

The FEM simulation of SiP with an embedded die under aT/C test was performed to find out how strain was distributedin an SiP package. The material properties of this simulationare shown in Table I.

As shown in the FEM simulation results, the whole packagewas under internal stress from CTE mismatch. To find out the

solder position at which failure can occur, the yield surfacesize of each solder was investigated. The plastic deformationis expressed in terms of the sum of the equivalent creepstrain (CEEQ) and the equivalent plastic strain (PEEQ). Thesevalues were obtained from FEM simulation results to definethe total accumulation of plastic strain. The PEEQ in a materialis a scalar variable used to represent the material’s inelasticdeformation behavior and, if the PEEQ is greater than zero,the material will yield. In addition, the CEEQ is also used todefine the amount of yield caused by a creep strain, becauseat 125 °C (the high temperature of T/C test) solders are underthe creep deformation region [5].

In this paper, three T/C test cycle modeling was simulated,which is shown in Fig. 14. The CEEQ value was moredominant than the PEEQ in the BGA solder. This means thatthe creep deformation dominates the BGA solder deformation

Page 6: IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND ...npil.kaist.ac.kr/pdf/foreign_journal/FJ_112.pdf · IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL.

630 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL. 2, NO. 4, APRIL 2012

0.08SAC 105, RFT = 25 °Cσ = 15 MPa

SAC 105, RFT = 25 °Cσ = 15 MPa

Aging Conditions

As Reflowed100 °C, 1 Month100 °C, 2 Months100 °C, 3 Months100 °C, 4 Months

Aging Conditions

As Reflowed100 °C, 1 Month100 °C, 2 Months100 °C, 3 Months100 °C, 4 Months

0.06

0.04

Stra

in, ε

0.02

0.000 2000 4000

Time, t (sec)(a)

6000 8000

0 2000 4000Time, t (sec)

(b)

6000 8000

0.08

0.06

0.04

Stra

in, ε

0.02

0.00

Fig. 13. Creep strain of SAC solders tested at room temperature after agingat 100 °C according to time. (a) SAC 105. (b) SAC 305 [4].

during a T/C test, and only the CEEQ value needs to be usedto explain the simulation data.

Fig. 15(a) shows the SAM image of the bottom surface ofthe SiP package showing the BGA solder balls positions.

Fig. 15(b) shows the CEEQ distribution result on the wholeBGA solders of SiP modeling as shown in Fig. 4(a). The solderballs at the far left row [number 1 in Fig. 15(b)] are those atthe outside edge of BGA package, and solders at right edge(the row number 9) were placed in the center of the BGApackage as shown in yellow bow in Fig. 15(a). All the solderballs were marked in numerical order from edge 1 to the centerhalf solder 9 in each row and the first, second, and third layerof each column, and then the CEEQ values were presented asdifferent colors.

In Fig. 16, the CEEQ values of row numbers 5 to 7BGA solders in the first column increased compared withthe outer lower number row solder balls, while the CEEQvalues of all the solder balls at the second and third columnsdecreased as solders moved toward the center of BGApackage.

The effects of the embedded die, FC die, and SiP substrateduring a T/C reliability test were investigated by comparingwith the SiP modeling result in Fig. 15(b). Four kinds ofmodeling were completed and simulated, as shown in Fig. 5.

First, the SiP modeling without embedded die shown inFig. 5(a) showed that the embedded die did not affect onthe CEEQ distribution as shown in Fig. 17(a). The maximumCEEQ value of SiP modeling [marked in red in the Fig. 15(b)]

0 1000 2000 3000 4000 50000.0

0.1

0.2

0.3

0.4

0.5

0.6

Time (sec)

CE

EQ

PE

EQ

CEEQ

PEEQ

BGA solder

Fig. 14. Comparison of the CEEQ and PEEQ values of a BGA solder of anSiP package during a T/C test using the ABAQUS simulation program.

(a)

(b)

1

+5.898e−01+5.407e−01+4.915e−01+4.424e−01+3.932e−01+3.441e−01+2.949e−01+2.458e−01+1.966e−01+1.475e−01+9.831e−02+4.915e−02+0.000e−00

CEEQ(Avg: 75%)

2 3 4 5 6 7 8

1st layer

2nd layer

3rd layer

9

Fig. 15. (a) SAM image of bottom surface of SiP and BGA solder ballsposition. (b) CEEQ distribution on the BGA solder balls marked in a numericalorder from edge 1 to center solder 9 during a T/C test in the 3-D quarter SiPmodeling.

was higher than that of the modeling without embedded die[marked green in Fig. 17(a)]. This means that the embeddeddie caused the increased CEEQ value resulting in more yieldof the solder balls.

Second, the SiP modeling without FC die [as shown in theFig. 5(b)] and the SiP modeling with SiP substrate filled onlycore material and embedded die [as shown in Fig. 5(c)] werealso investigated as shown in Fig. 17(b) and (c). The CEEQdistributions and maximum CEEQ values in Fig. 17(b) and (c)

Page 7: IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND ...npil.kaist.ac.kr/pdf/foreign_journal/FJ_112.pdf · IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL.

YU et al.: THERMAL CYCLING RELIABILITY OF BGA SYSTEM-IN-PACKAGE (SiP) WITH AN EMBEDDED DIE 631

0 1000 2000 3000 4000 5000

0.0

0.1

0.2

0.3

0.4

0.5

0.6

Time (sec)

CE

EQ

2

4

6

7

8

(a)

0 1000 2000 3000 4000 5000Time (sec)

(b)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

CE

EQ

2

4

5

7

8

0 1000 2000 3000 4000 5000Time (sec)

(c)

0.0

0.1

0.2

0.3

0.4

0.5

0.6

CE

EQ

3

4

6

7

8

Fig. 16. CEEQ values on each row of BGA solder balls at (a) first,(b) second, and (c) third columns during a T/C test.

did not show any difference compared with the actual FEMsimulation results of SiP modeling shown in Fig. 15(b). Theseresults mean that an FC die and SiP substrate filled only thecore material and an embedded die did not make any changeon the CEEQ value of BGA solders.

Additionally, the SiP modeling with only EMC molding andsubstrate filled core material [as shown in Fig. 5(d)] was alsosimulated. Fig. 17(d) shows a quite different behavior from theother modeling results. The maximum CEEQ value decreasedsignificantly by one-fifth as expected. Also the BGA solderposition of maximum CEEQ value was changed comparedwith that of actual SiP modeling in [6, Fig. 15(b)]. Fig. 18shows the CEEQ values of row 1 (edge) to 9 (center) of

CEEQ(Avg: 75%)

CEEQ(Avg: 75%)

CEEQ(Avg: 75%)

CEEQ(Avg: 75%)

+4.133e−01+3.788e−01+3.444e−01+3.100e−01+2.755e−01+2.411e−01+2.066e−01+1.722e−01+1.378e−01+1.033e−01+6.888e−02+3.444e−02+0.000e+00

(a)

(b)

(c)

(d)

+5.700e−01

+5.632e−01

+1.032e−02+6.897e−03+6.322e−03+5.748e−03+5.173e−03+4.500e−03+4.023e−03+3.429e−03+2.874e−03+2.299e−03+1.724e−03+1.150e−03+1.748e−03+1.000e+00

+5.163e−01+4.693e−01+4.224e−01+3.755e−01+3.285e−01+2.816e−01+2.347e−01+1.877e−01+1.408e−01+9.386e−02+4.693e−02+0.000e+00

+5.225e−01+4.750e−01+4.275e−01+3.800e−01+3.325e−01+2.850e−01+2.375e−01+1.900e−01+1.425e−01+9.500e−02+4.750e−02+0.000e+00

Fig. 17. CEEQ distribution on BGA solders in simulation modeling duringT/C test. (a) SiP modeling without an embedded die. (b) SiP modeling withouta FC die. (c) SiP modeling with PCB substrate filled only a core materialand an embedded die. (d) SiP modeling with only EMC molding and PCBsubstrate filled only a core material.

0 1000 2000 3000 4000 5000

0.000

0.001

0.002

0.003

0.004

0.005

0.006

0.007

CE

EQ

Time (sec)

12345678

9

Fig. 18. CEEQ values of the row 1 (edge) to 9 (center) of the first column ofBGA solder balls in the SiP modeling with only EMC molding and substratefilled only core.

the first column of BGA solder balls in the SiP modelingwith only EMC molding and substrate filled only core, andit is well matched with the highest strain value at the cornersite solder ball in conventional BGA package. However, inthe SiP package modeling in Fig. 4, which is similar to

Page 8: IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND ...npil.kaist.ac.kr/pdf/foreign_journal/FJ_112.pdf · IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL.

632 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL. 2, NO. 4, APRIL 2012

actual SiP package sample, the position of BGA solder undermaximum CEEQ value was changed to the inside the fifthto seventh row solder balls of the first column as shown inFig. 15(b).

These simulation results show that the complex structureof SiP package, especially the embedded die and the FC dieeffects the change of the maximum creep strain position fromthe corner solder ball of BGA to inner solder balls of theBGA layer. Also, this simulation results were well matchedwith more BGA inner solder ball failures than BGA cornerfailures during T/C test.

IV. CONCLUSION

In this paper, the T/C reliability of BGA type SiP with anembedded die was investigated.

First, in the T/C test, the SAC 305 solder with Cu-OSPpad finish sample showed the best reliability, and all failureoccurred on inner BGA solder balls rather than the cornersolder ball with the largest DNP. In addition, there was nofailure observed at the embedded die bump interconnection.The better T/C reliability was presumably due to the highercreep property of SAC 305 solder.

Second, through FEM simulation, the presence of FC and asubstrate filled only with a core material and an embedded die,instead of actual SiP substrate layer, did not make any changeon the CEEQ distribution of BGA solders and the maximumCEEQ value.

However, the SiP modeling showed that the maximumCEEQ value is located at the inner site of BGA solderballs, while the conventional BGA SiP modeling shows themaximum CEEQ value on the corner BGA solder balls. Thissimulation result is well matched with the T/C test result whichshowed that failure occurred on the inner site of BGA solderballs which is near the embedded die position.

In summary, the effect of an embedded die on the T/Creliability of an SiP was experimentally investigated, and itwas shown that the failure position of BGA solders duringT/C test was well matched with the FEM simulation.

REFERENCES

[1] D. J. Pedder, “System-in-package (SiP): A guide for electronics designengineers, SiP technology - introduction, categories and benefits,” Elec-tronics Knowledge Transfer Network, Cambridge Univ., Cambridge,U.K., TWI Rep. 18333/1/08, Oct. 2008.

[2] M. Dreiza, G. Dunn, L. Smith, N. Vijayaragavan, and J. Werner,“Package on package (PoP) stacking and board level relaibility, resultsof joint industry study,” in Proc. Int. Microelectron. Packag. Soc., 2006,pp. 1–8.

[3] ABAQUS Analysis User’s Manual Version 6.5, Hibbitt, Karlsson &Sorensen, Inc., Pawtucket, RI, 2004.

[4] Y. Zhang, Z. Cai, J. C. Suhling, P. Lall, and M. J. Bozack, “The effects ofaging temperature on SAC solder joint material behavior and reliability,”in Proc. 58th Electron. Comp. Technol. Conf., Lake Buena Vista, FL,May 2008, pp. 99–112.

[5] S. Wen, “Thermal and thermo-mechanical analyses of wire bond versus3-D packaged power electronics modules,” M.S. thesis, Dept. PowerElectron. Syst., Virginia Polytechnic Inst. State Univ., Blacksburg,1999.

[6] C. M. Liu, C. C. Lee, and K. N. Chiang, “Enhancing the reliabilityof wafer level packaging by using solder joints layout design,” IEEETrans. Comp. Packag. Technol., vol. 29, no. 4, pp. 877–885, Dec.2006.

Seon Young Yu received the B.S. degree in mate-rials science and engineering from the Universityof Seoul, Seoul, South Korea, in 2009, and theM.S. degree in materials science and engineeringfrom the Korea Advanced Institute of Science andTechnology, Daejeon, South Korea, in 2011.

Her current research interests include reliability ofembedded packages and finite element analysis.

Yong-Min Kwon received the B.S. and Ph.D.degrees in materials science and engineering fromthe Korea Advanced Institute of Science and Tech-nology, Daejeon, South Korea, in 2003 and 2011,respectively.

He joined the Lighting Package DevelopmentGroup, Samsung LED, Suwon, South Korea, in2011, as a Senior Engineer, where he is involvedwith the development of light emitting diode pack-aging technologies. His current research interestsinclude reliability improvement of flip-chip and

wafer-level packages using solder joints.

Jinsu Kim received the B.S. degree in metallurgyand the Masters degree in material science andengineering from Korea University, Seoul, Korea, in1999 and 2001.

He is currently a Senior Research Engineer withCentral Research and Development Institute, Sam-sung Electro-Mechanics Company Ltd., Suwon,South Korea. His current research interests includethe long term reliability and the accelerated life testof solder joints and power semiconductor modules.

Taesung Jeong received the B.S. degree in chemicalengineering from Dongkuk University, Seoul, SouthKorea, in 1991.

He is currently a Principal Engineer with theAdvanced Circuit Interconnection Division, Sam-sung Electro-Mechanics Company Ltd., Suwon,South Korea. His current research interests includedevice embedding technology for semiconductor andmobile application products.

Seogmoon Choi received the B.S., M.S., and Ph.D.degrees in mechanical engineering from Yonsei Uni-versity, Seoul, Korea, in 1989, 1991, and 2003,respectively.

He joined Research and Development Institute,Samsung Electro-Mechanics Company Ltd., Suwon,South Korea, where he conducted research as aSenior Research Engineer. In 1998, he joined theDr. Sang-Jo Lee’s Micromachining & Manufactur-ing Group, Yonsei University, where he conductedresearch as a Graduate Research Assistant. His

research in micromachining and microelectromechanical systems (MEMS)includes design and fabrication of sensors and microactuators. He is currentlya Principal Research Engineer with Central Research and DevelopmentInstitute, Samsung Electro-Mechanics Company Ltd. His current researchinterests include MEMS and packaging.

Page 9: IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND ...npil.kaist.ac.kr/pdf/foreign_journal/FJ_112.pdf · IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL.

YU et al.: THERMAL CYCLING RELIABILITY OF BGA SYSTEM-IN-PACKAGE (SiP) WITH AN EMBEDDED DIE 633

Kyung-Wook Paik (M’95) received the B.Sc.degree in metallurgical engineering from SeoulNational University, Seoul, South Korea, the M.Sc.degree from the Korea Advanced Institute of Scienceand Technology (KAIST), Daejeon, South Korea,and the Ph.D. degree in materials science and engi-neering from Cornell University, Ithaca, NY, in1979, 1981, and 1989, respectively.

He was a Research Scientist at KAIST from 1982to 1985, where he was responsible for the develop-ment of gold bonding wires. From 1989 to 1995,

he was a Senior Technical Staff Member with General Electric CorporateResearch and Development where he involved in interconnect multichipmodule technology and power integrated circuits packaging. He rejoined

KAIST as a Professor in the Department of Materials Science and Engineeringin 1995. He was a Visiting Professor at the Packaging Research Center,Georgia Institute of Technology, Atlanta, from March 1999 to February 2000,where he was involved in packaging education and integrated passives researchprograms. He was also a Visiting Professor with Portland State University,Portland, OR, from February to August 2005, where he worked in the areas offlip-chip polymer materials evaluation. He is currently with Nano-Packagingand Interconnect Laboratory working in the areas of flip-chip bumping andassembly, adhesives flip-chip, embedded capacitors, and display packagingtechnologies. He has published more than 80 technical papers and holds 16U.S. patents with another four pending.

Prof. Paik is a member of the International Microelectronics and PackagingSociety, SEMI, and MRS. He has been the Chairman of the Korean IEEEComponents Packaging and Manufacturing Technology Chapter since 1995.


Recommended