adhesion and reliability of direct cu metallization of ... and reliability of direct cu...

5
Adhesion and Reliability of Direct Cu Metallization of Through-Package Vias in Glass Interposers Timothy Huang, Venky Sundaram, P. Markondeya Raj, Himani Sharma, and Rao Tummala 3D Systems Packaging Research Center 813 Ferst Drive NW Atlanta, GA 30332-0250 [email protected] Abstract Direct metallization of bare glass with copper is required to reach the full potential low-cost benefit of glass interposers. However, this poses a fundamental materials challenge associated with copper-to-glass adhesion. Intermediate polymer liners on glass have been used by others, adding an extra material and processing step. In this paper, three approaches to direct metallization of copper to glass interposers are explored and reported. Electroless plating, sputtering followed by electrolytic plating, and sol-gel were investigated as Cu deposition methods with an emphasis on adhesion and reliability of copper to bare glass. The adhesion and reliability performance of films were characterized by tape-testing, peel- strength measurements, and thermal-shock testing. Based on these results, individual assessments are made for each approach and compared with others to assess future directions. Introduction The need for high logic-memory bandwidth in both mobile and high-performance applications with emerging 2.5D and 3D package architectures has been driving new advances in interposer and substrate materials. The semiconductor industry is primarily focusing on Si as the next-generation interposer material to address the challenges associated with dimensional instability and coefficient of thermal expansion (CTE) mismatch with organic packages. Based on recent pioneering technical advances by Georgia Tech, glass has been demonstrated as the best next-generation interposer and package material to meet the high-I/O, high-performance, and low-cost requirements for advanced packaging [1]. In addition, because of glass’s matched coefficient of thermal expansion (CTE) (few ppm/°C) with that of Si IC (2.5 ppm/°C), thermomechanical reliability is expected to be superior when compared to organic substrates of much higher CTE (18-22 ppm/°C), where CTE mismatch causes serious reliability concerns. Compared to organic packages, glass can be manufactured with a higher density of through-vias and with higher wiring density and I/Os to the chip because of its Si-like dimensional stability, and in addition, demonstrating better electrical performance and lower loss while scalable to large panels with high manufacturing through-put. One of the major challenges with glass, however, is its metallization with copper conductors with sufficient adhesion. Due to the differences in their chemical structures, most metallic materials such as Cu do not bond strongly to oxide networks of silicate glasses. This results in insufficient adhesion and failure to meet reliability criteria. One approach that has been demonstrated to circumvent this is to coat both sides of glass with a polymer film before the glass through package vias (TPVs) are made. After the creation of vias by excimer laser ablation, the polymer surface and the roughened- glass via sidewalls can be metallized by standard electroless processes [2]. The major disadvantage of this approach is its limited compatibility with via-formation methods. Glass manufacturers have recently developed proprietary TPV formation processes, some of which are not fully compatible with polymer-coated glass. Direct metallization of copper on smooth glass surfaces without polymer also provides more opportunities for lower cost, miniaturization as well as reliability and electrical performance improvements. In general, covalent-bonded oxide networks such as glass bond directly to materials composed of oxides or oxide interfaces that are stable. Since Cu is composed of metallic- bonded atoms, some surface modification that results in an oxide formation can be expected to result in enhanced bonding. Surface modifications can be physical (e.g., mechanical interlocking) or chemical, with the latter being ideal. For good chemical adhesion between bare copper metal and glass, fundamental adhesion theory states that it is important to have an intermediate metal oxide interface for strong and lasting adhesion. Based on this fundamental principle, a process that enhances chemical bonding through gradual, continuous material-property changes in the transition from glass to metal oxide to metal is preferred over mechanical-based methods to achieve strong adhesion. For example, it is well known that certain, reactive metals with high oxygen affinity (e.g., Cr, Ti, Zr, Mg) exhibit strong bonding “directly” to glass by forming thin oxide layers at the interfaces [3,4]. These oxides are very stable even at high temperatures and in reducing gas atmospheres. While standard vacuum deposition techniques benefit from this approach by depositing an adhesion layer such as Ti between Cu and glass, new innovations [5,6] are required to achieve this intermediate “adherence oxide” layer with wet metallization techniques as described next. Electroless copper deposition is a standard wet metallization process typically used for organic substrates. It is a relatively low-cost, wet processing method involving a series of steps which includes chemical surface treatments, deposition of catalyst (typically Pd/Sn) particles, and electroless copper reduction on the catalyst particles. This creates three separate, ionic-bonded interfaces between the glass (negatively charged OH - surface), conditioner (positively charged polymer electrolyte), and catalyst (negatively charged surface) (Fig. 1). Except for the bonding between the Pd catalyst and the electroless-deposited Cu, each of the interfaces between glass and catalyst rely on ionic bonding. To improve adhesion, surface etching and roughening is a typical part of electroless plating for organic substrates. Adhesion is improved through increased surface area and mechanical interlocking with the roughened surface. However, increasing the surface roughness 978-1-4799-2407-3/14/$31.00 ©2014 IEEE 2266 2014 Electronic Components & Technology Conference

Upload: nguyenhanh

Post on 02-May-2018

225 views

Category:

Documents


2 download

TRANSCRIPT

Adhesion and Reliability of Direct Cu Metallization of Through-Package Vias in Glass

Interposers

Timothy Huang, Venky Sundaram, P. Markondeya Raj, Himani Sharma, and Rao Tummala

3D Systems Packaging Research Center

813 Ferst Drive NW

Atlanta, GA 30332-0250

[email protected]

Abstract

Direct metallization of bare glass with copper is required to

reach the full potential low-cost benefit of glass interposers.

However, this poses a fundamental materials challenge

associated with copper-to-glass adhesion. Intermediate

polymer liners on glass have been used by others, adding an

extra material and processing step. In this paper, three

approaches to direct metallization of copper to glass interposers

are explored and reported. Electroless plating, sputtering

followed by electrolytic plating, and sol-gel were investigated

as Cu deposition methods with an emphasis on adhesion and

reliability of copper to bare glass. The adhesion and reliability

performance of films were characterized by tape-testing, peel-

strength measurements, and thermal-shock testing. Based on

these results, individual assessments are made for each

approach and compared with others to assess future directions.

Introduction

The need for high logic-memory bandwidth in both mobile

and high-performance applications with emerging 2.5D and 3D

package architectures has been driving new advances in

interposer and substrate materials. The semiconductor industry

is primarily focusing on Si as the next-generation interposer

material to address the challenges associated with dimensional

instability and coefficient of thermal expansion (CTE)

mismatch with organic packages. Based on recent pioneering

technical advances by Georgia Tech, glass has been

demonstrated as the best next-generation interposer and

package material to meet the high-I/O, high-performance, and

low-cost requirements for advanced packaging [1]. In addition,

because of glass’s matched coefficient of thermal expansion

(CTE) (few ppm/°C) with that of Si IC (2.5 ppm/°C),

thermomechanical reliability is expected to be superior when

compared to organic substrates of much higher CTE (18-22

ppm/°C), where CTE mismatch causes serious reliability

concerns. Compared to organic packages, glass can be

manufactured with a higher density of through-vias and with

higher wiring density and I/Os to the chip because of its Si-like

dimensional stability, and in addition, demonstrating better

electrical performance and lower loss while scalable to large

panels with high manufacturing through-put.

One of the major challenges with glass, however, is its

metallization with copper conductors with sufficient adhesion.

Due to the differences in their chemical structures, most

metallic materials such as Cu do not bond strongly to oxide

networks of silicate glasses. This results in insufficient

adhesion and failure to meet reliability criteria. One approach

that has been demonstrated to circumvent this is to coat both

sides of glass with a polymer film before the glass through

package vias (TPVs) are made. After the creation of vias by

excimer laser ablation, the polymer surface and the roughened-

glass via sidewalls can be metallized by standard electroless

processes [2]. The major disadvantage of this approach is its

limited compatibility with via-formation methods. Glass

manufacturers have recently developed proprietary TPV

formation processes, some of which are not fully compatible

with polymer-coated glass. Direct metallization of copper on

smooth glass surfaces without polymer also provides more

opportunities for lower cost, miniaturization as well as

reliability and electrical performance improvements.

In general, covalent-bonded oxide networks such as glass

bond directly to materials composed of oxides or oxide

interfaces that are stable. Since Cu is composed of metallic-

bonded atoms, some surface modification that results in an

oxide formation can be expected to result in enhanced bonding.

Surface modifications can be physical (e.g., mechanical

interlocking) or chemical, with the latter being ideal. For good

chemical adhesion between bare copper metal and glass,

fundamental adhesion theory states that it is important to have

an intermediate metal oxide interface for strong and lasting

adhesion. Based on this fundamental principle, a process that

enhances chemical bonding through gradual, continuous

material-property changes in the transition from glass to metal

oxide to metal is preferred over mechanical-based methods to

achieve strong adhesion. For example, it is well known that

certain, reactive metals with high oxygen affinity (e.g., Cr, Ti,

Zr, Mg) exhibit strong bonding “directly” to glass by forming

thin oxide layers at the interfaces [3,4]. These oxides are very

stable even at high temperatures and in reducing gas

atmospheres. While standard vacuum deposition techniques

benefit from this approach by depositing an adhesion layer such

as Ti between Cu and glass, new innovations [5,6] are required

to achieve this intermediate “adherence oxide” layer with wet

metallization techniques as described next.

Electroless copper deposition is a standard wet

metallization process typically used for organic substrates. It is

a relatively low-cost, wet processing method involving a series

of steps which includes chemical surface treatments, deposition

of catalyst (typically Pd/Sn) particles, and electroless copper

reduction on the catalyst particles. This creates three separate,

ionic-bonded interfaces between the glass (negatively charged

OH- surface), conditioner (positively charged polymer

electrolyte), and catalyst (negatively charged surface) (Fig. 1).

Except for the bonding between the Pd catalyst and the

electroless-deposited Cu, each of the interfaces between glass

and catalyst rely on ionic bonding. To improve adhesion,

surface etching and roughening is a typical part of electroless

plating for organic substrates. Adhesion is improved through

increased surface area and mechanical interlocking with the

roughened surface. However, increasing the surface roughness

978-1-4799-2407-3/14/$31.00 ©2014 IEEE 2266 2014 Electronic Components & Technology Conference

of the glass-metal interface will result in higher conductor

losses, especially at higher frequencies due to the skin-effect

[7]. Additionally, glass roughening processes are difficult to

control precisely and can introduce flaws or cracks, resulting in

increased risk of mechanical failure. Since chemical, covalent

bonds are not directly made between the Cu and the glass oxide

network, glass surface roughness is mandatory for achieving

sufficient adhesion. To maximize electrical performance and

mechanical integrity, it is therefore of interest to find the

minimum roughness required to achieve sufficient adhesion

through roughening for electroless plating. Companies such as

Atotech, Inc. are pioneering novel surface treatments and

nanoparticle-based copper deposition approaches to achieve

metal-glass adhesion with minimal roughness [5,6].

Figure 1. Schematic of chemical interfaces between

electroless Cu and glass (not to scale).

In the MCM-D (MultiChip Module – Deposition) and flat-

panel display industries, sputtering is a standard metallization

process for ceramic substrates and large panels of bare glass.

High-performance packaging requires thicker metal lines than

what sputtering alone can achieve cost-effectively. The

approach in this paper uses sputtered adhesion (Ti) and seed

(Cu) layers, which are subsequently electroplated at high

deposition rates to reach the necessary Cu thickness. Since

glass/Ti and Ti/Cu are known to be strong interfaces

individually, the glass/Ti/Cu composite structure is also

expected to be strong.

Sol-gel is a common approach to deposit thin films due to

its versatility with available chemistries and its relative

simplicity as a process. In this wet deposition process, a metal-

organic precursor solution (sol) is deposited onto the substrate.

Because of the low viscosity of the solutions, the substrate can

be dip-coated, spin-coated, or sprayed with controlled

thickness to form a three-dimensional network (gel). The film

is then subjected to a temperature high enough to pyrolyze the

organic components, resulting in a three-dimensional network

of the metal oxide. Finally, it is sintered at a higher temperature

to crystallize the film, creating a glass/CuOx/Cu structure.

This paper investigates three approaches to metallize glass

directly, without relying on organic materials that may hinder

subsequent processing: electroless Cu deposition, sputtering

with electrolytic plating, and sol-gel synthesis. For sputtering

and sol-gel, the resulting glass-metal oxide-metal structure is

expected to demonstrate strong adhesion. As the electroless Cu

does not introduce an oxide interface, weaker adhesion is

expected. Based on the results from each of the approaches

described above, individual assessments are made in terms of

potential feasibility and future directions.

Experimental Methods

The three approaches to metallize glass are briefly

described here. The key characterization techniques employed

in this work are also outlined. Smooth and roughened glass-

substrates were provided by Corning, Inc. and Life BioScience,

Inc. Roughened samples were characterized by atomic force

microscopy (Veeco AFM) prior to processing to obtain

roughness (Ra), and ranged from < 1 nm to 0.559 m.

Electroless processes were used to deposit electroless

copper. The electroless-plating steps are substrate cleaning,

conditioning (deposition of polymer electrolyte), catalyst

deposition, catalyst activation, and finally electroless copper

deposition. Triple rinses in water were performed between each

step. Electroless deposition was performed until 0.2 m of Cu

was deposited. Finally, samples were annealed at 155°C for 30

minutes in air. Tape-tests were then performed according to

IPC-TM-650 test standards. For samples which failed the tape-

test, the substrate and peeled interfaces were analyzed by X-ray

photoelectron spectroscopy (XPS) (Thermo K-Alpha XPS) for

the presence of Cu, Si, and C [8].

Sputtering was performed by Tango Systems, Inc. Ti (100

nm) was first deposited as an oxide-forming adhesion layer,

followed by Cu deposition to various thicknesses. Samples

were then electroplated with a current density of 2.0 ASD

(amperes per square decimeter) until the desired thickness was

reached, as measured by a Cu gauge. This was followed by

annealing at 155°C for 30 minutes in air. Lines and spaces of

1.0 and 0.5 cm were patterned by photolithography and

subtractive etching (Transene Company, Inc.) for peel-testing.

Peel-testing was performed at a pull-rate of 12 in/min. Thermal

shock testing was performed according to JEDEC Standard

JESD22-A106B Condition C.

Cu sol-gel solution was prepared by dissolving Cu-ethoxide

in 2-methoxyethanol and acetic acid. The solution was then

spin-coated on glass and baked at 300°C. This was repeated

until a sufficient thickness was deposited. It was then placed in

a rapid thermal processing (RTP) chamber in a forming gas

atmosphere at 560°C for 10 minutes for Cu metal

crystallization in the bulk, and simultaneous copper oxide

interface formation with glass. Characterization of the metal

and glass-metal interfaces was performed with X-ray

diffraction (XRD) and Scanning Electron Microscopy (SEM).

This paper shows preliminary results of using the sol-gel route

to deposit a copper-oxide film directly on glass, followed by

thermally converting the oxide film to metallic copper.

Results and Discussion

Electroless Cu Approach: Results of the tape-test of

electroless Cu on glass are shown in Fig. 2, in the order of

increasing adhesion. It can be seen that there is a regime of

surface roughness values (Ra) where the structure of peeled

copper changes. In the samples with worst adhesion, Cu outside

of the tape dimensions were peeled, as can be seen by the

jagged peels in Fig. 2a,b. As the adhesion increases, the copper

film does not peel beyond the edge of the tape (Fig. 2c), and

then smaller amounts of Cu peel from the glass (Fig. 2d-f).

2267

Finally, the glass chemically roughened to 0.559 μm passed the

tape test, exhibiting no peeled Cu (Fig. 2g).

(a) (b) (c)

(d) (e)

(f) (g)

Figure 2. Tape tests for electroless copper on various glasses:

(a) Corning SGW8.5 (Ra = 0.5 nm) (b) Corning SGW3 (Ra =

1.2 nm) (c) Corning SGW3 (Ra = 0.9 nm) (d) Corning

SGW8.5 (Ra = 1.3 nm) (e) Corning SGW8.5 (Ra = 19 nm) (f)

Corning SGW8.5 (Ra = 46 nm) (g) Life BioScience APEX (Ra

= 0.559 µm).

The large range in tape-test results for similar roughness

values (Fig. 2a-d) can be attributed to several factors.

Roughness values were measured with AFM and could vary

depending on the selected area because the topography was

generally not homogenous in the areas inspected (10x10 μm

and 30x30 μm). The values in this range are within the

instrumental error. In addition to surface roughness, adhesion

to glass depends on the glass chemical composition. High CTE

glasses (SGW8.5) are composed of more glass network-

modifying ions than low CTE glasses (SGW3); as these ions

are positively charged, they can assist with adhesion to the Pd

catalysts.

Representative XPS elemental Pd and survey scans of the

peeled Cu and the exposed glass are shown in Fig. 3. It is

apparent that Pd was detected only on the peeled Cu, while C

was detected only on the glass surface. Assuming that the only

source of carbon in the system is the polymer electrolyte

conditioner, this suggests that the weakest interface in the

electroless copper-glass is that between the conditioner and the

Pd catalyst.

With the current electroless process, a surface roughness on

the order of hundreds of nanometers is necessary to achieve

sufficient adhesion to pass the tape test. Based on the skin-

effect, Brist et al. [7] modeled the effects of surface roughness

on electrical loss and introduces an additional electrical loss

factor Ksr, where a value of 1 represents the factor for smooth

surfaces (i.e., no additional loss). At an electrical frequency of

10 GHz, surface roughnesses of 0.2 μm and 0.6 μm result in Ksr

values of ~1.1 and 1.55, respectively [7]; i.e., Cu with 0.6 μm

roughness will have 55% more loss (dB/length) than smooth

Cu.

Figure 3. XPS elemental Pd and survey scans of the peeled

Cu and the exposed glass.

Sputtering Approach: To investigate the effects of the

sputtered Cu seed-layer thickness and total thickness of the Cu

layer, three different seed layer thicknesses (200, 500, and 800

nm) were sputtered on unroughened glass. Each film was

electroplated to a total Cu thickness of 5 µm and 10 µm, for a

total of six different samples. Peel-tests of these samples

showed no significant correlation with the seed-layer thickness

for the ranges investigated.

Peel-testing for smooth and roughened (Ra = 10, 50, 100,

200, 300, 400, 500, 1000 nm) glass electroplated to 10 µm Cu

was performed before and after thermal shock testing (5 µm

thick Cu strips tore during peel testing, and could not be

measured). Prior to thermal shock testing, smooth glass

exhibited peel strength values over 0.2 kg/cm (Fig. 4). On the

other hand, the Cu lines on all roughened glass samples could

not be peeled from the glass and therefore exceeded the

strength limits that the technique can measure. After 1000

thermal shock cycles, the smooth glass was again the only

peelable sample, and did not show any significant change in

peel strength when compared to those from pre-thermal shock

testing.

In the microelectronics packaging industry, a typical

benchmark for the peel strength is 0.7 kg/cm for Cu thickness

of 30 µm on organic substrates. The peel strength 𝑊𝑜 of a metal

strip as a function of film thickness 𝑡𝑠 has been modeled by

Bikerman to follow the relationship [9]:

𝑊𝑜 ∝ 𝑡𝑠3 4⁄ Equation (1)

Using this relationship, a comparable peel strength value for Cu

plated to 10 μm thickness was calculated to be 0.18 kg/cm. As

seen in Fig. 4, the peel strengths before and after thermal shock

testing were over 0.25 kg/cm, well-above this scaled

benchmark value.

2268

Figure 4. Peel strength of 10 µm Cu on smooth glass before

and after 1000 thermal shock cycles.

Sol-gel Approach: The steps of Cu film deposition by sol-

gel are shown in Fig. 5. The resulting Cu film is continuous,

and the adhesion was sufficient to pass the tape test. The film

was confirmed to be reduced to the metallic state by XRD (Fig.

6).

(a) (b) (c)

Figure 5. Cu deposited on glass by sol-gel after (a) spin-

coating, (b) baking, and (c) rapid thermal processing.

Figure 6. X-ray diffraction scan of sol-gel Cu film.

While the film appeared to be continuous across the glass

surface, SEM images show that the Cu film thickness was not

uniform (Fig. 7). The non-uniform film thickness is partly due

to insufficient wetting of the Cu solution onto the glass surface

during spin-coating, which can lead to minor islanding effects.

Additionally, the Cu solution showed some cloudiness,

indicating that some Cu hydroxyl-complexes precipitated as

particles in the solution, adding to the non-uniformity of the

resulting film. Further optimization of the sol-gel process to

increase glass-wettability and Cu solution quality should result

in more uniform and higher-quality Cu films.

Figure 7. SEM image of cross section of Cu on glass

deposited by sol-gel.

Conclusion

Three approaches for direct copper metallization of glass

substrates were investigated for the resulting metal-to-copper

adhesion.

With electroless deposition, the effect of glass surface

roughness on adhesion strength was analyzed with tape-test

results. The required roughness for mechanical adhesion of

copper on glass is of the order of hundreds of nanometers,

which will negatively affect the electrical performance at high

frequencies. Analysis of the electroless Cu/glass interface

through XPS suggests that the weakest interface is the ionic

bond between the polymer electrolyte and the Pd catalyst

particles.

With sputtered Ti/Cu, the seed-layer thickness had no

measurable effect on the resulting peel strength. Sputtering on

smooth glass, as known previously, met the peel-strength

performance targets before and after thermal shock testing.

Ti/Cu sputtered on roughened-glass exceeded the peel-strength

measurement capability of the technique even after thermal

shock testing. Therefore, reliability has been demonstrated for

sputtering on smooth and roughened glass.

Cu oxide deposited with sol-gel and reduced to metallic

copper passed the tape test, but exhibited non-uniform Cu

films. Enhancements to sol-gel processing can improve the

quality of Cu films.

As expected from fundamental glass-to-metal bonding

theory with stable oxides at the interface, Cu deposited by

sputtering and sol-gel showed significantly greater adhesion to

smooth glass than electroless plating because of the presence

of an intermediate “adherence oxide”. The stability of this Cu

oxide is the subject of future work.

Acknowledgments

The authors would like to acknowledge Jason Bishop and

Chris White for their guidance and support in sample

fabrication, as well as Dibyajat Mishra for XPS data acquisition

and Yuya Suzuki for assistance in peel strength analysis.

References

1. V. Sukumaran, “Low-cost thin glass interposers as a

superior alternative to silicon and organic interposers for

packaging of 3-D ICs,” IEEE Trans. CPMT, vol. 2, no. 9,

pp. 1426-1433, Sept. 2012.

20

200

2000

40 42 44 46 48 50 52 54

Inte

nsi

ty (

Arb

itra

ry U

nit

s)

2 theta

2269

2. V. Sukumaran, “Through-package-via formation and

metallization of glass interposers,” in Proc. IEEE

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

Vegas, NV, Jun. 1-4, 2010, pp. 557-563.

3. P. Benjamin, “The adhesion of evaporated metal films on

glass,” Proc. R. Soc. Lond. A, vol. 261, no. 1307, pp 516-

531, May 1961.

4. N. Jiang, “Observations of reaction zones at

chromium/oxide glass interfaces,” J. Appl. Phys., vol.87,

no. 8, pp. 3768-3776, Apr. 2000.

5. S. Bamberg, “Novel wet chemical metallization for glass

interposers”, 3D Materials and Processes Session,

International Conference on Device Packaging,

Scottsdale/Fountain Hills, AZ, Mar. 7-10, 2011.

6. R. Taylor, “Challenges of Adhesion Promotion for

Metallization of Glass Interposers”in Global Interposer

Technology Workshop (GIT), Atlanta, GA, Nov. 17-20,

2013.

7. G. Brist, “Non-classical conductor losses due to copper

foil roughness and treatment,” in Proc. Electron. Circuits

World Conv. (ECWC), Anaheim, CA, Feb. 22-24, 2005,

pp. 22-24.

8. X. Cui, “Copper deposition and patterning for glass

substrate manufacture,” in IEEE Trans. Electronics

Packaging Technol. Conf. (EPTC), Singapore, Dec. 10-12,

2007, pp. 37-42.

9. C. Jouwersma, “On the theory of peeling,” Journal of

Polymer Science, vol. 45, issue 145, pp. 253-255, July

1960.

2270