904 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND ...folk.uio.no/olem/papers/tollefsen2013.pdf · Technology: Au–Sn SLID Bonding Torleif André Tollefsen, Andreas Larsson, Ole

904 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND MANUFACTURING TECHNOLOGY, VOL. 3, NO. 6, JUNE 2013

High Temperature Interconnect and Die AttachTechnology: Au–Sn SLID Bonding

Torleif André Tollefsen, Andreas Larsson, Ole Martin Løvvik, and Knut E. Aasmundtveit

Abstract— Au–Sn solid–liquid interdiffusion (SLID) bondingis a novel and promising interconnect and die attach technologyfor high temperature (HT) applications. In combination withsilicon carbide (SiC), Au–Sn SLID has the potential to be akey technology for the next generation of HT electronic devices.However, limited knowledge about Au–Sn SLID bonding forHT applications is a major restriction to fully realizing the HTpotential of SiC devices. Two different processing techniques—electroplating of Au/Sn layers and sandwiching of eutectic Au–Sn preform between electroplated Au layers—have been stud-ied in a simplified metallization system. The latter processwas further investigated in two different Cu/Si3N4/Cu/Ni–P/Au–Sn/Ni/Ni2Si/SiC systems (different Au-layer thickness). Die sheartests and cross-sections have been performed on as-bonded,thermally cycled, and thermally aged samples to characterizethe bonding properties associated with the different processingtechniques, metallization schemes, and environmental stress tests.A uniform Au-rich bond interface was produced (the ζ phase witha melting point of 522 °C). The importance of excess Au on bothsubstrate and chip side in the final bond is demonstrated. It isshown that Au–Sn SLID can absorb thermo-mechanical stressesinduced by large coefficient of thermal expansion mismatches (upto 12 ppm/K) in a packaging system during HT thermal cycling.The bonding strength of Au–Sn SLID is shown to be superb,exceeding 78 MPa. However, after HT thermal ageing, the ζ phasewas first converted into the more Au-rich β phase. This createdphysical contact between the Sn and Ni atoms, resulting in brittleNixSn y phases, reducing the bond strength. Density functionaltheory calculations have been performed to demonstrate that theformation of NixSn y in preference to the Au-rich Au–Sn phasesis energetically favorable.

Index Terms— Au–Sn solid–liquid interdiffusion (SLID)bonding, density functional theory (DFT), die attach, hightemperature (HT), interconnect technology.

Manuscript received April 3, 2012; revised December 13, 2012; acceptedMarch 9, 2013. Date of publication May 3, 2013; date of current versionMay 29, 2013. This work was supported in part by the HTPEP Project, theResearch Council of Norway under Project 193108/S60, Badger, SmartMotor,Fairchild, Roxar, and Norbitech. Recommended for publication by AssociateEditor T.-C. Chiu upon evaluation of reviewers’ comments.

T. A. Tollefsen is with SINTEF ICT Instrumentation, Oslo 0373, Norway,and also with Vestfold University College, Institute for Micro and Nanosys-tems Technology, Borre 3184, Norway (e-mail: [email protected]).

A. Larsson is with SINTEF ICT Instrumentation, Oslo 0373, Norway(e-mail: [email protected]).

O. M. Løvvik is with SINTEF Materials and Chemistry, Oslo 0314, Norway,and also with the Department of Physics, University of Oslo, Oslo 0318,Norway (e-mail: [email protected]).

K. E. Aasmundtveit is with Vestfold University College, Institutefor Micro and Nanosystems Technology, Borre 3184, Norway (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.2013.2253353

I. INTRODUCTION

M ICROELECTRONIC packaging plays a vital role inelectronic devices. It serves the purposes of electri-

cal interconnection, heat dissipation, mechanical support, andphysical protection [1]. The electrical performance, size, cost,and reliability is to a large degree governed by the package,which is often referred to as the bottleneck of microsystemindustry [2]. The choice of interconnection, i.e., the conductivepath required to achieve connection from a circuit element tothe rest of the circuit, is therefore of the utmost importance.

Commonly used interconnect techniques include soldersand conductive adhesives [3]. However, for high temperature(HT) applications like automotives, drilling and well inter-vention systems, aerospace, space exploration, and nuclearenvironments, the standard interconnect materials do not meetthe requirements regarding, e.g., HT stability [4]. Currently,there is no clear definition of the temperature range of a HTelectronic system. In this paper, it is considered to be above200 °C.

There is a limited range of HT interconnect techniques[4]–[6]. One alternative is sintered nano-particle Ag, whichhas good electrical and thermal conductivity [7], [8]. A nano-particle Ag joint has a high melting point (960 °C) comparedto the low processing temperature (<300 °C) [7]. However,Ag migration is reported to be a problem in HT applications(particularly in combination with high power), limiting thelifetime of the joint [9], [10].

Other prospective HT interconnect techniques includeliquid-based solder joints [11], composite solder joints [12],bismuth-based solder joints [13], and solid–liquid inter-diffusion (SLID) joints [14], [15]. Of these techniques,SLID bonding—also called transient liquid phase bonding[16], [17], isothermal solidification [18], or off-eutectic bond-ing [19]—has shown great potential [9], [19]–[21]. SLIDbonding utilizes a binary system with one HT melting metaland one low temperature melting metal (general principlesare shown in Fig. 1). The applied processing temperatureis higher than the melting point of the low melting pointmetal, and new intermetallic compounds (IMCs) are formed.The solidification is isothermal, and the final joint has ahigher melting point than the processing temperature. Thisopens a window for new subsequent manufacturing stepswithout the need for ever decreasing process temperaturesfor each step [14], [15]. Another important advantage withSLID bonding is that often a processing temperature in closeproximity to the final application temperature can be applied

2156-3950/$31.00 © 2013 IEEE

TOLLEFSEN et al.: HIGH TEMPERATURE INTERCONNECT AND DIE ATTACH TECHNOLOGY 905

Fig. 1. Schematic illustration of SLID bonding (TB bonding temperature, Tmmelting temperature, RT room temperature, IMC intermetallic compound).

(since the melting point of the final joint is much higherthan the processing temperature). This can help reduce thethermo-mechanical stresses induced by coefficient of thermalexpansion (CTE) mismatches in the final package.

SLID bonding was performed in various metal systems.Examples include Ag/In [14], [15], Ag/Sn [22], Au/In [14],[23], Au/Sn [6], [9], [19]– [21], [23]–[25], Cu/Sn [26]–[29],and Ni/Sn [28], [29]. Ag/In, which is among the first SLIDsystems [14], [15], was investigated for HT applications inseveral studies [9], [30]–[32]. Here, a HT stable joint can beachieved (stable up to 700 °C [30]) using a processing temper-ature of only 210 °C, followed by annealing at 150 °C [32].However, the HT lifetime of the joint is reported to belimited (especially in combination with high power), due to Agmigration [9], [10].

Au/Sn is a promising SLID system for HT applications [9],[19]–[21], [33], [34]. Based on the Au–Sn phase (shown inFig. 2), several Au–Sn phases can be appropriate for HTapplications. However, when long time stability is taken intoaccount, the ζ phase is the most promising. The final bondstructure is reported to be layered [21]—Au/ζ /Au—wherethe ζ ′ phase undergoes a phase transition to the ζ phaseat 190 °C [35], [36]. The ζ phase has a melting point of522 °C [35], [36], making it desirable for HT applications.Thorough investigations of the Au–Sn phase diagram suggestthat the actual layered bond structure probably is Au/ζ /Au,since the ζ phase is stable down to −5 °C (depending on Auconcentration). Another possible bond structure is Au/β/Au.The β phase has an even higher melting point than the ζphase, i.e., 532 °C [35], [36]. However, the β phase is reportedto be more brittle than the ζ phase [37], indicating that aAu/ζ /Au bond probably is more reliable than a Au/β/Au bond.Au/Sn SLID was already shown good HT stability and thermalcycling abilities in studies performed by Johnson et al. [9],[19].

Fig. 2. Au–Sn phase diagram [35], [36].

Silicon carbide (SiC), a wide bandgap semiconductor, iscommonly considered as the best alternative for the nextgeneration of innovative, high performance, cost-effective, andenvironmental-friendly drilling and well-intervention systemsfor the oil industry [38]. SiC has a high breakdown fieldstrength, a high thermal conductivity, and offers excellentperformance in HT (up to 600 °C) and high power appli-cations [39]–[41]. However, lack of qualified HT packagingtechnology is a major limitation to fully realize the potentialof SiC.

This paper presents a study of Au–Sn SLID bonding ina package utilizing a bipolar junction transistor (BJT) SiCchip. First, two different processing techniques—electroplatingof Au/Sn layers and sandwiching a eutectic Au–Sn preformbetween electroplated Au layers—are investigated in a simpli-fied metallization system to find the best technique. The latterprocess is further investigated for metallized Si3N4 substrates,in two different Cu/Si3N4/Cu/Ni–P/Au–Sn/Ni/Ni2Si/SiC sys-tems (different Au-layer thickness). Die shear tests and cross-sections are performed on the different samples to characterizethe bonding properties associated with the different processingtechniques, metallization schemes, and environmental stresstests. Furthermore, density functional theory (DFT) calcula-tions were performed to investigate the relative thermodynamicstability of different Au–Sn and Ni–Sn phases, to enhance theunderstanding of how Ni acts as a diffusion barrier betweensubstrate/chip metallization and a AuSn SLID bond.

II. METHODOLOGY

A. Test Assemblies

1) SLID 1a & b Samples: Oxidized Si wafers with sputteredTi-W (60 nm)/Au (100 nm) adhesion/seed layers were used asboth substrate (diced in 4.3 × 6.6 mm2 after plating) and chip(diced in 2 × 2 mm2 after plating) in the simplified bondingsamples, hereby referred to as SLID 1 samples. These werethen electroplated with a uniform Au layer (5 μm). The Auelectroplating was performed in a gold cyanide solution ata temperature range 60 °C–65 °C, with a current density of5.4 mA/cm2. Two different types of SLID 1 samples weremanufactured.


(a) (b)

Fig. 3. SLID 1a samples. (a) Layers as plated. (b) Expected structure afterbonding.

(a) (b)

Fig. 4. SLID 1b samples. (a) Layers as plated. (b) Expected structure afterbonding.

a) SLID 1a: Sn (2 μm)/Au (0.1 μm) layers were elec-troplated on the chip side using a tin sulfate solution atroom temperature, with a current density of 10 mA/cm2.The thin Au layer was applied to minimize oxidation,making fluxless bonding possible [24]. A Sn/Au-platedchip was then bonded to a substrate (see Fig. 3 forillustration).

b) SLID 1b: A eutectic Au 80 wt.% Sn 20 wt.% preform(7.5 μm) was sandwiched between a chip and substrateto make the joint. The preform was purchased fromMicro Joining KB (see Fig. 4 for illustration).

The bonding was performed in two steps; first, a flip chipbonder was used to pick and place at a moderate temperature(120 °C) applying a force of 35 N for 30 s. Secondly, thepositioned samples were bonded using a hotplate in a vacuumchamber and a clamping force to ensure intimate contact. Theapplied force translated to a pressure 2.5 MPa on the surfaceto be bonded. First, the samples were heated to 250 °C andheld there for 5 min (to bake out any residual moisture andto assure a uniform temperature distribution in the bondinglayers). Then, the samples were heated to 350 °C, and keptthere for 20 min to ensure that the desired phases were created.

2) SLID 2a & b Samples: Commercialy purchased Si3N4substrates (from Denka Chemicals) with active metal bondedCu (150 μm) and plated Ni–P (7 wt% P) were used assubstrates for both SLID 2a & b samples. The substrates hadsymmetrical metallization (Cu/Ni–P layers on both top andbackside) to minimize warpage of the substrate due to CTEmismatches between Si3N4 and Cu. An additional Au layer (3or 5 μm) was electroplated on the substrates in a gold cyanidesolution at a temperature range 60 °C–65 °C, with a current

(a) (b)

Fig. 5. SLID 2a & b samples. Sketch of expected layer structure afterbonding. (a) SLID 2a. (b) SLID 2b.

Fig. 6. Picture of manually aligned SLID 2a & b samples on a hot platefastened with a clamping force.

density 2.7 mA/cm2. The substrate was diced in 6 × 6 mm2

samples after plating.The BJT SiC dummy chips, delivered from Fairchild, had

sputtered Ni2Si (140 nm)/Ni (300 nm)/Au (100 nm) metal-lization. The chips were electroplated with a uniform Au layer(5 μm), and diced in 1.855×3.4 mm2 samples. Two differenttypes of SLID 2 samples were produced.

a) SLID 2a: As for SLID 1b samples, an eutectic Au80 wt.% Sn 20 wt.% preform (7.5 μm) was sandwichedbetween the chip and the substrate to make the joint.For SLID 2a samples, the electroplated Au layer on thesubstrate was only 3 μm. This means that there wouldbe no excess Au left on the substrate side of the joint[see Fig. 5(a) for illustration].

b) SLID 2b: Same as SLID 2a samples, but with 5 μmelectroplated Au on both sides, resulting in excess Auon both the substrate and the chip side in the final joint[see Fig. 5(b) for illustration].

The bonding was performed in two steps: first, the substrate,the preform, and chip were aligned manually on a hot plateand fastened with a clamping force (see Fig. 6). Secondly,the samples were bonded using the hotplate in a vacuumchamber. The same bonding profile as for SLID 1 sampleswere used. However, the final bonding time was reduced from20 to 10 min based on work published in [20].


0

50

100

150

200

0 50 100

Tem

pera

ture

(°C

)

Time (min)

Fig. 7. Cycling profile for thermal cycling. The temperature in the joint wasmeasured by a k-type thermocouple and an Agilent 34970A data acquisitionunit.

B. Test Methods and Equipment

Thermal cycling tests were peformed in a Heraeus HT7012S2 thermal cycling chamber. Both SLID 1a & b andSLID 2a & b samples were cycled between 0 °C and 200 °C,with a gradient 10 °C/min, and a dwell time 15 min atthe temperature extremes (see Fig. 7 for cycling profile—thetemperature in the joint was measured by a k-type thermocou-ple and an Agilent 34970A data acquisition unit). Two levelsof cycling were performed (500 and 1000 cycles). The cycledsamples were shear tested in a Dage 2400A shear tester witha 50 kgf load cartridge, and the results were compared tomeasurements on as-bonded samples.

SLID 1a & b and 2a & b samples were aged in air at 250 °Cfor 6 months in a Binder laboratory oven. The aged sampleswere shear tested in a Dage 2400A shear tester with a 50 kgfload cartridge, and compared to measurements on as-bondedsamples.

Cross-sectioning was performed on all groups of samples.The samples to be cross-sectioned were embedded in epoxyresin prior to grinding, and then ground (on SiC paper grade320 through 4000, using water cooling) and polished (using6 μm diamond particles and an alcohol-based lubricant priorto fine polishing using 3 and 1 μm diamond particles with awater- and oil-based lubricant). Note that SiC is a very hardmaterial compared to the Au and the Au–Sn phases, makingit difficult to prepare planar cross-sections.

The cross-sectioned samples and the fracture surfaces ofshear tested samples were investigated by optical microscopy(Neophot 32), scanning electron microscopy (SEM: JEOLJSM-5900LV) and energy-dispersive spectroscopy (EDS:Oxford X-MAX 50).

C. Ab Initio Calculations

Periodic DFT calculations were performed using theVienna ab initio simulation package (VASP) [42], [43]within the generalized gradient approximation according toPerdew et al. [44]. The projector augmented wave (PAW)method was used to treat the core regions within an all-electron framework [45]. The plane wave basis set cut-offenergy was 450 eV, and the density of the k-point gridwas at least 0.25 Å−1 for all the structures. Standard PAWpotentials were used, which implied that the following valence

0

20

40

60

80

100

SLID 1a SLID 1b

Die

shea

r stre

ngth

(MPa

)

Sample group

As bonded500 cycles1000 cycles

Fig. 8. Die shear strength of SLID 1a & b samples as a function of thermalcycles (0 °C–200 °C, 10 °C/min, dwell time of 15 min). The bars show thestandard deviation for the different groups.

Fig. 9. Optical microscopy image of a cross-section of an as-bonded SLID 1bsample showing the challenges associated with co-planarity during bonding.

orbitals were used in: 5d106s1 (Au), 5s25p2 (Sn), and 3d84s2

(Ni). The criterion for self-consistency was that the energydifference between two consecutive iterations was <10−6 eV.All structures were relaxed with a quasi-Newton method (theresidual minimization scheme with direct inversion in theiterative subspace) using a force convergence criterion of0.05 eV/Å. The atomic positions as well as the unit cell sizeand shape were relaxed simultaneously. Structural relaxationwas performed twice, before a final high-accuracy calculationof the total energy. Experimental structures from a database(Inorganic Crystal Structure Database) were used as input forthe structural relaxations [46].

III. RESULTS AND DISCUSSION

A. Thermal Cycling

1) SLID 1a & b Samples: After 500 and 1000 thermalcycles, the die shear strength was determined and comparedto the as-bonded strength. Five samples from each group weretested. The results have some clear trends (shown in Fig. 8).

a) Variations: There are large variations (standard devia-tion) in the shear strength. The reason for this mostlikely originate from the sample manufacturing. A clampforce was used to ensure sufficient contact between thebonding surfaces (see Fig. 6). The clamp force is onlyin contact with a restricted part of the surface area ofthe chip and the substrate. During the bonding process,


parts of the bonding materials will be liquified, meaningthat if the clamp force is not symmetrically placedon the chip/substrate, problems with co-planarity willoccur. This was confirmed by cross-section pictures (seeFig. 9).

b) Die shear strength: The shear strength has a substantialincrease after 500 cycles. The shear strength after 1000cycles also increased compared to the as-bonded sam-ples, but it had decreased compared to the strength after500 cycles. The explanation for the relative increase inshear strength probably stems from the challenges asso-ciated with co-planarization. During thermal cycling,the samples are regulary heated and kept at 200 °C,increasing the diffusion rate between the bonding part-ners, i.e., making a stronger joint. The reduction in theshear strength after 1000 cycles compared to 500 cyclesindicates that there is a reduction in fatigue lifetimeduring thermal cycling. However, due to the limitedamount of samples, and the large standard deviation,this should be further investigated.

c) SLID 1a versus 1b: The SLID 1b samples has higher dieshear strength than the SLID 1a samples after thermalcycling. There are no obvious reason for this trend, butcareful investigations of the cross-sections of 1a and1b samples revealed that there was a through crack inall SLID 1a samples (Fig. 10). There were no cracksin the SLID 1b samples, so the through crack in the1a samples probably explains the lower strength. Thecracks, located at the interface between the electroplatedAu and Sn layers on the chip side, probably originatedfrom contaminatons during Sn plating (e.g. on the Ausurface, not pure enough Sn, etc.). Hieber et al. [47]experienced problems regarding the pureness of theelectroplated Sn in a CuSn SLID joint. They reportedthat only e-beam evaporated Sn is pure enough for CuSnSLID bonding. This shows the importance of the Snquality. However, strong and uniform CuSn SLID jointshave been produced using electroplated Sn [26].

d) Bond configuration: The bond interface of the the goodsamples (without co-planarity problems) was uniformwith regard to IMC formation. Fig. 11 shows a opticalmicroscopy and a SEM image of the cross-section of aSLID 1b sample. EDS was used to identify the bondingphase as most probably the ζ phase. This supports ourassumption that a AuSn SLID bond most probably hasa Au/ζ /Au structure when the overall Au:Sn ratio in thebondline is designed appropriately.

e) Fracture surfaces: The fracture surfaces of sampleswith high-bond strength were located at the interfacesbetween the substrate/chip and Ti-W. The fracture sur-faces of samples with low-bond strength were locatedin the actual bond layer. This indicates that the sam-ples with low die shear strength fail because of co-planarization issues. However, the samples with high dieshear strength fail in the chip/substrate metallization,indicating that a good AuSn SLID bond has higherbond strength than 80 MPa (the highest measured bondstrength). This is also supported by previous work by

Fig. 10. Optical microscopy image of a cross-section of a SLID 1a sampleshowing the through crack in the electroplated Au/Sn interface.

TABLE I

DIE SHEAR STRENGTH OF SLID 2a SAMPLES AS A FUNCTION OF NO OF

THERMAL CYCLES (0 °C–200 °C, 10 °C/min, DWELL TIME OF 15 min)

As Bonded (MPa) 500 Cycles (MPa) 1000 Cycles (MPa)

>78 >78 >78

>78 >78 >78

>78 69.0 43.8

71.8 66.9 36.4

64.3 50.3 35.5

TABLE II

DIE SHEAR STRENGTH OF SLID 2b SAMPLES AS A FUNCTION OF NO OF

THERMAL CYCLES (0 °C–200 °C, 10 °C/min, DWELL TIME OF 15 min)

As Bonded (MPa) 500 Cycles (MPa) 1000 Cycles (MPa)

>78 >78 >78

>78 >78 >78

>78 >78 >78

>78 >78 >78

>78 >78 >78

>78 >78 >78

>78 >78 69.9

>78 >78 68.4

>78 >78 66.3

50.9 52.9 65.5

Johnson et al. [19], who reported a die shear strengthabove 90 MPa for AuSn SLID joints.

EDS analysis indicates that a AuSn SLID bond is con-structed of a layered Au/ζ /Au structure. When the joining iswell performed, high-bond strengths are achieved. Importantly,a AuSn SLID joint can withstand thermal cycling between0 °C and 200 °C in a package with small CTE mismatches(Si chip and substrate).

2) SLID 2a & b Samples: After 500 and 1000 thermalcycles, the die shear strength was determined and comparedto the as-bonded strength. Ten SLID 2b samples and five 2asamples were tested. The results have some clear trends(shown in Fig. 12, Tables I and II).

a) Variations: The standard deviation in the shear strengthwas greatly reduced compared to SLID 1a & b samples.The assembly was performed as for SLID 1a & bsamples, but experience assured a higher yield. Notethat the majority of the samples had a bond strengthabove the equipment limit (50 kgf), indicating that therecould be a considerable variation not measureable withthe applied equipment.


(a) (b)

Fig. 11. Optical microscopy image of (a) a SLID 1a sample with (b) a magnified SEM image of a representative region. The rectangles show regions whereEDS was performed. The atomic percent for the regions is included. Note that the features on the left side of the SEM image most probably stems fromcross-section preparations.

0

20

40

60

80

100

SLID 2a SLID 2b

Die

shea

r stre

ngth

(MPa

)

Sample group

As bonded 500 cycles 1000 cycles

Fig. 12. Die shear strength of SLID 2a & b samples as a function of thenumber of thermal cycles (0 °C–200 °C, 10 °C/min, dwell time 15 min). Notethat this figure is only included to visualize the main trends. Since many ofthe tested samples did not fracture during testing, due to the equipment limit(50 kgf), a proper average cannot be made.

b) Shear strength: The bond strength of SLID 2b samplesremained relatively unchanged and superb (>78 MPa)during thermal cycling (note that the majority of thesamples had a bond strength above the equipment limit,indicating that there could be a degradation in the bondstrength not measureable with the applied equipment).For SLID 2a samples, there was a decrease in the bondstrength as a function of the number of thermal cycles.In SLID 2b samples, there was excess Au on boththe substrate and the chip side (Fig. 13). In SLID 2asamples, there was no excess Au left on the substrateside (Fig. 14), causing formation of brittle Au–Ni–SnIMCs during thermal cycling. These brittle IMCs arebelieved to be the primary cause of the degradation ofthe bond strength.

Fig. 13. Optical microscopy image of a cross-section of a SLID 2b sampleshowing a uniform bondlayer, with excess Au on both substrate and chip side.The different phases were identified by SEM and EDS.

Fig. 14. Optical microscopy image of a cross-section of a SLID 2a sampleshowing a uniform bondlayer, but with excess Au on only the chip side. Onthe substrate side brittle Au–Ni–Sn phases have been created, weakening thedie shear strength. The different phases were identified by SEM and EDS.

c) Bond configuration: As for SLID 1a & b samples,the bond interface was uniform with regard to IMCformation, EDS was used to identify the bonding phaseas most probably the ζ phase.


Fig. 15. SEM images of the fracture surface of SLID 2a & b samples. Thefracture surfaces were identified with optical microscopy, scanning electronmicroscopy (SEM), and energy dispersive X-ray spectroscopy (EDS).

d) Fracture surfaces: In Fig. 15, SEM images of thefracture surfaces of thermally cycled (1000 cycles) SLID2a & b samples are shown. The fracture surface of 2bsamples was located at the chip/Ni2Si/Ni interface, againindicating that a good AuSn SLID bond has higher bondstrength than 78 MPa. The fracture surface of 2a sampleswas located at the ζ phase/Au–Ni–Sn IMCs interface,confirming that these brittle IMCs are the primary causeof the degradation of the bond strength for 2a samples.

Notably, the AuSn SLID joints withstand thermal cyclingbetween 0 °C and 200 °C for a package with large CTEmismatches (12 ppm/K difference between SiC and the thickCu film). The importance of excess Au on both substrateand chip side in the final bond is also demonstrated. Thesoft Au layer is important since it absorbs thermo-mechanicalstresses in the package, induced by, e.g., CTE mismatchesbetween the chip and the conducting layer. Excess Au onboth sides of the final joint is also important since it actsas a diffusion barrier between the Au–Sn phases and thechip/substrate metallization, which tends to develop brittleIMCs like Ni3Sn4 [48], greatly reducing the lifetime and thereliability of the package. Furthermore, excess Au is centralfor the prediction of the properties of the joint, since it assuresstable material phases with predictable/known properties.

The superb die shear strength of a AuSn SLID joint showsthat this is one of the most promising SLID bonding materialschemes for HT applications. In addition, the ability of AuSnSLID joints to absorb stress makes them very attractive.However, for applications with temperatures above ∼0.5×Tm ,

0

20

40

60

80

100

SLID 1a SLID 1b

Die

shea

r stre

ngth

(MPa

)

Sample group

As bonded Aged

Fig. 16. Die shear strength of SLID 1a & b samples as a function ofthermal ageing (250 °C, 6 months). The bars show the standard deviation forthe different groups.

TABLE III

DIE SHEAR STRENGTH OF AS BONDED AND THERMALLY AGED

(250 °C, 6 MONTHS) SLID 2a SAMPLES

As Bonded (MPa) Aged (MPa)

>78 >78

>78 >78

>78 43.8

71.8 36.4

64.3 35.5

creep is considered to be important for the long time reliabil-ity [2]. (Tm is the melting point given in absolute tempera-ture.) For a ζ phase bond, 0.5 × Tm = 125 °C, making itvulnerable for creep during long time HT applications withconstant stresses in the system (e.g., from CTE mismatches).But, by, e.g., applying a process temperature in close rangeto the operation temperature (which often is possible forSLID bonds, since the melting point of the final joint is wellabove the processing temperature), stresses induced by theCTE mismatches in the system can be minimized. The effectof creep in a Au–Sn SLID joint should be more carefullyinvestigated.

B. Thermal Ageing

1) SLID 1a & b Samples: After six months of thermalageing, the die shear strength was determined and comparedto the as-bonded strength. Five samples from each group weretested, and the results are shown in Fig. 16.

The results had the same tendency as the thermally cycledSLID 1a & b samples. There was a large standard deviation,the shear strength increased after ageing, the bond strength of1b samples was higher than that of 1a samples, and the fracturesurface of samples with high-bond strength was located atthe interfaces between the subsrate/chip and Ti-W, while thefracture surface of samples with low bond strength was locatedin the actual bond layer. For a more thorough discussion aboutthese results, see the thermal cycling test section.


TABLE IV

DIE SHEAR STRENGTH OF AS BONDED AND THERMALLY AGED

(250 °C, 6 MONTHS) SLID 2b SAMPLES

As Bonded (MPa) Aged (MPa)

>78 >78

>78 >78

>78 >78

>78 >78

>78 >78

>78 72.9

>78 66.9

>78 58.5

>78 55.9

50.9 54.1

0

20

40

60

80

100

SLID 2a SLID 2b

Die

shea

r stre

ngth

(MPa

)

Sample group

As bonded Aged

Fig. 17. Die shear strength of SLID 2a & b samples as a function ofthermal ageing (250 °C, 6 months). Note that this figure is only included tovisualize the main trends. Since many of the tested samples did not fractureduring testing, due to the equipment limit (50 kgf), a proper average cannotbe made.

2) SLID 2a & b Samples: After six months of thermalaging, the die shear strength was determined and comparedto the as-bonded strength. Five SLID 2a samples and 10SLID 2b samples were tested. The results are shown in Fig. 17,Tables III, and IV.

The results can be compared to the thermally cycledSLID 2a & b samples. However, the bond stength of theaged SLID 2a samples was reduced more than that of thethermally cycled samples. Furthermore, the bond strength ofthe SLID 2b samples wasalso reduced (the bond strength ofSLID 2b samples was not reduced during thermal cycling).Cross-sections of the thermally aged samples revealed largealterations in the bond structure compared to the as-bondedstructure. In Fig. 18, a SEM image of a cross section of an agedSLID 2a sample is shown. It reveals that all the Sn in the AuSnSLID bond has diffused into the diffusion barrier, Ni, creatingNi3Sn2 and Ni3Sn4, leaving only pure Au in the original bond.Notice that the composition analysis was performed with EDS.Due to the uncertainty in EDS analyses, and the thickenss ofthe Ni–Sn IMC layers, there probably is some remaining Snin the bond layer. As seen from the Au–Sn phase diagram inFig. 2, the Au phase can contain up to 2–3 at.% Sn in solidsolution at room temperature.

Fig. 18. SEM image of a SLID 2a sample. The composition shown wasfound by EDS analyses.

Notably, these results demonstrate that the Au/ζ /Au bondis not stable at 250 °C. Sn continues to diffuse into the pureAu layers/Au continues to diffuse into the ζ layer, creating aβ/ζ layer with no excess Au between the diffusion barrier,Ni, and the bond layer. When there is a physical contactbetween Sn and Ni, interdiffusion results in the formation ofthe brittle Ni3Sn2 and Ni3Sn4 IMCs, leaving only pure Auin the original bond (probably with some Sn). In Fig. 19, aschematic overview over the AuSn SLID bond structure atdifferent stages is shown.

C. DFT Calculations

DFT calculations utilizing VASP were performed to enhancethe understanding of interdiffusion at the Au–Sn/Ni inter-face. The enthalpy of formation for hypothetical reactionstaking place at the interface was calculated by comparing theground state electronic total energy of the various compounds.Only combinations in the relevant part of the phase diagram(Au/Ni > 0.75) were included. The results are shown inTable V and Fig. 20, and confirm that a combination of pureAu and intermetallic Ni–Sn is thermodynamically more stablethan a combination of Au–Sn and pure Ni. This is valid forall combinations of Au–Sn phases (the ζ -phase Au31Sn5, theβ-phase Au11Sn, or Au5Sn) and Ni–Sn phases (Ni3Sn2 orNi3Sn4). It is furthermore instructive to see how the enthalpyof reaction defines a convex hull of the final combination ofAu and Ni–Sn when plotted as a function of the Au/Ni ratio inFig. 20. That is, depending on the effective Au/Ni ratio, bothNi3Sn2 and Ni3Sn4 will be thermodynamically accessible inthe relevant part of the phase diagram. The local, effectiveAu/Ni ratio will depend on a number of parameters like theglobal Au/Ni ratio, the temperature profile, the morphology,other phases present, as well as the kinetic barriers of solid-state diffusion.

The calculations above were performed at the ground state,i.e., with no contributions from temperature, entropy, or zero-point motion. This means that we have in reality only probedthe phase diagram at 0 K. We have also neglected the possibil-ity of other phases in this part of the phase diagram, includingamorphous ones. Finally, the factors governing the effectiveAu/Ni ratio, mentioned above, have also been disregarded.Nevertheless, the results are consistent and conclusive enoughto rationalize the tendency of Sn to diffuse from the Au–Snphases into Ni when they are in physical contact. As long as


Fig. 19. Schematic overview of the AuSn SLID bond structure at different stages in a SLID 2b sample. Note that the As bonded and After HT ageingstructures are based on experimental observations, whereas the after TC structure is deduced as an intermediate state between the two aforementioned, basedon the phase diagram (Fig. 2).

TABLE V

GROUND STATE ENTHALPY OF REACTION Hr OF POTENTIAL CHEMICAL

REACTIONS AT THE Au–Sn/Ni INTERFACE CALCULATED BY DFT. A

NEGATIVE ENTHALPY OF REACTION INDICATES THAT THE REACTION IS

ENERGETICALLY FAVORABLE. Hr IS GIVEN IN kJ/mol PER Au ATOM

Potential Reaction Au/Ni Hr (kJ/mol)

4Au11Sn + 3Ni → 44Au + Ni3Sn4 0.94 −2.474Au31Sn5 + 15Ni → 124Au + 5Ni3Sn4 0.89 −3.192Au11Sn + 3Ni → 22Au + Ni3Sn2 0.88 −4.144Au5Sn + 3Ni → 20Au + Ni3Sn4 0.87 −3.022Au31Sn5 + 15Ni → 62Au + 5Ni3Sn2 0.81 −6.152Au5Sn + 3Ni → 10Au + Ni3Sn2 0.77 −6.70

-8-7-6-5-4-3-2-10

0,75 0,80 0,85 0,90 0,95 1,00

ΔH r

(kJ/mol)

Au/Ni

Fig. 20. Enthalpy of reaction Hr of the tentative reactions listed in Table V,as a function of the Au/Ni ratio. The enthalpy is given in kJ/mol per Auatom. The dotted line designates the convex hull of the reaction products(Ni–Sn intermetallic phases and Au) in this part of the ternary Au–Ni–Snphase diagram at 0 K.

the solid-state diffusivity is high enough for such transport totake place, there is a strong thermodynamic driving force forthe formation of intermetallic Ni–Sn.

D. Some Final Remarks

When the DFT results are seen together with the experi-mental results from the ageing of SLID 2a & b samples, wecan understand how Ni–P acts as a diffusion barrier betweenCu and Au; Sn from intermetallic Au–Sn will diffuse intoNi and form intermetallic Ni–Sn. This furthermore stressesthe importance of having a residual pure Au layer acting asa diffusion barrier between substrate and chip metallizationas well as on both sides of the final Au–Sn SLID bond. Onethen needs to take into account that the final Au–Sn SLID bondmay be a Au/ζ /Au bond. For a Au–Sn SLID bond to withstandlong-time HT exposure, the bondline should be designed witha Sn:Au ratio corresponding to max. 8 at.% Sn, to ensuresurplus Au also after transformation to a ζ phase.

It should be mentioned here that Ti seems to act as a betterdiffusion barrier than Ni for interdiffusion of Sn. The SLID 1bsamples had exactly the same bonding metallization as theSLID 2b samples. However, the SLID 1b samples used Ti-Was a diffusion barrier instead of Ni. After six months of thermalageing at 250 °C, there was no trace of Sn in the Ti-W layer.This is consistent with the work performed by Anhock et al.[49], where Ti was found to be a better diffusion barrier thanCr, Ni, Pd, and Pt.

IV. CONCLUSION

Two different AuSn SLID processing techniques were inves-tigated, where sandwiching of a eutectic Au–Sn preformbetween electroplated Au layers was found to be the preferredscheme. Initial processing issues regarding co-planarizationand through cracks in the Au/Sn interface were solved, andstrong uniform Au/ζ /Au joints were produced.

The bond strength of a Au–Sn SLID bond is shown to besuperb, >78 MPa. However, for joints without excess Au on


both substrate and chip side, the shear strength is reduced bythermal cycles.

Importantly, it is demonstrated that a Au–Sn SLID jointcan absorb thermo-mechanical stresses induced by large CTEmismatches (12 ppm/K) in a package during HT thermalcycling. However, after long-time HT exposure, the jointis transformed from Ni/Au/ζ /Au/Ni to Nix Sny /Au/NixSny ,apparently through a Ni/β/Ni stage. The primary cause of theformation of the NixSny IMCs has been investigated by DFTcalculations. The results indicate that the combination of pureAu and intermetallic Ni–Sn is energetically more favorablethan pure Ni and intermetallic Au–Sn. The Sn:Au ratio shouldthus be below 8 at.%. When the insight above is implemented,Au–Sn SLID bonding is an excellent candidate for HT dieattach and interconnect technology.

ACKNOWLEDGMENT

The authors would like to thank T. T. Luu andDr. K. Wang for manufacturing the SLID 1a&b samples and toDr. M. M. V. Taklo for proofreading.

REFERENCES

[1] R. R. Tummala and E. J. Rymaszewski, Microelectronics PackagingHandbook. New York, NY, USA: Van Nostrand, 1989.

[2] R. R. Tummala, Fundamentals of Microsystems Packaging. New York,NY, USA: McGraw-Hill, 2000.

[3] H. Oppermann, “The role of Au/Sn solder in packaging,” in Materialfor Informations Technology, E. Zschech, Ed. New York, NY, USA:Springer-Verlag, 2005.

[4] H. S. Chin, K. Y. Cheong, and A. B. Ismail, “A review on die attachmaterials for SiC-based high-temperature power devices,” Metall. Mater.Trans. B, vol. 41, no. 4, pp. 824–832, Aug. 2010.

[5] L. Coppola, D. Huff, F. Wang, P. R. Burgos, and D. Boroyevich,“Survey on high-temperature packaging materials for SiC-based powerelectronics modules,” in Proc. IEEE Power Electron. Specialists Conf.,vols. 1–6. Jun. 2007, pp. 2234–2240.

[6] P. Zheng, “High temperature electronics packaging processes and materi-als development,” Ph.D. dissertation, Dept. Technol. Eng., Auburn Univ.,Auburn, AL, USA, 2010.

[7] J. G. Bai, Z. Z. Zhang, J. Calata, N. Lu, and Q. Guo-Quan,“Low-temperature sintered nanoscale silver as a novel semiconductordevice-metallized substrate interconnect material,” IEEE Trans. Compon.Packag. Technol., vol. 29, no. 3, pp. 589–593, Sep. 2006.

[8] C. R. Chang and J. H. Jean, “Effects of silver-paste formulationon camber development during the cofiring of a silver-based, low-temperature-cofired ceramic package,” J. Amer. Ceramic Soc., vol. 81,no. 11, pp. 2805–2814, Nov. 1998.

[9] P. Zheng, P. Zheng, P. Henson, and L. Chen, “Metallurgy for SiC Dieattach for operation at 500 °C,” in Proc. Int. Conf. Exhibit. High Temp.Electron., 2010, pp. 8–17.

[10] S. A. Yang and A. Christou, “Failure model for silver electrochemicalmigration,” Ieee Trans. Device Mater. Rel., vol. 7, no. 1, pp. 188–196,Mar. 2007.

[11] S. H. Mannan and M. P. Clode, “Materials and processes for implement-ing high-temperature liquid interconnects,” IEEE Trans. Adv. Packag.,vol. 27, no. 3, pp. 508–514, Aug. 2004.

[12] G. Muralidharan, “Composite die-attach materials for high-temperaturepackaging applications,” in Proc. High Temp. Electron. Conf., 2006,pp. 593–601.

[13] Y. Takaku, I. Ohnuma, R. Kainuma, Y. Yamada, Y. Yagi, Y. Nishibe, andK. Ishida, “Development of Bi-base high-temperature Pb-free solderswith second-phase dispersion: Thermodynamic calculation, microstruc-ture, and interfacial reaction,” J. Electron. Mater., vol. 35, no. 11,pp. 1926–1932, Nov. 2006.

[14] L. Bernstein, “Semiconductor joining by solid-liquid-interdiffusion(SLID) process,” J. Electrochem. Soc., vol. 113, no. 12, pp. 1282–1288,Dec. 1966.

[15] L. Bernstein and H. Bartholomew, “Applications of solid-liquid interdif-fusion (SLID) bonding in integrated-circuit fabricatoin,” Trans. Metall.Soc. AIME, vol. 236, no. 3, pp. 405–411, 1966.

[16] D. S. Duvall, W. A. Owczarshi, and D. F. Paulonis, “TLP bonding:A new method for joining heat resisant alloys,” IEEE J., vol. 53, no. 4,pp. 203–214, Apr. 1974.

[17] W. D. Macdonald and T. W. Eagar, “Transient liquid-phase bonding,”Ann. Rev. Mater. Sci., vol. 22, no. 1, pp. 23–46, Aug. 1992.

[18] R. Schmid-Fetzer, Design Fundamentals of High-Temperature Com-posites, Intermetallics and Metal-Ceramic Systems, R. Y. Lin, Ed.Warrendale, PA, USA: TMS, 1995, pp. 75–97.

[19] W. R. Johnson, C. Wang, Y. Liu, and D. J. Scifield, “Power devicepackaging technologies for extreme environments,” IEEE Trans. Elec-tron. Packag. Manuf., vol. 30, no. 3, pp. 182–193, Jul. 2007.

[20] K. E. Aasmundtveit, K. Y. Wang, N. D. Høivik, J. M. Graff, andA. Elfving, “Au-sn SLID bonding: Fluxless bonding with high temper-ature stability to above 350 °C,” in Proc. Eur. Microelectron. Packag.Conf., Jun. 2009, pp. 723–728.

[21] K. Aasmundtveit, T. T. Luu, H.-V. Nguyen, R. Johannessen,N. D. Høivik, K. Y. Wang, “Au-Sn fluxless SLID bonding: Effect ofbonding temperature for stability at high temperature, above 400 °C,”in Proc. Electron. Syst. Integr. Technol. Conf., Sep. 2010, pp. 1–6.

[22] J. F. Li, P. A. Agyakwa, and C. M. Johnson, “Kinetics of Ag3Sn growthin Ag-Sn-Ag system during transient liquid phase soldering process,”Acta Mater., vol. 58, no. 9, pp. 3429–3443, May 2010.

[23] C. L. Lee, Y. W. Chen, and M. Goran, “Advances in bonding tech-nology for electronic packaging,” J. Electron. Packag., vol. 115, no. 2,pp. 201–207, Jun. 1993.

[24] K. Y. Wang, K. E. Aasmundtveit, and H. Jakobsen, “Surface evolutionand bonding properties of electroplated Au/Sn/Au,” in Proc. 2nd Elec-tron. Syst., Integr. Technol. Conf., vols. 1–2. Sep. 2008, pp. 1131-1134.

[25] G. S. Matijasevic, C. C. Lee, and C. Y. Wang, “Au-Sn alloy phase-diagram and properties related to its use as a bonding medium,” ThinSolid Films, vol. 223, no. 2, pp. 276–287, Feb. 1993.

[26] N. D. Høivik, K. E. Aasmundtveit, G. Salomonsen, A. Lapadatu,G. Kittilsland, and B. Stark, “Fluxless wafer-level Cu-Sn bonding formicro-and nanosystems packaging,” in Proc. Present. Electron. Syst.Integr. Technol. Conf., Sep. 2010, pp. 1–5.

[27] H. Huebner, S. Penka, B. Barchmann, M. Eigner, W. Gruber, M. Nobis,S. Janka, G. Kristen, and M. Schneegans, “Microcontacts with sub-30 mu m pitch for 3D chip-on-chip integration,” Microelectron. Eng.,vol. 83, nos. 11–12, pp. 2155–2162, Nov.-Dec. 2006.

[28] S. Bader, W. Gust, and H. Hieber, “Rapid formation of intermetalliccompounds by interdiffusion in the Cu-Sn and Ni-Sn systems,” ActaMetall. Et Mater., vol. 43, no. 1, pp. 329–337, Jan. 1995.

[29] P. F. Yang, Y.-S. Lai, S.-R. Jian, J. Chen, and R.-S. Chen, “Nanoinden-tation identifications of mechanical properties Of Cu6Sn5, Cu3Sn, andNi3Sn4 intermetallic compounds derived by diffusion couples,” Mater.Sci. Eng. A, Struct. Mater. Properties Microstruct. Process., vol. 485,nos. 1–2, pp. 305–310, Jun. 2008.

[30] W. W. So and C. C. Lee, “High temperature joints manufactured at lowtemperature,” in Proc. IEEE 48th Electron. Compon. Technol. Conf.,May 1998, pp. 284–291.

[31] R. W. Chuang and C. C. Lee, “Silver-indium joints produced at low tem-perature for high temperature devices,” IEEE Trans. Compon. Packag.Technol., vol. 25, no. 9, pp. 453–458, Sep. 2002.

[32] P. Quintero, “Development of a shifting melting point ag-in paste viatransient liquid phase sintering for high temperature environments,”Ph.D. dissertation, Dept. Mech. Eng., Univ. Maryland, College Park,Md, USA, 2008.

[33] T. A. Tollefsen, A. Larsson, O. M. Løvvik, and K. Aasmundtveit,“Au-Sn SLID Bonding—Properties and possibilities,” Metall. Mater.Trans. B, vol. 43, no. 2, pp. 397–405, Apr. 2012.

[34] T. A. Tollefsen, A. Larsson, O. M. Løvvik, and K. Aasmundtveit,“Au-Sn SLID bonding for high temperature applications,” in Proc.IMAPS High Temp. Electron. Network Conf., 2011, pp. 58–67.

[35] H. Okamoto, “Au-Sn (Gold-Tin),” J. Phase Equilibria Diffusion, vol. 28,p. 490, Oct. 2007.

[36] H. S. Liu, C. L. Liu, K. Ishida, and Z. P. Jin, “Thermodynamicmodeling of the Au-In-Sn system,” J. Electron. Mater., vol. 32, no. 11,pp. 1290–1296, Nov. 2003.

[37] R. R. Chromik, D.-N. Wanga, A. Shugara, L. Limataa, M. R. Notisa,and R. P. Vincia, “Mechanical properties of intermetallic compoundsin the Au-Sn system,” J. Mater. Res., vol. 20, no. 8, pp. 2161–2172,Aug. 2005.


[38] R. Johannessen, “High Temperature Power Electronic Packaging forOil Well Applications,” in Proc. IMAPS High Temp. Electron. NetworkConf., Sep. 2009, pp. 1–9.

[39] Z. J. Shen, “High temperature, high power module design for widebandgap semiconducters: Packaging architecture and materials consid-erations,” in Proc. High Impact Technol. Exchange Conf., May 2008,pp. 1–6.

[40] M. N. Yoder, “Wide bandgap semiconductor materials and devices,”IEEE Trans. Electron Devices, vol. 43, no. 10, pp. 1633–1636,Oct. 1996.

[41] N. G. Wright, B. H. Alton, and K. Vassilevski, “Prospects for SiCelectronics and sensors,” Mater. Today, vol. 11, nos. 1–2, pp. 16–21,Jan.–Feb. 2008.

[42] G. Kresse and J. Hafner, “Abinitio molecular-dynamics for liquid-metals,” Phys. Rev. B, vol. 47, no. 1, pp. 558–561, Jan. 1993.

[43] G. Kresse and J. Furthmuller, “Efficient iterative schemes for ab initiototal-energy calculations using a plane-wave basis set,” Phys. Rev. B,vol. 54, no. 10, pp. 11169–11186, Oct. 1996.

[44] J. P. Perdew, J. A. Chevary, and S. H. Vosko, “Atoms, molecules, solids,and surfaces: Applications of the generalized gradient approximation forexchange and correlation,” Phys. Rev. B, vol. 46, no. 11, pp. 6671–6687,Sep. 1992.

[45] G. Kresse and D. Joubert, “From ultrasoft pseudopotentials to theprojector augmented-wave method,” Phys. Rev. B, vol. 59, no. 3,pp. 1758–1775, Jan. 1999.

[46] FIZ Karlsruhe Inorganic Crystal Structure Database (ICSD). (2010,Aug. 26) [Online]. Available: http://www.fiz-karlsruhe.de/icsd.html

[47] H. Hieber, “Heat-resistant contacts with use of liquid phase transition,”in Proc. 7th Eur. Hybrid Microelectron. Conf., 1993, pp. 7–13.

[48] D. R. Frear, Mechanical-Behavior Of Solder Joint Interfacial Inter-metallics. Materials Park, OH, USA: ASM International, 1991.

[49] S. Anhock, H. H. Oppermann, C. Kallmayer, R. Aschenbrenner,L. K. Thomas, and C. H. Reichl, “Investigations of Au/Sn alloys ondifferent end-metallizations for high temperature applications,” in Proc.IEEE Int. Electron. Manuf. Technol. Symp., Apr. 1998, pp. 156–165.

Torleif André Tollefsen was born in Kristiansand,Norway, in 1978. He received the M.Sc. degreein material physics from the University of Oslo,Oslo, Norway, in 2008. He is currently pursuingthe Ph.D. degree in applied micro- and nanosystemswith Vestfold University College, Borre, Norway.

He was a System Engineer at FMC Technologies,Bergen, Norway, from 2008 to 2010, working withsystems for the subsea production of oil and gas.In 2010, he joined the Research Institute SINTEFand Vestfold University College to pursue a Ph.D.

degree. From 2013, he has been a Research Scientist at SINTEF. He hasauthored more than 25 journal and conference papers. His current interestsinclude packaging for micro and nanosystems, with a special focus onmaterials for harsh environments.

Andreas Larsson was born in Eskilstuna, Sweden,in 1980. He received the M.Sc. degree in engineeringphysics–applied physics from Uppsala University,Uppsala, Sweden, in 2006.

He has been a Structural Engineer with the Nor-wegian Oil and Gas industry since 2006, wherehe performed finite element analysis and conceptdevelopment. Since 2008, he has been with theResearch Institute SINTEF, currently as a SeniorScientist responsible for harsh environment pack-aging. He has authored more than 40 publications

and holds patents. His current research interests include packaging andinterconnect technology for harsh environments, with the important topicsbeing multiphysics simulations and concept development.

Ole Martin Løvvik was born in Norway in 1968.He received the M.Sc. degree in theoretical physicsand the dr.scient. degree in energy physics from theUniversity of Oslo (UiO), Oslo, Norway, in 1993and 1998, respectively.

He was a Post-Doctoral Researcher at UiO from1999 to 2006 and the Institute for Energy Technol-ogy, Kjeller, Norway. He was an Associate Professorwith the Norwegian University of Life Sciences, As,Norway, in 2001, a Consultant with the ResearchCouncil of Norway, Oslo, in 2006, a Research Fel-

low with the Institute for Energy Technology, Kjeller, from 2006 to 2008,a Visiting Scientist with the University of Hiroshima, Hiroshima, Japan, in2008, and a Visiting Professor with Osaka University, Osaka, Japan, in 2012.He is currently a Senior Researcher with the research institute SINTEF andan Adjunct Professor with the University of Oslo. He has authored around65 papers in international journals with referee, and has given more than 90invited and contributed talks at international conferences. His current researchinterests include atomic-scale modeling of materials for energy technologies,with the important topics being electronic materials, materials for hydrogentechnology, and light weight metals.

Knut E. Aasmundtveit received the M.S. degreein technical physics from the Norwegian Instituteof Technology, Trondheim, Norway, in 1994, andthe Ph.D. degree in materials physics from theNorwegian University of Science and Technology,Trondheim, in 1999.

He was with Alcatel Space Norway/Ame Space(now Kongsberg Norspace), Horten, Norway, as aRadio Frequency System Design Engineer, until2004, when he joined the Department of Microand Nano Systems Technology, Vestfold University

College, Borre, Norway, as an Associate Professor. He has authored or co-authored four book chapters and more than 60 journal and conference papers.His current research interests and teaching activities include packaging andintegration technology for micro and nano systems, such as intermetallicbonding, polymer-based bonding, nanomaterials integration in microsystems,and biomedical packaging.

Documents

904 IEEE TRANSACTIONS ON COMPONENTS, PACKAGING AND ...folk.uio.no/olem/papers/tollefsen2013.pdf · Technology: Au–Sn SLID Bonding Torleif André Tollefsen, Andreas Larsson, Ole