Optics & Laser Technology 91 (2017) 85–91

Contents lists available at ScienceDirect

Optics & Laser Technology

journal homepage: www.elsevier.com/locate/optlastec

Full length article

Process design based on temperature field control for reducing the thermal MARK
residual stress in glass/glass laser bonding
⁎
Yanyi Xiaoa,b, Wen Wangb, Xingyang Wua, Jianhua Zhanga,b,
a
School of Mechatronic Engineering and Automation, Shanghai University, Shanghai 200072, China
b
Key Laboratory of Advanced Display and System Applications of Ministry of Education, Shanghai University, Shanghai 200072, China

A R T I C L E I N F O A BS T RAC T

Keywords: This paper presents a process design method based on temperature field control for reducing the thermal
Thermal residual stress residual stress in glass/glass laser bonding. The effect of the starting point, the shutting down rate of the laser
Laser bonding and the substrate temperature on the temperature field and sequential residual stress of the substrates was
Viscoelasticity investigated by ANSYS and experimental validation. The results show that the residual stress could be reduced
ANSYS
by starting the bonding avoiding the corner of the path, turning the laser off gradually and heating the glass
Temperature field
substrates, which influence the temperature field at the scale of point, line and surface, respectively. Thereby,
the formation probability of cracks is inhibited, and the bonding quality is improved.

1. Introduction residual-stress was discussed. The viscoelasticity characteristic [11,12]

of glass was taken into account during the analysis using a Maxwell
Laser encapsulation based on local and non-contact laser heating Model. The analysis reckoned on the residual stresses field by using
technology is expected to be widely applied to high-power LED [1], Finite Element Analysis (FEA) to provide a preliminary estimation of
MEMS (Micro-electromechanical Systems) [2], solar cells [3] and other the thermal residual stress during laser bonding processing. Finally,
optical and semiconductor products [4]. The glass/glass laser bonding the optimized bonding processing was validated.
creates a dynamic molten pool of glass frit between upper and lower
glass substrate [5]. Glass/glass laser bonding technique could over- 2. Experimental studies
come the drawbacks of traditional encapsulation because of the
inherent merits such as high thermal stability, low water vapor An 808 nm semiconductor laser was used to carry out the laser
transmission rate (WVTR) and small heat affected zone (HAZ) [6]. bonding experiments at the laser power of 6.5 W, the welding speed of
However, the small HAZ means large temperature gradient causing 2 mm/s and the laser beam diameter of 0.8 mm [13]. The TFT glass of
more crucial thermal residual-stress and cracks compared with tradi- Corning EAGLE XG® was selected as the glass substrates. The size of
tional processes. Therefore, a well-controlled temperature distribution the glass substrates was 30 mm×40 mm×0.7 mm and the thickness of
is of paramount importance to reduce the thermal residual stress and the glass frit was 10 µm. The glass frit and the operation platform were
enhance the reliability of the laser bonding. developed by this laboratory. The glass/glass laser encapsulation
Chen [7] and Paczkowska [8] have reported that the laser bonding schematic diagram was showed in Fig. 1.
quality is strongly dependent on the parameters of the laser processing. The normal laser bonding method starts the processing path at the
These works focus on the effect of the bonding parameters on the corner of the rectangle and immediately shut down the laser after
reduction of the residual stress. In the cases of glass [9] and stainless finished the bonding without heating the substrates. Three factors of
steel [10], the tempering process is an effective method to reduce the the processing improvement were carried out in this study. They were
residual stress. Hence, to get a precise and overall control of the heating the substrates through the whole processing from pre-bonding
temperature field is the key factor to reduce the thermal residual stress. to reaching a stable state of the substrate temperature distribution after
In this study, by considering the temperature field control at the bonding, starting bonding from the midpoint of the straight line and
scale of point, line and surface, the residual stress of glass/glass turning off the laser power gradually after the bonding. Based on the
bonding was investigated by ANSYS based on APDL (ANSYS para- improvement, a three-factor two-level full factorial experimental
metric design language), and the generation mechanism of thermal method was used in the experimental design, in which the two levels

⁎
Corresponding author at: School of Mechatronic Engineering and Automation, Shanghai University, Shanghai 200072, China.
E-mail address: jhzhang@oa.shu.edu.cn (J. Zhang).

http://dx.doi.org/10.1016/j.optlastec.2016.12.016
Received 25 November 2015; Accepted 17 December 2016
0030-3992/ © 2016 Elsevier Ltd. All rights reserved.
Y. Xiao et al. Optics & Laser Technology 91 (2017) 85–91

Fig. 3. The maximum temperature–node number.

and convection [14,15].

According to the measurement of the maker, ZCN® from Zhejiang,

China, the laser intensity of the semiconductor laser (ZFL-D45-400
Fig. 1. Glass/glass laser encapsulation schematic diagram.
45W) used in this work follows closely to Gaussian distribution. Thus, a
three-dimensional model was generated in this study. As can be seen
meant that adopt or do not adopt the improvement. The eight groups of from Fig. 2, fine, medium and coarse meshes were used in the bonding
experiments were conducted by the operation platform equipped with a zone, the HAZ and the glass substrate, respectively, due to the different
closed-loop controlled heating mica, two servomotors and a PID changes in the temperature in these three zones.
controlled semiconductor laser system. An optical microscope was The quality of the mesh was confirmed by increasing 20% elements
used to observe the morphologies of the glass/glass bonding area. to the whole model until the calculated temperature are almost
consistent ( < 1%) [16]. The mesh convergence curve based on max-
3. Numerical simulation of laser bonding imum temperature vs node number is shown as Fig. 3.

3.1. Finite element model

The finite element model was based on the assumptions for

3.2. Material properties
simplifying the laser-bonding model as follows:

For the numerical simulation, the material properties of the glass

(1) All materials are isotropic thermal behavior.
frit and glass substrate were set as viscoelasticity whose relaxed
(2) Thermal radiation is ignorable because the thermal radiation
modulus was in Prony series form [12]. The basic form could be given
influence is much smaller compared with the thermal conductivity
by the representation of Prony series [17]:

Fig. 2. Mesh of the finite element analysis.

86
Y. Xiao et al. Optics & Laser Technology 91 (2017) 85–91

Fig. 4. The material properties of the glass substrate and frit.

Table 1
The physical properties of the glass frit and the glass substrate used in the FEA model. other physical properties of the glass substrate and the frit remain
unchanged to the temperature, shown as Table 1.
Material Density, ρ (kg/m3) Poisson's ratio
3.3. Thermal stress analysis
Glass substrate 2540 0.23
Glass frit 2768 0.186
The temperature distribution was calculated by APDL. The load of
the surface heat was achieved through the nested loops to establish a
nG ⎛ t ⎞ four-dimensional matrix (x-coordinate value, y-coordinate value, time
G (t ) = G∞ + ∑ Gi exp ⎜− ⎟
and laser intensity). The whole processing time was divided into 550
i =1 ⎝ τiG ⎠ (1)
segments and the laser flux of each segment was calculated for the
⎛ whole model. The total laser flux of all the segments was loaded on the
t ⎞
nK
K (t ) = K∞ + ∑ Ki exp ⎜− ⎟ top surface of the glass frit. The calculated temperature distribution
i =1 ⎝ iK ⎠
τ (2) was set as a load to reload on the model and the element type was
transformed to SOLID 185 unit to obtain the final thermal stress. The
where G∞ and Gi were shear modulus, K∞ and Ki were bulk modulus,
thermal stress was calculated under thermal to structure coupling. The
τiG andτik were the Relative time of Prony series components,
choice of the element type was SOLID 70 unit [19]. In the stress
respectively.
analysis, eight cases of the laser processing based on the three-factor
In the heating source model, laser intensity distribution was
two-level full factorial experimental method was simulated.
Gaussian distribution [18]:

2Pα ⎛ 2r 2 ⎞ 4. Results and discussion

I (r ) = exp ⎜ − 2 ⎟
πR2 ⎝ R ⎠ (3)
4.1. Thermal stress generation mechanism
where α was the laser absorption efficiency of the glass frit, P was the
laser output power, R was the laser beam radius, r was the transverse The contour of von Mises stress obtained from the simulation
coordinate, and r = 0 was the beam center. starting the bonding at the corner was shown in Fig. 5. Glass is a brittle
The material properties dependent on temperature. The thermal material at room temperature. During the laser bonding process
conductivity, specific heat, Young Modulus, and CTE of the glass and discussed in this work, the temperature increases to around the soften
frit changed significantly, and the curve of each property - temperature point of the glass and the glass shows the viscoelasticity. When the
are shown in Fig. 4. The date under each temperature can be found by temperature around the soften point of glass, the von Mises equivalent
interpolation. The thermal conductivity and specific heat of glass frit stress could be used [20]. The simulated temperature field was similar
were measured by FL4010, TA Instruments, the Young Modulus and to that of the stress and not shown here. The thermal stress
CTE of glass frit measured DMA Q800 from TA Instruments. The concentrated at the ending point of laser flux while ended the bonding
material properties of glass substrate were measured by supplier. The at the time of 49 s. After the bonding, the highest thermal stress and

87
Y. Xiao et al. Optics & Laser Technology 91 (2017) 85–91

Fig. 5. The thermal stress contour of the upper substrate.

Table 2
Thermal stress of the bottom surface of upper glass substrate via different processing factors of improvement (×106 Pa).

Time (s) C/NH/D C/NH/G M/NH/D C/H/D M/NH/G M/H/D C/H/G M/H/G

49 156 109 149 131 180 128 78.4 164

49.41 137 88.1 131 116 91.6 113 61.5 60.1
49.71 123 87.1 118 105 78.6 102 56.7 42.3
49.97 110 84.9 106 94.8 69.1 92.1 50.6 32.4
52.14 37.2 50.9 26.0 30.8 46.5 23.1 32.3 20.0
56.4 37.2 33.1 26.0 30.8 18.9 23.1 22.3 10.7
63.06 18.0 17.8 12.5 20.3 18.9 17.0 15.4 10.8
79.72 8.22 8.65 6.06 13.7 5.27 12.1 10.6 5.64
146.39 2.95 2.95 2.94 3.27 2.94 3.47 2.72 2.58
163.06 2.94 2.94 2.94 2.69 2.94 2.95 2.67 2.55
243.06 2.92 2.92 2.90 2.58 2.90 2.49 2.58 2.48

the stress of the whole sample gradually decreased and the thermal 4.2. Impact of the improvements
stress expanded to the whole sample along the glass frit. In the end, the
stress only existed at every corner of the glass frit rectangle. The eight experiments with different laser processing factors were
Based on the temperature distribution analysis, the temperature simulated and the maximum thermal stresses at the bottom surface of
showed a steep decrease after turning off the laser, namely at 49 s. With the upper substrate at each key time were recorded in the Table 2. M
the time passing by, the decrease rate slowed down, and the thermal was the starting point in the middle of rectangle's side, C was the
stress exhibited a corresponding abatement. The heat conducted from starting point at the corner of the rectangle; H was using the heating
the laser ending point to the whole substrate along the rectangle of the substrate during pre-bonding to post-bonding, NH was no heating the
glass frit. At each corner of the rectangle, owing to the relatively large substrate; G was the gradually reducing the laser power, D was directly
superficial area, the temperature showed a steeper decline compared to shut down the laser.
the side of the rectangle. The decreasing behavior of the temperature The eight stress contours under the different factors were similar
after the bonding is closely related to the thermal residual stress. except the values. Based on the Table 2, the thermal stress of each
processing improvement was drawn in Fig. 6 and the partially enlarged
view of 49–50.4 s and 140–250 s were shown in Fig. 7 and Fig. 8,

88
Y. Xiao et al. Optics & Laser Technology 91 (2017) 85–91

In Fig. 6, all the lines exhibited the same trend. At 79.72 s, the
CHD, MHD, CHG lines had an increase compared to the CNHD
obtained from the normal process. However, in Figs. 7 and 8, the lines
with the H factor at least had a discount of 15% of the initial and final
stresses compared to the normal line. Namely, the method of heating
the substrate can reduce the initial and final stress, and have a
relatively stable and low decreasing rate of thermal stress. The MHG
line in Fig. 6, under all the three factors, had the smallest thermal
stress among the eight lines. The laser processing with three factors
would reach the lowest thermal residual stress according to the
simulation results.
The M method is a point-scaled control of laser bonding to optimize
the temperature field, and it has a slight reducing effect to the thermal
stress. The G method is a line-scaled control of laser bonding, which
has a better reduction effect compared to the M method. The H
method, which is a surface-scaled control, could get the best optimiza-
tion among the three methods. It is clear that the larger the controlling
Fig. 6. The thermal stress after bonding. scale, the more reduction of the thermal stress. By combining the three
temperature field controlling methods, the thermal stress could have
get well controlled.

4.3. Validation

According to Table 2, the experiments of eight groups were carried

out for six times. The samples were observed via optical microscope
after bonding for 24 h. The typical morphology of the glass frit was
shown in Fig. 8 in the order as the sequence of the row in Table 2.
In the micrograph of the samples, the white spot was the pores
generated during the laser bonding. Nevertheless, if the area of the
pores was too large compared to the glass frit area, the bonding quality
would decrease and the WVTR of the sample would increase. Fig. 9(a)
was the sample without any processing improvements. Plenty of cracks
were observed along the glass frit indicating the poor bonding quality.
By using the G method, the cracks decreased slightly as shown in
Fig. 9(b). Using the M method, most cracks were eliminated, only slices
of cracks were left showed as in Fig. 9(c). Using the H method, the
Fig. 7. The thermal stress during 49–50.4 s. cracks were almost removed as in Fig. 9(d).
The samples improved under double processing factors almost
removed cracks. The cracks only can be seen for the sample without H
method as shown in Fig. 9(e), and the cracks in this sample were quite
small. The sample improved under the three processing factors
achieved the best bonding quality, and few pores and no crack was
observed as in Fig. 9(h). It is clear that the thermal stress could be
reduced via the three factors. With a precise and overall temperature
control, the thermal residual stress could be deduced to the minimum.

5. Conclusions

The thermal residual stress of the substrates due to laser bonding

was investigated via ANSYS and experimental validation. The result
shows that the scale of the temperature field control is closely related to
the residual stress. Based on the simulation and the experimental
results, the following conclusions can be drawn:

1) The point-scaled control can homogenize the maximum residual

stress at the four corners, and can only reduce the systematic
Fig. 8. The thermal stress during 140–250 s. thermal residual stress deviation.
2) The line-scaled control can reduce the initial thermal stress after
bonding and inhibit the formation of the cracks, which is more vital
respectively. Based on the MNHD line, the stress had a slight decrease than the point-scaled control.
than the normal line, and it meant that avoiding the starting point to 3) The surface-scaled control can reduce the initial and finial thermal
the corner could decrease the stress during the whole process. As the stress, which can directly improve the laser bonding quality.
CNHG and the CHG lines shown in Fig. 7, the stress had almost 30% 4) Combining the point, line, surface-scaled controls, the thermal
and 50% discount compared to the normal line, meaning that using the residual stress can be reduced and the bonding quality can be
method of gradually reduced the laser power can deduct the initial improved markedly.
stress after bonding.

89
Y. Xiao et al. Optics & Laser Technology 91 (2017) 85–91

Fig. 9. The micrograph of (a) CNHD (b) CNHG (c) MNHD (d) CHD (e) MNHG (f) MHD (g) CHG (h) MHG samples.

Acknowledgements LED processing by using femtosecond laser, Opt. Lasers Eng. 74 (2015) 17–21.
[2] F. Bardin, S. Kloss, C. Wang, A.J. Moore, A. Jourdain, I.D. Wolf, et al., Laser
bonding of glass to silicon using polymer for microsystems packaging, J.
This work is supported by the Shanghai Subject Chief Scientist (No. Microelectromach. Syst. 16 (3) (2007) 571–580.
14XD1401800), Shanghai Engineering Research Center of Flat Panel [3] F. Ribeiro, J. Macaira, R. Cruz, J. Gabriel, L. Andrade, A. Mendes, Laser assisted
glass frit sealing of dye-sensitized solar cells, J. Sol. Energy Mater. Sol. Cells 96
Display, the Science and Technology Commission of Shanghai (2012) 43–49.
Municipality (No. 14DZ2280900). [4] Wang Tao, Yinan Ma, Yanbin Chen, et al., The influence of adhesive viscosity and
elastic modulus on laser spot weld bonding process, Int. J. Adhes. Adhes. 51 (2014)
111–116.
References [5] H. Kind, E. Gehlen, M. Aden, A. Olowinsky, A. Gillner, Laser glass frit sealing for
encapsulation of vacuum insulation glasses, Phys. Procedia 56 (2014) 673–680.
[1] I. Reklaitis, T. Grinys, R. Tomašiūnas, et al., A new geometrical approach for rapid [6] Duan Yu, Yong-Qiang Yang, Zheng Chen, et al., Recent progress on thin-film

90
Y. Xiao et al. Optics & Laser Technology 91 (2017) 85–91

encapsulation technologies for organic electronic devices, Opt. Commun. (2016) (in [14] K. Zeng, D. Pal, H.J. Gong, et al., Comparison of 3DSIM thermal modelling of
press) (Available online 25 August 2015). selective laser melting using new dynamic meshing method to ANSYS, Mater. Sci.
[7] Shuhai Chen, Jihua Huang, Jun Xia, et al., Influence of processing parameters on Technol. 31 (8) (2015) 945–956.
the characteristics of stainless steel/copper laser welding, J. Mater. Process. [15] H. Wei, Y. Zhang, L. Tan, et al., Energy efficiency evaluation of hot-wire laser
Technol. 222 (2015) 43–51. welding based on process characteristic and power consumption, J. Clean. Prod. 87
[8] Marta Paczkowska, The evaluation of the influence of laser treatment parameters (2015) 255–262.
on the type of thermal effects in the surface layer microstructure of gray irons, Opt. [16] Y. Xiao, W. Wang, J. Zhang, Accurate predetermination of the process parameters
Laser Technol. 76 (2016) 143–148. for glass/glass laser bonding based on the temperature distribution analysis, J.
[9] Fabien Béchet, Norbert Siedow, Dominique Lochegnies, Two-dimensional finite Electron. Packag. 138 (2) (2016) 021006.
element modeling of glass forming and tempering processes, including radiative [17] Ali Sarhadi, Jesper Henri Hattel, Hans N.ørgaard Hansen, Evaluation of the
effects, Finite Elem. Anal. Des. 94 (2015) 16–23. viscoelastic behaviour and glass/mould interface friction coefficient in the wafer
[10] Gopa Chakraborty, C.R. Das, S.K. Albert, A.K. Bhaduri, et al., Study on tempering based precision glass moulding, J. Mater. Process. Technol. 214 (7) (2014)
behaviour of AISI 410 stainless steel, Mater. Charact. 100 (2015) 81–87. 1427–1435.
[11] S.P. Donegan, A.D. Rollett, Simulation of residual stress and elastic energy density [18] S.Z. Shuja, B.S. Yilbas, Laser heating of a moving slab: influence pulse intensity
in thermal barrier coatings using fast Fourier transforms, Acta Mater. 96 (2015) parameter on temperature and stress fields, Opt. Laser Technol. 70 (2015) 7–16.
212–228. [19] Bappa Acherjee, Arunanshu S. Kuar, Souren Mitra, Effect of carbon black on
[12] Laurent Daudeville, H.élène Carré, Thermal tempering simulation of glass plates: temperature field and weld profile during laser transmission welding of polymers: a
inner and edge residual stresses, J.: J. Therm. Stress. 21 (6) (1998) 667–689. FEM study, Opt. Laser Technol. 44 (3) (2012) 514–521.
[13] Wen Wang, Yanyi Xiao, Xingyang Wu, et al., Optimization of laser-assisted glass [20] M. Utz, P.G. Debenedetti, F.H. Stillinger, Atomistic simulation of aging and
frit bonding process by response surface methodology, Opt. Laser Technol. 77 rejuvenation in glasses, Phys. Rev. Lett. 84 (7) (2000) 1471.
(2016) 111–115.

91