First Demonstration of Copper-plated Through-Package-Via (TPV) Reliability in Ultra-thin

3D Glass Interposers with Double-side Component Assembly

Kaya Demir, Saumya Gandhi, *Tomonori Ogawa, Raghu Pucha, Vanessa Smet, Venky Sundaram,

P. Markondeya Raj and Rao Tummala

3D Systems Packaging Research Center

Georgia Institute of Technology

Atlanta, GA, USA

*Asahi Glass Company (AGC), Japan

Email: kaya@gatech.edu

Abstract

This paper reports the first reliability demonstration of

copper-plated laser-drilled through-package-vias (TPV) in

ultra-thin 3D glass interposers with double-side component

assembly. This was accomplished with four major

innovations: 1) TPV geometry design for low stress, 2)

fabrication of TPV with minimum defects, 3) interconnection

and assembly process development, and 4) package design for

minimal warpage during assembly and thermal cycling test

(TCT). Finite element models were used to obtain design

guidelines for reliability and estimate critical regions of the

3D package during thermal cycling. Following the design

guidelines from modelling, 6”x6” glass interposer panels of

100µm thickness were fabricated with TPVs. Vias were

formed with 355nm UV laser at 60µm diameter. Test dies

were assembled on both sides and connected with TPVs.

Distance between the solder bump of test die and TPV was

varied in order to assess its effect on reliability. Additionally,

TPV daisy-chain coupons without dies were fabricated to

investigate the reliability of UV laser drilled TPVs in free-

standing glass. Test coupons were subjected to liquid-to-liquid

thermal cycling test between 125°C and -40°C with 5 minutes

dwell time at each temperature extremes. All TPV daisy-

chains in free-standing glass and all double-side assembly

coupons passed 1000 thermal cycles. Further thermal cycling

up to 2000 thermal cycles resulted in failures related to solder

bump-glass interposer connection. There were no failures in

TPV chain in both free-standing and double-side die

assembled test coupons as was predicted by simulations.

Introduction

Glass Packaging that Georgia Tech started in 2010, as a

new promising interposer concept, is now being widely

developed as an ideal package over silicon for reduced cost

and improved performance due to large panel processing, high

electrical resistivity, low dielectric constant, high-density I/O

capability, silicon-matched coefficient of thermal expansion

(CTE), low-loss, and availability in large and ultra-thin form

factors [1, 2]. However, glass has one major shortcoming – its

brittleness. Moreover, glass has high CTE-mismatch with

copper, raising concerns related to reliability of interposers

that are built from glass. This problem is further aggravated

by the process defects that typically form during via

formation. The earlier reports from Georgia Tech at ECTC

focused on reliability of TPVs in free-standing glass

interposers [3, 4]. The real benefits of ultra-thin glass

interposers are, however, seen with double-side or 3D

assembly of active and passive components on glass.

Assembled components on both sides introduce different

thermomechanical loadings on TPVs. Further, the process

defects during glass via formation and copper plating create

additional reliability concerns.

Several previous studies investigated the reliability of

through-silicon-vias (TSVs) and plated through holes in

organic interposers [5, 6]. However, reliability of TPVs in

glass with double-side component assembly has not yet been

studied. This paper accomplishes this through modeling,

design, test vehicle fabrication and reliability demonstration

through thermal cycle tests and failure analysis. This paper,

therefore, comprehensively addresses the thermomechanical

design and processing issues.

Mechanical Modeling and Design

This section focuses on modeling of TPVs in 3D glass

packages to assess the effect of parameters such as glass CTE,

and other material properties and via geometry on reliability

of TPVs under thermomechanical loading. AnsysTM

finite

element modeling software was used using temperature-

dependent properties of materials used for fabricating the

glass TPV test vehicles. These materials are glass, polymers

laminated onto the glass substrate, solder resist, silicon die

and lead-free solder bumps. Material properties of these are

summarized in Table 1 and temperature-dependent properties

of solder are compiled in Table 2. Bi-linear kinematic

hardening model with 170 MPa yield stress and 1034 MPa

modulus is used for copper in order to capture the plastic

deformations.

Table 1. Material Properties

Elastic

Modulus

(GPa)

Poisson’s

Ration

CTE(ppm/oC)

Glass 77 0.22 3.8

Polymer 6.9 0.3 31

Copper 121 0.3 17.3

Underfill 10.5 .261 75

Silicon 130 .28 2.7

Solder temp-

dependent

.4 22

978-1-4799-8609-5/15/$31.00 ©2015 IEEE 666 2015 Electronic Components & Technology Conference

Table 2. Temperature- Dependent Properties of Solder

Temperature

(K)

233 263 293 333 373

Modulus

(GPa)

31.8 27 20.1 16 12.1

Yield Stress

(MPa)

35 30 25 16 12

Tangent

Modulus

(MPa)

116.7 116.7 66.7 25 16.7

Geometric properties for modelling are summarized in

Table 3.

Table 3. Geometric Properties

Glass Thickness (µm) 100

Glass width / length (mm) 12

Polymer thickness (µm) 20

Solder Resist thickness (µm) 10

Solder bump width/height (µm) 100

Solder bump pitch 100

TPV Diameter / pitch (µm) 60 / 200

Die Thickness (mm) 0.6

Die width / length (mm) 5/5

For enabling double-side assembly processes, warpage

should be controlled after the first assembly on one side to

enable assembly of the second die on the other side. It is,

therefore, important to predict warpage of the glass

interposers after assembly. Finite element models in 2D were

created and simulated using the material and geometric

properties. The simulation starts with heating the system up to

underfill cure temperature followed by cool down to -55°C.

Simulation results, as illustrated in Fig. 1, show that the

maximum warpage of glass interposer is around the edge and

it is approximately 0.02mm for a 12mm square glass

interposer, which is low enough for enabling double-side

assembly.

Fig. 1. Scaled view of glass warpage after assembly (mm).

Reliability of vias in free-standing glass interposer was

investigated in previous studies [3, 4]. From simulations, it

was observed that material junctions and interfaces were the

critical regions for failure during thermal cycling. The plastic

strains in copper were low, indicating high fatigue life.

Moreover, laser drilling of glass results in ultra-small surface

defects which do not lead to cracking of glass under thermal

load. These results were confirmed experimentally for

different geometric configurations [3, 4]. In order to study the

stress around TPV in package, cut-boundary method was used

based on Saint Venant's principle. Local TPV model was

simulated using boundary conditions from a global model.

Two-dimensional models of the free-standing glass interposer

with TPV and package with an assembled die are shown in

Fig. 2.

Fig. 2. (a) Global model of tpv in free-standing glass

interposer (b) global model of TPV with die, and (c) local

model of TPV.

A sample simulation result illustrating the shear stress

distribution of the global model is shown in Fig. 3.

Fig. 3. Global model showing shear stress distribution after

cooling (MPa).

Local model of the TPV in free-standing glass interposer

and package is shown in Fig. 4, illustrating the von Mises

stress distribution around TPV with more resolution. It was

observed that, material junctions are critical regions for failure

in both cases and stress distributions are similar.

Fig. 4. Local TPV model showing Von Mises stress (MPa)

distribution after cooling (a) package (b) free-standing glass

interposer.

Fig. 5 shows the out-of-plane displacement for both TPV

in free-standing glass, and TPV with die assembly. For TPVs

Maximum warpage ~ 0.02mm

(a) (b)

(c)

Die

Polymer

GlassCopper

viavia

Critical

Regions

Die

Glass

Critical Regions

667

in free-standing interposers, displacements are dominated by

thermomechanical deformations, whereas TPV deformation in

assembled packages is mainly due to warpage.

Fig. 5. Out-of-plane (Uy) displacement in (a) (left) free-

standing TPV, and (b) (right) TPV in package.

Fig. 6 shows the distribution of radial stress around TPV in

a free-standing glass and in glass with assembled dies. Radial

stress in TPV in package is mainly compressive whereas

radial stress in TPV in glass without dies is mainly tensile. As

a result, delamination of copper from via wall and glass

cracking due to tensile stresses is more likely in a free-

standing glass panel compared to TPV in a glass package with

assembled dies

Fig. 6. Radial stress distribution in (a) (left) free standing

TPV, and (b) (right) TPV in package.

For high-bandwidth applications, short distances are

preferred to reduce the time for signal propagation from

solder bump to TPV. In order to assess the effect of this

distance, the via stresses were modeled when this parameter is

systematically decreased from 500µm to 0µm (0µm refers to

via-on-pad structure). Von Mises stress around via is

considered as the design parameter. Fig. 7 shows the change

of stress with distance. It is observed that the distance

between bump and TPV does not have much effect unless it is

a via-on-pad structure. Placing the solder bump directly on

TPV leads to higher von Mises stress around TPV.

Fig. 7. Effect of bump-TPV distance on via stress.

Fig. 8 shows the plastic strain distribution in the package.

Higher plastic deformation occurs in solder bump compared

to copper-plated TPV. Using a Coffin-Manson type equation,

lifetime of TPVs in different combinations can be roughly

estimated

pf

c

fN

75.0 (1)

c: Fatigue ductility exponent (Copper: -0.6, Solder: -0.7)

εf : Fatigue ductility coeff. (Copper: 0.3 Solder: 0.65)

Based on the Coffin-Manson’s relations and strain

amplitudes from simulation, fatigue life estimation for solder

bumps is around 1700 thermal cycles fatigue life whereas, for

copper plated TPVs, it is around 4000 thermal cycles.

Fig. 8. (a) (left) Plastic strain distribution in package, and (b)

(right) in copper plated TPV.

Fig. 9 shows that plastic strain distribution in package

when solder bump is directly on TPV. High plastic

deformation around solder bump is observed and plastic strain

values are higher compared to when solder bump is not

directly placed on TPV.

Fig. 9. Plastic strain distribution in solder bump for a pad-on-

via type structure.

Deformation due to

warpage

Deformation due to

thermal strains

Tensile Radial

Stresses

(~7MPa)

Compressive

Radial

Stresses (~12

MPa)

300

320

340

360

380

400

0 100 200 300 400 500

Vo

n M

ise

s St

ress

aro

un

d t

pv

(Mp

a)

Distance between bump and tpv (µm)

Plastic Strain in

Solder ~0.038

tpv ~ 0.03

Plastic Strain in

Solder ~0.041

668

Simulation results show that: 1) 60µm UV laser drilled in

100µm glass is expected to have high fatigue life; 2) the low

warpage of glass interposer enables double-side assembly of

dies; 3) distance between solder bump and via does not have

significant impact unless it is a via-on-pad type structure; and

4) Solder bump has higher plastic deformation compared to

TPV in glass interposer. Based on these results, TPV in glass

interposer with double-side component assembly is estimated

to be reliable. Failures are expected to be related to solder

bumps but not to TPVs.

Test Vehicle Design and Fabrication

Glass interposer test vehicles were designed based on

guidelines from the modeling study. Schematic design of test

vehicle is shown in Fig. 10. Test vehicles are based on daisy-

chains formed from top die solder bump, TPV and bottom die

solder bump. When any interconnect in this chain fails, the

whole daisy-chain fails. This design enables to test reliability

of both solder bumps and TPV simultaneously.

Fig. 10. Schematic layout of package test vehicle.

A commercial flip-chip daisy-chain test die was used for

fabricating the test vehicles (PBO8) and is shown in Fig. 11. It

has 22 solder bumps on each side. In each test coupon, there

are in total 20 TPVs that connect the top and bottom dies. As

a result, a whole daisy-chain is made up of 22 solder bumps,

10 TPVs and redistribution layers between them.

Fig. 11. PBO8 Flip-chip Daisy-chain test die layout.

Test vehicles were fabricated and subjected to reliability

tests. The TPV hole formation was performed using a 355nm

UV laser. The UV laser drilling of glass is a thermal based

process which leads to smooth via walls, which reduces

reliability concern related to stress concentrations at defects.

The glass interposers were fabricated following the processes

described in Fig. 12 as detailed in previous work [2].

Fig. 12. Glass interposer fabrication process.

After fabricating the 6”x6” glass interposer, it was

singulated into 12mm x 12mm individual coupons by

mechanical dicing. Distance between TPVs and bump pads in

each test coupon was varied in order to assess the effect of

interactions between them as shown in Fig. 13.

Fig. 13. Test vehicle surface showing via and bump pads.

Flip-chip test dies were sequentially assembled on both

sides of glass coupons by solder reflow followed by applying

underfill. Test dies were connected through TPV chains in a

way to enable reliability testing of the entire package. Sample

test vehicle is shown in Fig. 14.

Fig. 14. Fabricated test vehicle of glass substrate with

assembled die.

X-ray image of the sample test vehicle is shown in Fig.

15.Bump array and via array can be observed on this image.

Via arrayBumppads

Connected

solder bumps

Glass PanelCleaning/SilaneTreatment

Polymer lamination

Copper Protection layer deposition

TPV Formation(Excimer Or UV laser)

Copper etching And surface treatment

Electroless Cu plating

Photoresist lamination

Lithography

Electrolytic CuPlating

PhotoresistStripping

Seed layerEtching

Solder Resist Lamination

Lithography and Surface Finish

110µm

Varied 500µm-> 0µm

TPV

Bump pads

12mm

Test pads

12mm

669

Fig. 15. X-Ray image of the assembled test vehicle.

Cross-sections of test vehicle depicting solder bumps and

TPV array after fabrication are shown in Figs. 16-18.

Fig. 16. Cross-section of a representative test vehicle with

double-side assembly.

Fig. 17. Cross-section of the interposer showing the pad and

solder resist.

Fig. 18. TPV Array after double-side assembly.

Reliability Test and Failure Analysis

The fabricated test vehicles were then subjected to

accelerated liquid-to-liquid thermal cycling test between

-55°C and 125

°C with a dwell-time of 5 minutes at each

temperature extreme. In total, eight test coupons were tested.

Bump-to-TPV distance is varied from 500µm to 0µm. Along

with assembled test coupons, chains of 10 TPVs in 4 free-

standing glass coupons were subjected to accelerated

reliability test in order to observe the effect of die assembly

on tpv reliability. Test samples are summarized in Table 4.

Table 4. Tested samples and TPVs

Test Coupon

# of

coupons

Total # of

TPVs

Pass/

Fail

TPV chain in free-standing

glass

4 40 4/4

Assembled test vehicle

with dies

4 80 4/4

Electrical resistances of the TPV-solder bump daisy-chains

were monitored periodically to identify the initiation of

cycling-induced failures. Resistance of the whole chain varied

between 0.8 - 1Ω and did not change by more than 10% up to

1000 thermal cycles. The reliability characterization results

with 3D interposer samples did not show any failures up to

1000 cycles, as predicted by the models. A daisy chain hat

passed reliability test is shown in Fig.19. However, after

further thermal cycling up to 2000 thermal cycles, resistance

increase was observed on majority of chains. The samples

were characterized using optical cross-section images to

investigate the failure locations and mechanisms. The failures

were mainly attributed to solder bump cracks, with a majority

found around the bump-die interfaces. These are shown in

Fig. 20 and Fig. 21.

(a) (b)

Fig 19. (a) Cross-section of a passed saple (b) Passed

Solder bump

Fig. 20. Cross-section of a failed sample after thermal cycle

test.

Fig. 21. Failure in solder ball after thermal cycle.

There were no failures related to TPV and glass interposer

in the form of major cracking around high stressed regions.

An array of TPV after 2000 cycles is shown in Fig. 22.

Thermal cycling did not lead to any growth of defects in TPV

arrays both in free-standing glass and double-side assembled

configuration.

Top die bump array

Bottom die bump array

Copper plated vias(TPV)

Glass

Polymer

Polymer

Solder Resist

Solder Resist

glass

Cracks in solder

underfill

200µm

glass

Bottom die

200µm

670

(a)

(b)

Fig. 22. (a)Cross-section of TPV after thermal cycling test (b)

TPV after thermal cycling test

Conclusions

This paper presents the first comprehensive study on

reliability of copper-plated TPVs in ultra-thin 3D glass

interposer packages with double-side component assembly.

Thermomechanical reliability of TPVs was studied, starting

with finite element modeling for stress in glass and strains in

copper when ICs are assembled on these. These are followed

by design guidelines for reliable TPV geometries, leading to

fabrication of test vehicles and accelerated testing to validate

the models. Copper-plated TPV daisy-chains, both in free-

standing glass and in glass package with double-side

assembly, were subjected to accelerated thermal cycling tests.

All of test coupons passed 1000 thermal cycles. Further

cycling up to 2000 thermal cycles resulted in failures which

were related to solder bump interconnection failures between

glass and die. The TPV reliability exceeded the solder joint

reliability and is hence not expected to be a concern with glass

interposers.

Acknowledgements

This research was supported by the Low Cost Glass

Interposers and Packages (LGIP) Industry Consortia. The

authors would also like to thank Jason Bishop and Chris

White for their assistance with assembly and reliability test.

References

1. Rao R. Tummala et al, “Trend from ICs to 3D ICs to 3D

systems,” in Custom Integrated Circuits Conference, 2009,

pp 439-444

2. Sukumaran, Vijay, et al. "Through-package-via formation

and metallization of glass interposers," in Proc. IEEE

Electronic Components and Technol. Conf. (ECTC), 2010

3. Demir, Kaya, et al. "Thermomechanical and

electrochemical reliability of fine-pitch through-package-

copper vias (TPV) in thin glass interposers and packages,"

in Proc. IEEE Electronic Components and Technol. Conf.

(ECTC), 2013 .

4. Demir, Kaya, et al. "First demonstration of reliable

copper-plated 30μm diameter through-package-vias in

ultra-thin bare glass interposers," in Proc. IEEE Electronic

Components and Technol. Conf. (ECTC) 2014.

5. Liu, Xi, et al. "Thermo-mechanical behavior of through

silicon vias in a 3D integrated package with inter-chip

microbumps." Electronic Components and Technology

Conference (ECTC), 2011 IEEE 61st. IEEE, 2011.

6. Karmarkar, Aditya P., et al. "Material, process and

geometry effects on through-silicon via reliability and

isolation." MRS Proceedings. Vol. 1249. Cambridge

University Press, 2010.

7. Iannuzzelli, Ray. "Predicting plated-through-hole

reliability in high temperature manufacturing

processes." Electronic Components and Technology

Conference, 1991. Proceedings., 41st. IEEE, 1991.

8. Vecchio, K. S., and R. W. Hertzberg. "Analysis of long

term reliability of plated-through holes in multilayer

interconnection boards Part B: Fatigue results and fracture

mechanisms." Microelectronics Reliability 26.4 (1986):

733-751.

Glass

No cracks around critical region

671