You are on page 1of 9

Microelectronics Reliability 44 (2004) 823–831

www.elsevier.com/locate/microrel

Investigation on bondability and reliability of


UV-curable adhesive joints for stable mechanical properties
in photonic device packaging
a,b a,*
C.W. Tan , Y.C. Chan , H.P. Chan a, N.W. Leung b, C.K. So b

a
Department of Electronic Engineering, City University of Hong Kong, 83, Tat Chee Avenue, Kowloon, Hong Kong
b
Photonic Manufacturing Service Ltd., Suites 2207-11, 22/F, Tower 1, The Gateway, 25 Canton Road, TST, Kowloon, Hong Kong
Received 19 June 2003; received in revised form 1 October 2003

Abstract
The demand for volume deployment of photonic components has increased, and with it the need to effectively
manufacture in a reproducible and cost effective way. Therefore, it is important to keep the assembly process consistent
and stable. A simple shear test setup was used in this study to determine the shear force that can be used as an instant
indicator to the process stability of fiber arrays assembly. The shear test was then used to study the effect of curing
conditions, surface roughness, and reliability test on the adhesion of the adhesive joints of single channel fiber arrays. In
general, shear force increases with the curing conditions. However, thermal-induced residue stress during improper
controlled curing process might lead to loss in ductility and reliability of the joints. Results showed that rough bonding
surface for the coverlid can provide better wettability and enhances the shear strength. However, it might result in
incomplete polymerization due to transmission loss of UV radiation during assembly process that was caused by the
uneven surface. Furthermore, rough surface can affect the flow of adhesive that could result in incomplete-fills. Adhesives
at the surrounding of V-grooves and edges of the joints were attacked by moisture at elevated temperature. This paper
also discussed in details the possible failure mechanism of adhesive joints after temperature and humidity test.
Ó 2003 Published by Elsevier Ltd.

1. Introduction more advantages in terms of mass production though


they are typically not reworkable [1,2].
The packaging of photonic components represents Epoxies have been used widely as adhesives, coatings,
unique requirements and challenges. The major chal- underfills and encapsulants in the production of elec-
lenge that is uniquely associated with photonic packag- tronic and optoelectronics devices. Epoxy adhesives are
ing relates to the extremely tight assembly tolerance. In used because of their fluidity in uncured stage, which
order to achieve and maintain alignment, it is impor- allows them to flow easily into complex shapes, but once
tant to pay special attention to the bond joints and fully cured the highly cross-linked three-dimensional
joining techniques. Soldering, laser welding and adhe- networks have superior adhesive and thermo-mechanical
sive bonding are the three most popular joining tech- properties [3,4]. The resulting networks permanently affix
niques in photonic packaging. Compensation offset can the components; and the intractability of these materials
minimize or eliminate bond shift. Epoxies lend them- is desirable from a reliability aspect. Because of their
selves to shrinkage compensation, design flexibility and epoxide, hydroxyl, amine and other polar groups, the
epoxies have high specific adhesion to glass substrate.
Epoxies cure with only a fraction of the shrinkage of
vinyl type adhesives. Also, the epoxies do not pull away
*
Corresponding author. Tel.: +852-2788-7130; fax: +852- from glass fibers.
2788-7579. Many photonic devices go into underground
E-mail address: eeycchan@cityu.edu.hk (Y.C. Chan). or undersea optical network, so device reliability is
0026-2714/$ - see front matter Ó 2003 Published by Elsevier Ltd.
doi:10.1016/j.microrel.2003.10.003
824 C.W. Tan et al. / Microelectronics Reliability 44 (2004) 823–831

Table 1
Curing conditions used in this experiment
Run Condition Power Time (s)
(mW/cm2 )
1 UV post-cure 1700 60
2 UV-85 °C 1700 60
3 UV 2000 60
4 UV 2000 30
5 UV 1700 60
6 UV 1700 30
7 UV 1700 15
8 UV 1500 60
Fig. 1. Cross-sectional view of a single channel fiber array. 9 UV 1500 30
10 UV 1000 60
11 UV 1000 30
paramount. In order to optimize their uses, it is neces-
sary to understand their behavior when exposed to cli-
matic and environmental aggressions. In particular, this surface, and surfaces induced by sand papers of 400 and
study reports on the mechanical stability of adhesive 600 grit sizes. Scan length of 500 lm with scan speed of
joint in fiber array. V-grooves on the fiber block are to 50 lm/s and stylus force of 66.8 mg was used in this
provide pitch alignment of multiple optical fiber and experiment.
epoxy adhesive is hereof used to form joining between
coverlid and fiber block in order to fix optical fiber in 2.3. Reliability test
position. Based on the three-point contact principal as
shown in Fig. 1, misalignment due to curing shrinkage The single channel fiber arrays that were prepared
can be ruled out. using curing condition 2 in Table 1 were subjected to a
This paper reports the performances, i.e., bondability temperature and humidity test, i.e., 85 °C and 85% RH
and reliability of epoxy adhesive joints. The effects of for 0, 25, 50, 75, 100, 125, and 150 h respectively.
curing condition and surface roughness to adhesion Chemical and structural changes of the cured and aged
strength and mechanical properties of epoxy joints were adhesive specimens were identified by comparing their
also reported. IR spectra obtained by Fourier transform infrared
spectroscopy (FT-IR). Perkin–Elmer Spectrum ONE
FT-IR spectrometer equipped with an attenuated total
2. Experimental procedures reflectance (ATR) of 4 cm1 resolution was utilized to
assemble the spectra range from 4000 to 650 cm1 of the
Fiber blocks with single channel V-groove and its samples. In quantitative analysis, relative absorbance
coverlids were prepared using Toshiba USM-20A Slic- intensities of the molecular bonds and groups were
ing machine. Epoxy-based UV-curable adhesive was determined. The corresponding relative intensities of the
used to affix the single channel optical fiber in between cured samples and the aged samples were compared. The
fiber block and coverlid. This model was selected be- base line method was used to calculate the absorbance
cause it represents most of the construction of adhesive intensities of different groups and molecules. The rela-
joint in optical devices, especially passive devices. tive absorbance intensities of any two peaks were cal-
culated as the ratio of their intensities. The relative
2.1. Curing conditions intensity, Er between the reactive peak, Ereactive and ref-
erence peak, Eref would be calculated using the following
The curing process of this adhesive consists of a rapid equation [5]:
polymerization initiated by UV-reactive photo-initiators Ereactive
and two stages of slow thermal polymerization, i.e., 85 Er ¼ : ð1Þ
Eref
°C for 2 h and followed by 60 °C for 10 h. These samples
were prepared by using various curing conditions, as The 1437 cm1 band which is due to CH2 deformation
shown in Table 1. vibration was used as the internal reference peak.

2.2. Surface roughness, Ra, induced by different grit size 2.4. Shear test

Alpha-Step Profiler Model 500 was used to determine The lid shear test was carried out by using an IN-
the surface roughness, Ra, of the glossy surface, abrasive STRON Mini 44 Tester with a cross-head speed of 5
C.W. Tan et al. / Microelectronics Reliability 44 (2004) 823–831 825

mm/min on 20 samples [6]. The shear blade was placed Adhesive


approximately 0.5 mm from the top surface of fiber Cover lid 1
2
block. The shear blade was moving in the X -direction of 3
Jacket Fiber optic 4
the cross-sectional view of Fig. 1. Micro-structural 5
analyses of the fractured surface of sheared-samples Fiber block
were then obtained by using a Philips XL 40 FEG
scanning electron microscope (SEM).
Fig. 3. A cross-sectional view of the fiber array showing the
possible failure modes in location 1––break at coverlid, location
2––break at adhesive/coverlid interface, location 3––cohesive
3. Results and discussion failure, location 4––break at adhesive/fiber block interface and
location 5––break at fiber block.
3.1. Shear force

As the manufacturing of photonic device revolves the epoxy or at the interface. There are five potentially
into mass production mode, the consistency of the observed failure modes in this experiment as listed below
assembly process is becoming relatively important. and shown in Fig. 3:
Therefore, a reliable in situ monitor system is required in
addition to an on-line visual inspection. The purpose of 1. break at coverlid,
this shear test set-up is to provide immediate indication 2. break at adhesive/coverlid interface which is a adhe-
to line-personnel whether the process and/or the process sive failure,
parameters are operating within specifications. By 3. break at adhesive layer which is a cohesive failure,
referring to six-sigma standard, upper control limit 4. break at adhesive/fiber block interface which is an-
(UCL) and lower control limit (LCL) have been deter- other adhesive failure and,
mined instead of only an average value, as shown in Fig. 5. break at fiber block.
2. The UCL was determined to be 126.9 N and LCL was
35.8 N in which an average value of 81.4 N falls within As a result, all of these 20 samples have demonstrated
this range. These values can be used as the process the failure mode of number 2. Obviously, this bonding
control of the assembly process. Fig. 2 also shows the interface between adhesive and coverlid is the weakest
distribution of the shear force of 20 samples. among the five interfaces mentioned above.
Failure modes of sheared samples are remarkably as When these fracture surfaces were subjected to SEM
important as the shear force value. When the adhesive is examination, it was found that the optical fiber was still
properly cured, the cohesive strength within the glue line sitting in the V-groove and being covered by adhesive on
is great, and adhesion of the epoxy to other materials top of it. This observation can be clearly seen in Fig. 4a
should be so good, that failure under stress condition and b. There were adhesive stains and small quantity of
should often occur in one of the adherents rather than in adhesive remnants, mostly around the location of fiber
optic and V-groove were remaining on the coverlid, as
shown in Fig. 4c. These observations clearly showed that
140 this joining was fractured at the adhesive/coverlid
UCL
interface after shear test.
120

100 3.2. Curing process


Shear force, N

80 The degree of polymerization of adhesive strongly


Average
depends on the curing conditions and parameters as
60 tabulated in Table 1. Fiber arrays were assembled using
these curing conditions and followed by shear test. Fig. 5
40 LCL shows the required shear force-to-fracture of these
adhesive joints that were assembled using the conditions
20 in Table 1. Generally, the higher the UV radiation
power, the higher the shear force obtained. The longer
0
dwell time also contributes to a higher resistance to an
0 5 10 15 20
Sample
external shear mechanism. There are more adhesives
observed on the coverlid after the shear test for samples
Fig. 2. Shear force obtained from 20 samples that were that were cured at a lower curing degree. The failure
assembled at the optimized curing conditions. mode has shifted to a cohesion failure from an adhesive
826 C.W. Tan et al. / Microelectronics Reliability 44 (2004) 823–831

Fig. 4. SEM micrograph show the fracture surface at (a) top view at the fiber block, (b) enlarged view of (a), and (c) at coverlid.

90

80

70
Shear force, N

60

50

40

30
0 1 2 3 4 5 6 7 8 9 10 11 12
Run
Fig. 6. Shear force-vs.-displacement curves with the fracto-
Fig. 5. Shear force strength obtained from samples assembled
graph of samples prepared under different curing conditions.
under various curing conditions.

failure with increasing degree of cure. This indicates that the three curing conditions, i.e., UV curing only, UV
lower the curing degree, the less adhesive has been curing plus curing temperature of 85 °C, and UV plus
solidified that result in lower resistance to the force in post-cure at lower temperature for longer time frame.
shear direction. Since the curing condition of run 3 did The shear force-vs.-displacement curve of samples that
not increase the shear force significantly compared to did undergo UV curing only shows that the plastic
condition of run 6, the latter was chosen as the opti- deformation has taken place before the breaking point.
mized parameters for the rest of this study. A relatively long displacement, approximately 0.7 mm
The effects of post-cure and heat treatment on top of was observed before the adhesive joints reached the
the photo-initiation of UV radiation have significantly breaking point. In the attached SEM micrograph, it is
increased the shear force of fiber arrays. Fig. 6 compares clearly shown that the fracture surface is relatively
C.W. Tan et al. / Microelectronics Reliability 44 (2004) 823–831 827

rough, more plastic deformations and dimples are visi- sandpaper with grit size of 600, 400, and the abrasive
ble, thus they are less brittle compared to the joints with surface processed by vendor.
the other two curing conditions. Fig. 8 shows that rougher surface can provide
For samples that were exposed to UV radiation and stronger interfacial bonding to the adhesive, thus a
85 °C of heat curing, the shear force-vs.-displacement higher shear force is required to remove the coverlids
curve is much sharper. Though it broke at higher shear from the fiber arrays. Rough surface undoubtedly con-
force if compared to samples cured only with UV, tributed to a better adhesion force. However, rougher
shorter displacement was observed. It was observed that bonding surfaces of the coverlids will lead to uneven
the fracture surfaces of these samples were extremely surfaces of the adhesive, as shown in Fig 9. This rough
brittle. This reveals that the adhesive joints produced
under this condition might have suffered from a certain
240
extent of stresses such as thermal expansion stress, shear force 3500
220
shrinkage stress, and other forms of residual stress due surface roughness
200 3000
to non-uniform polymerization.
180
Among all samples the post-cured ones provide the 2500

Shear force, N

Ra, angstrom
160
smoothest curve and highest shear force. In addition, the 140
2000
fracture surface showed a less brittle surface with very 120
little plastic deformation. This indicates that the post- 100 1500
curing condition might have removed the residual stress 80
at the joints induced by the exposure to 85 °C during 1000
60
first stage of post-curing. Therefore, in order to get the 40
500
strongest and most reliable adhesive joints out of this 20
curing process, the condition with post-curing at lower 0 0

ABRASIVE SURFACE

SD 400

SD 600
temperature with longer time if favorable. Glossy SURFACE

3.3. Surface roughness

Four types of coverlid bonding surfaces were exam-


ined by using SEM as shown in Fig. 7. The smoothest is Fig. 8. Roughness of surfaces that were produced by different
the glossy surface, followed by surface produced by lapping processes.

Fig. 7. SEM micrographs of surfaces with various surface roughness (a) glossy surface, (b) abrasive surface, (c) by SD 600, and (d) by
SD 400.
828 C.W. Tan et al. / Microelectronics Reliability 44 (2004) 823–831

3.4. Reliability test

The effect of the temperature and humidity test, i.e.,


85 °C and 85% RH, is obvious where the shear force of
fiber arrays degrades from 81.4 N before the test to
approximately 15.2 N after 150 h of testing, as shown in
Fig. 10. The shear forces of samples that were exposed to
the reliability test for 50 , 75 and 100 h have marginally
exceeded the LCL limit that was determined in the first
part of this paper. However, it is obvious that samples
after an exposition of 125 and more hours to the test
condition were relatively weak.
Table 2 shows that the relative absorbance intensity
of the hydroxyl group is in increasing trend with the test
time increases. A diminutive decrease in the ester char-
acteristic peak was observed at 1730 and 1180 cm1 , as
Fig. 9. SEM micrograph of top surface of fiber block after
shear test. The rough morphology of adhesive is caused by the shown in Fig. 11. A slight increase in the hydroxyl group
abrasive bonding surface of coverlid. at 3450 cm1 and in the carboxylic group at 1058 cm1 ,
and a corresponding drop in the ester peak at 1730 and
surface will affect the effectiveness of UV radiation 1180 cm1 support the idea of hydrolysis [3,8–10] in the
transmission and result in an inconsistent curing degree sample which was exposed to the humidity test. There-
which might affect the local core pitch. In Fig. 9, some fore, the degradation in adhesion strength that we have
adhesive remained on surrounding of the exposed fiber observed in Fig. 10 is probably not due to the chemical
optic which are suspected as the uncured portion of
adhesive. In addition, 40% of the failure mode is cohe-
sion failure and greater amount of the remnants on the
90
bonding surface of coverlids was observed.
Fig. 9 also shows that the gap at the surrounding of 80
the top contact point of the optic fiber and the cover is
not perfectly filled as observed at the polished bonding 70
surface of the coverlid. It has been reported by Wake [7]
60
Shear force, N

that rough surface can affect the spread of adhesive ei-


ther because the adhesive cannot penetrate into the pits 50
or because it penetrates too slow and sets before it
40
completes the penetration. These observations might
result in misalignment and coupling efficiency loss of the 30
fiber optic assembly.
The abrasive surface is not the preference in mass 20
production of fiber arrays because of difficulties found in
10
maintaining its cleanliness and an extra grinding or
lapping process is required. Thus, additional cost would -20 0 20 40 60 80 100 120 140 160
be caused. Therefore, a rougher bonding surface of the Test time, hour
coverlid should be avoided, especially in mass produc-
tion since the polished surface is able to provide sufficient Fig. 10. Shear force-vs.-temperature humidity test time curve.
adhesion strength. In order to improve and further en-
hance the adhesion strength of the cut edges of the cov-
erlid with the adhesive at the fillet, one may produce Table 2
coverlids’ cut edges with moderate roughness. This can Relative absorbance intensity of hydroxyl at different reliability
test time
be achieved by choosing a blade with right grit size for
dicing and flat-grooving. Though it provides relatively Test time (h) Relative absorbance intensity of hydroxyl
good shear adhesion, a blade with a grit size of 400 will 0 9.716
cause undesired chipping along the edges, especially on 25 9.875
quartz substrates. Therefore, a dicing blade with a grit 50 9.962
size of 600 was selected for the dicing process. The similar 75 10.497
design was applied to the surface finishing of a flat- 100 10.494
125 10.708
groove that forms adhesion with the adhesive fillet, too.
C.W. Tan et al. / Microelectronics Reliability 44 (2004) 823–831 829

Fig. 11. Characteristic bands of ester (a) (mðCAOÞ  1180 cm1 ) and (b) (mðC@OÞ  1730 cm1 ).

Fig. 12. SEM micrographs of the top surface of fiber block after shear test for samples that have underwent temperature and humidity
test for (a) 0 h, (b) 50 h, (c) 100 h, and (d) 125 h.
830 C.W. Tan et al. / Microelectronics Reliability 44 (2004) 823–831

Fig. 13. SEM micrographs of bonding surfaces of coverlids after shear test for samples that have underwent temperature and humidity
test for (a) 0 h, (b) 50 h, and (c) 125 h.

structure degradation of the bulk adhesive. Con- the experimental results and observations obtained in
versely, it should be due to the degradation of the this study, it can be concluded as below:
adhesion strength at the interface within coverlid and (a) A shear test was setup and the shear force can be
adhesive. used as internal process control to provide immediate
Fig. 12 shows the hydrolysis marks (craters) at the indication of the process stability. The mean is approx-
surrounding of the exposed optical fiber. It begins with imately 81.4 N, while the upper control limit is about
bubble-like vacant; then these vacant were connected and 126.9 N and the lower control limit is about 35.8 N.
spread slowly. In Fig. 13 the moisture penetration path Only one failure mode was observed, a break at an
can clearly be seen which begins from the V-groove, i.e., adhesive/coverlid interface.
location of the optical fiber and the edges of the bonding. (b) Generally, the shear force increases with the cur-
The dark areas are the plasticizing mark of the adhesive. ing degree. Extreme curing condition, i.e., elevated
It has been well known that the bonding interface has temperature and short dwell time at post-curing, caused
the lowest surface energy if compared to the bulk mate- the adhesive joints losing their ductility. They became
rials. It is believed that moisture was penetrating via more brittle and relatively unreliable.
the bonding interface which lead to degradation of the (c) The surface roughness, Ra, has strongly enhanced
adhesion strength at the bonding interface. In view of the strength of the adhesive joints. However, in the
the fact that the adhesive/fiber block interface has been meantime it was affecting the curing uniformity and
strengthened by the fiber optic that was clamped firmly spread of adhesive that have huge impact to the coupling
by V-groove and the cured adhesive, the coverlid/adhe- efficiency. In addition, rough surface will cause flow
sive interface remains as the weakest point at this con- difficulty, thus affecting the fill up efficiency of adhesive.
struction. Selectively, this layer will consistently be So, rough surface or uneven surface should be avoided
observed as the failure mode, unless there is inconsis- in adhesive joints.
tency in this process or in the experimental setup. (d) The strength of the adhesive joint can be further
enhanced by using a coverlid with a rougher cutting edge
and a rougher flat-groove of the fiber block which can
form a stronger adhesion to the adhesive fillet.
4. Conclusions (e) Temperature and humidity tests have apparently
degraded the adhesion strength due to moisture
A simple shear test experimental setup was used to absorption, and FTIR results suggested that the joint
study the effect of curing conditions (thermal history), has experienced hydrolytic attacks when an increase in
surface roughness and reliability test on the adhesion carboxylic and hydroxyl group but a corresponding
strength of the adhesive joint in fiber arrays. Based on decrease in ester peaks was observed. The V-groove/
C.W. Tan et al. / Microelectronics Reliability 44 (2004) 823–831 831

adhesive interface was found to be one of the favorite vices: problems in device packaging. Opt Fiber Technol
moisture penetration paths apart from the edges that 1995;1:283–8.
result in hydrolysis attack to the adhesive at the sur- [3] Chen JS, Ober CK, Poliks MD. Characterization of
rounding of optical fiber. Hence, up to a certain extent it thermally re-workable thermosets: materials for environ-
mentally friendly processing and re-use. Polymer 2001;43:
will loosen the fastening mechanism and potentially lead
131–9.
to coupling efficiency loss. [4] Cho JD, Kim EO, Kim HK, Hong JW. An investigation of
the surface properties and curing behavior of photocurable
cationic films photosensitized by antracene. Polym Test
Acknowledgements 2002;21:781–91.
[5] Tan CW, Chiu YW, Chan YC. Corrosion study of
The authors would like to acknowledge the grant anisotropic conductive joints on polyimide flexible circuits.
from the Government of the Hong Kong Special Mat Sci Eng B 2003;98(3):255–64.
Administrative Region––Innovation and Technology [6] Tan CW, Chan YC, Yeung NH. Behavior of anisotropic
conductive joints under mechanical loading. Microelectron
Fund (ITF), University–Industry Collaboration Pro-
Reliab 2003;43(3):481–6.
gram (Project no. UIT/033, and CityU ref. 9440025). [7] Wake WC. Adhesion and the formulation of adhesives.
Thanks to Ms. Tan Sai Choo, Mr. Jack Wang and Mr. 2nd ed. London: Elsevier Applied Science Publishers LTD;
Sean Peng for their technical supports. 1986.
[8] Park CE, Han BJ, Bair BE. Humidity effects on adhesion
strength between solder ball and epoxy underfills. Polymer
References 1997;38(15):3811–8.
[9] Money L, Balali R, Vebrel J, Dubois C, Chambaudet A.
[1] Ling YM, Liu WN, Shi FG. Adhesive joint design for high Photochemical degradation study of an epoxy material
yield & low cost assembly of fiber optic devices. In: Pro- by IR-ATR spectroscopy. Polym Degrad Stabil 1988;62:
ceedings of IEEE ECTC 2002. p. 662–6. 353–9.
[2] Nagata H, Shiroishi M, Miyama Y, Mitsugi N, Miyamoto [10] Hepburn DM, Kemp IJ, Cooper JM. Degradation of filled
N. Evaluation of new UV-curable adhesive material for epoxy resin surfaces. Polym Degrad Stabil 2000;70:245–
stable bonding between optical fibers and waveguide de- 51.

You might also like