- Email: [email protected]
Flexible printed circuit boards laser bonding using a laser beam homogenization process
Optics and Lasers in Engineering 50 (2012) 1643–1653
Contents lists available at SciVerse ScienceDirect
Optics and Lasers in Engineering journal homepage: www.elsevier.com/locate/optlaseng
Flexible printed circuit boards laser bonding using a laser beam homogenization process Joohan Kim a, Haewoon Choi b,n a b
Dept. of Mechanical and Automotive Engineering, Seoul National University of Science and Technology, Nowon-Gu Gongneung-2Dong 172, 139-743, Seoul, South Korea Dept. of Mechanical and Automotive Engineering, Keimyung University, 704-701, Daegu, South Korea
a r t i c l e i n f o
abstract
Article history: Received 23 May 2011 Received in revised form 19 May 2012 Accepted 21 May 2012 Available online 27 June 2012
A laser micro-bonding process using laser beam shaping is successfully demonstrated for flexible printed circuit boards. A CW Ytterbium fiber laser with a wavelength of 1070 nm and a laser power density of 1–7 W/mm2 is employed as a local heat source for bonding flexible printed circuit boards to rigid printed circuit boards. To improve the bonding quality, a micro-lens array is used to modify the Gaussian laser beam for the bonding process. An electromagnetic modeling and heat transfer simulation is conducted to verify the effect of the micro-lens array on the laser bonding process. The optimal bonding parameters are found experimentally. As the measured temperature ramp rate of the boards exceeds 1100 K/s, bonding occurs within 100–200 ms at a laser power density of 5 W/mm2. The bonding quality of the FPCB is verified with a shear strength test. Process characteristics are also discussed. & 2012 Elsevier Ltd. All rights reserved.
Keywords: Flexible printed circuit boards FPCB Laser welding Micro-bonding
1. Introduction As multifunctional portable electrical appliances are developed and commercialized, advanced micro-packaging technology for highdensity electronic devices becomes necessary, and related research has been carried out [1,2]. In particular, flexible printed circuit boards (FPCB) have been widely adopted in portable electrical devices such as cell phones and laptops, in which folding or sliding mechanisms are required [3–6]. This FPCB is usually connected to other electrical components, such as rigid printed circuit boards (PCB) or transparent substrates of other display units, and the bonding or reflowing process can be accomplished with various heating sources, including hot bars, soldering irons, ultrasonic sources, and convection ovens [7–11]. A hot bar or soldering iron offers a broad heating source, but imperfect or irregular surface contact may lead to bonding or reflowing failure, and tool wear is also a concern. Ultrasonic bonding reduces the local negative thermal effect on electrical boards, but precise contact for vibration transfer can be a problem. A convection oven can be used to heat the entire product for reflowing in non-contact mode, and high throughput can be expected. However, because heat is applied to all parts of the product, other temperature-sensitive components can also be affected. In sum, although each of these processes offers distinct advantages, each is limited to high-precision bonding of
n
Corresponding author. Fax: þ82 53 580 6725. E-mail address: [email protected] (H. Choi).
0143-8166/$ - see front matter & 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.optlaseng.2012.05.005
ultrafine pitch electrical boards in high-density micro-system packaging. Moreover, it is generally difficult to control the bonding or reflowing temperature of the components to achieve consistent bonding results, as the boundary conditions for the parts being heated (including the heat sources) cannot be defined precisely. Heat loss to the surroundings by radiation or convection is not easily expressed, and irregular contact between the heat source and the FPCB can lead to inconsistent heat resistance. Consequently, imperfect bonding can occur in any of these processes, resulting in a reliability issue for the overall packaging procedure. A focused laser beam offers unique advantages as a heat source, including high spatial resolution, absence of contact, precision control, and rapid energy delivery. Accordingly, micro-system packaging using lasers has been developed as an advanced manufacturing process [12,13]. Laser irradiation transfers heat to a material and thus provides the mechanism for bonding to a substrate. A laser beam can heat metal inter-connectors, and the heat can be transferred by conduction to a solder paste for bonding. In addition, a laser beam with an appropriate wavelength can penetrate a polymer layer of a PCB, and direct heating can then be applied to a solder paste beneath it. In this case, the upper layer is bonded to the substrate via the melted solder paste. This type of laser soldering is also known as a laser reflow soldering process [14–16]. Laser reflow soldering provides precise control of energy to the reflow solder paste, and hence a precise soldering temperature of the solder paste can be achieved. In this type of laser process, reliability and repeatability can be improved in comparison with conventional procedures. Moreover, the processing time can be
1644
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
considerably reduced because of the rapid non-contact heat transfer of radiation. The spot size of a focused laser beam can easily be held to less than 100 mm, and thus local heating can be achieved with minimal heating of the surroundings. Because the laser beam path can be also controlled with micro-motion stages or a galvano scanner, not only a 1-D heat source, but also a 2-D or 3-D heat source path can be applied. This enables the generation of an optimized heat-source pattern for controlling the temperature of the solder paste with respect to various boundary conditions and material properties. In this paper, we propose a laser reflow process for bonding an FPCB to a PCB. An FPCB film consists of polymer and metal layers. Solder paste is inserted between the FPCB and the PCB. Therefore, a careful process design must be adopted to achieve successful bonding. A focused laser beam with a Gaussian energy distribution has a higher energy density in the center and can produce an unintended local temperature increase in the material, leading to FPCB damage. Moreover, large spatial and temporal temperature differences can induce internal stress in the boundary, causing micro-cracks. We use a CW IR laser system to reflow the solder paste between the FPCB and the PCB. Laser process parameters to achieve successful bonding are investigated. The laser beam is homogenized with a micro-lens array, and the effect is examined via a finite difference analysis of electromagnetic theory. The mechanical bonding strength is tested to evaluate the bonding quality, and related issues are also discussed.
2. Experimental setup and procedures 2.1. Experimental materials A commercial FPCB, of the type used in cell phones, was prepared for laser reflowing, and its configuration is shown in Fig. 1. The FPCB consists of a polymer stiffener, polymer insulators, connectors, and conducting lines. Electrical signals are transferred through the connectors of the PCB and FPCB, which are bonded with lead-free solder. This connecting layer is usually thinner than the other layers, making the component flexible and enabling physical movement of the FPCB. It consists of a polymer stiffener (polyimide: PI) and Cu. Micro-holes are fabricated to connect the top and bottom of the Cu layer when the solder paste
Fig. 1. Photographs of the FPCB and PCB.
is melted. The bonding substrate is a PCB, and solder paste is placed at each connecting point (Cu layer). The solder paste is Sn–3.0Ag–0.5Cu (wt%), and the melting temperature is 217–220 1C (or 490–493 K). The alignment of the FPCB and PCB connectors is achieved with a bonding jig, which is used to install other optical delivery devices such as lenses. The material properties of the solder paste and the FPCB composites are given in Table 1. Since bonding is accomplished using a laser beam source, material interaction with the laser (such as laser absorption) is important to the design and evaluation of the reflowing process. As mentioned above, the main components of the FPCB are a PI layer and a Cu layer with an Au coating. When the laser beam irradiates the FPCB layer, the Cu layer and Au coating directly interact with the laser beam. As the laser beam spot moves across the FPCB, the polymer layer is also exposed to the beam. The absorption characteristics of these two materials (PI and Cu) as well as Au with respect to wavelength were measured. The optical properties can be dependent on the surface topography. So, we measured the optical properties of the material used in the experiment by using spectroscopy. The measured data are shown in Fig. 2. Generally speaking, a high absorption rate is obtained for Cu in the UV wavelength range. However, the absorption decreases rapidly for relatively long wavelengths (in the vicinity of 1 mm). Moreover, the Cu layer is coated with a thin gold film with prior Nickel diffusion barrier, and thus the absorption rate is assumed to be slightly lower than that of a bulk Cu layer. Because the absorption rate is also dependent on the surface roughness, the actual rate will be somewhat different. For polyimide (the base material for the FPCB stiffener), the absorption also decreases rapidly as the wavelength increases. At a wavelength of 1.07 mm (the wavelength used herein), most of the laser energy will be transmitted to the solder paste or the PCB layers. In terms of the optical properties, the metal layer is mainly heated by the laser beam, and the heat is diffused to the contacting layers (solder paste and polyimide). When an IR laser beam is irradiated, it is to
Fig. 2. The optical absorption spectra of PI, Au and Cu.
Table 1 Thermal properties of the solder paste and the FPCB [20]. Material
Thermal Conductivity (W/(m K))
Specific Heat (J/g K)
Melting Point (1C or K)
Polyimide Cu Solder paste (Sn–3.0Ag–0.5Cu)
0.174–0.317 401 57
1.00–1.42 0.39 0.22
None 1083 (or 1356 K) 217–220 (or 490–493 K)
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
be expected that more than 80% of the energy will be reflected. However, the minimum energy required to melt the solder paste is relatively small compared to the power produced by commercial material-processing lasers (10 100 W). In our work, we use a 50 W ytterbium CW IR fiber laser (wavelength¼1070 nm) as a heat source.
2.2. Laser processes The experimental setup for the laser reflowing process is shown in Fig. 3. The laser system was combined with a scanner system (Scanlab) and computer-controlled micro-motion stages for the experiment. A bonding sample is aligned with the motion stages, and the laser beam is scanned by a galvanometer scanner with a computer-programmed path. The maximum power of the laser is 50 W, but the material could be damaged by a focused laser beam spot with a diameter of around 20 mm. As the power density needed to increase the temperature of the solder paste is less than 5 W/mm2 at a scanning speed of approximately 100 mm/s, the laser beam is defocused to reduce the power density on the target, and the spot diameter is held at around 3 mm. The laser parameters used in the experiment are given in Table 2. Because the connector assembly is longitudinal in shape, a linear or elliptical laser spot is preferable for scanning the reflowing area, instead of the circular shape obtained from a Gaussian beam. The elliptical or linear spot cannot cover the whole connector assembly in a single path, and hence a multiple scanning scheme is introduced. As previously noted, the irradiated laser beam is absorbed in the Cu and PI layers of the FPCB, raising the temperature of the layers. The heat is diffused into the solder paste, and the temperature of the solder paste rises to the melting point. After laser irradiation, the components are cooled, and bonding is completed. Prior to the
1645
bonding process, a flux is applied to prevent oxidation and improve the wetting of the solder by eliminating any contaminants. Actual bonding occurs between the FPCB and PCB when the solder material melts. However, pressure must be applied to the assembly to obtain a reliable and homogeneous bonding quality. The cylindrical lens is used not only to reshape the laser beam spot, but also to exert pressure on the FPCB and solder layer. The laser beam from the scanner is transmitted through the focusing lens, and an elliptical-shaped beam is created. This cylindrical lens is mounted on the jig holding the samples. A clamping pressure is also applied on the bonding materials on the jig (see Fig. 4). To improve the homogeneity of the heat source on the target, a microlens diffuser is adopted. The exact coordinates of the bonding targets are aligned with the motion stages before the scanner
Fig. 4. Schematics of the sample jig.
Fig. 3. Schematic of the experimental setup.
Table 2 Experimental laser parameters. Laser type
Wavelength (nm)
Beam mode
Power density (W/mm2)
Scanning speed of the laser beam (mm/s)
Target materials
Ytterbium fiber laser (CW)
1070
TEM00, CW
1–7
100–500
FPCB, PCB, and solder ball
1646
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
is used. The controllable process parameters are the scanning speed, the laser power density, and the number of laser scans. The laser beam is chopped with a signal from a function generator to control the energy and prevent any damage by
overheating. The configuration of each layer and specifications including thickness are shown in Fig. 5.
2.3. Beam shaping The laser beam used in the experiment has a Gaussian energy profile, which is generally expressed as follows: Iðr,zÞ ¼ ð1RÞI0
Fig. 5. Configuration of the layers.
Fig. 6. Effects of beam profiles on the laser bonding process.
w0 wz
2
2r 2 , exp w2z
ð1Þ
where R is the reflectivity of the target material, I0 is the intensity at the center of the beam, w0 is the beam waist, wz is the laser beam width at which the intensity is 1/e2 of the axial value, and r is the radial distance from the center. The maximum power density occurs at the center of the laser beam. In a laser process, this type of beam distribution can produce a large temperature difference in the absorbing layer, and the material exposed to the center of the beam can become overheated. The glass transition temperature of the PI attached to the Cu layer is roughly 360– 410 1C (or 633–683 K). Therefore, damage can be expected when the Cu layer is overheated. The heat conductivities of both Cu and the solder paste are high, and thus heat can readily be transferred through the contacting solder layer. Hence a relatively uniform temperature in these layers can be also assumed. However, PI has a low thermal conductivity, and thus is vulnerable to overheating. Consequently, the Gaussian profile of the laser beam (which is desirable for the fine cutting and drilling of materials) may not be appropriate in a bonding process, which requires uniform heat transfer to the sample. A relatively smooth laser beam profile with a low peak at the center seems to be more appropriate for a bonding process (see Fig. 6). When a homogenized heat source is formed on the surface, it produces a fairly uniform temperature distribution on the absorbing layer. To achieve this effect, we employ a diffuser with a micro-lens array (Thorlabs; ED1). The material of diffuser is injected molded ZEONORTM and its index of refraction is 1.53. The size of the diffuser is 1.5 mm thick and 25 mm diameter. The diffusion angle can be limited in 20 degree and the transmission was optimized for 380–1100 nm. As shown in Fig. 7 (source Thorlabs Inc.), the micro-lens array (MLA) is formed on the diffuser surface to convert a Gaussian laser beam input into a square-shaped profile. This reshaped laser spot is used in the reflowing process to improve the bonding quality. The smoothing beam is condensed by a cylindrical lens, so that an elliptical-shaped beam is finally applied to the FPCB surface.
Fig. 7. Photos of the MLA: (a) the magnified lens array and (b) the He–Ne laser beam profile with the lens array.
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
3. Results and discussion 3.1. Numerical analysis of beam homogenization The modified laser energy distribution from the MLA is obtained by numerical analysis. Finite difference time domain (FDTD) is a computational method that can be used to analyze electromagnetic field propagation in a medium, and laser beam propagation can be estimated by this technique [17–19]. The theory of FDTD is basically derived from Maxwell’s equations, given by Eqs. (2) and (3): !
r E ¼ m !
r H ¼e
! @H @t
1647
beam propagation was closer to collimated or parallel beam propagation due to the focusing effect of the lens. To analyze the beam propagation through the combined optical elements (focusing lens þMLA), an additional simulation was conducted, and the results are shown in Fig. 10. As can be seen, the laser homogeneity was much improved by the combined optical elements. Another type of simulation was also used to analyze the beam propagation through the combined optical elements (focusing lens þMLA). This simulation was based on plane waves and a combination of a cylindrical lens and an MLA. As Fig. 11(a) shows,
ð2Þ
! ! @E þs E , @t
ð3Þ
where E is the electric field, H is the magnetic field, e is the dielectric constant, and m is the magnetic permeability. From Maxwell’s equations, the differential form of FDTD can be expressed as Eqs. (4) and (5): F n ði þ 12 ,j,kÞF n ði12,j,kÞ @F n ði,j,kÞ ¼ þOðDx2 Þ @x Dx 1
ð4Þ
1
@F n ði,j,kÞ F n þ 2 ði,j,kÞF n2 ði,j,kÞ ¼ þ OðDt 2 Þ: @t Dx
ð5Þ
Eqs. (4) and (5) can be expanded along the y-axis and z-axis, and the final form of FDTD is then given by Eqs. (6) and (7) 1 nþ2
Hx
ði,j þ 12 ,kþ 12Þ 1 n2
¼ Hx
ði,j þ 12 ,kþ 12 Þ
Dt
m
1 n ðEx ði,j þ 1,k þ 12ÞÞ ðEnx ði,j,kþ 12ÞÞ Dy
1 n ðE ði,j þ 12 ,kþ 1ÞðEny ði,j þ 12,kÞÞ Dz y 1 nþ2
Ex
ðiþ 12 ,j,kþ 12Þ
ðiþ 12 ,j,kÞ ¼ En1 x
1
Dz
1 n2
ðHy
Dt
e
Fig. 8. (a) simulation model and (b) laser beam propagating through air.
ð6Þ
1 1 1 n n ðHx 2 ði þ 12 ,j þ 12,kÞÞ ðHx 2 ði þ 12 ,j12,kÞÞ Dy 1 n2
ði þ 12 ,j,k þ 12 ÞHy
ði þ 12 ,j,k12ÞÞ:
ð7Þ
To analyze the homogeneity of the MLA and inverse Kepler optical device, this FDTD was input to the commercial simulation software EM Explorer (EM Explore Inc, 2007) for comparison of the 3-D lens assembly. Assuming that the software has already been proven for such an application, only a simple beam propagation simulation was used as the validation procedure in this study. The boundary conditions used in the simulation were perfectly matched boundary layer (PML) along the sidewalls of beam propagation and full absorption at the free end of beam propagation. As shown in Fig. 8(a), a laser beam with a wavelength of 1070 nm was irradiated through an excitation port or slit to the air domain (free space). It propagated in spherical form, with weaker intensity along the propagation direction, as shown in Fig. 8(b). As expected, the laser beam propagates through open air domain. Under the same conditions depicted in Fig. 8, a focusing lens with an index of refraction of 1.4 was used to converge the laser beam to the air domain, as shown in Fig. 9. As expected, the divergence of the laser decreased, and the laser beam was focused along the center line in the direction of propagation. Compared to the free-space propagation experiment shown in Fig. 8, the laser
Fig. 9. (a) simulation model with focusing lens and (b) laser beam propagating through air.
Fig. 10. (a) simulation model with focused lens and MLA and (b) laser beam propagating through air.
1648
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
Fig. 11. Plane laser beam propagation through air with (a) cylindrical lens only and (b) (c) cylindrical lens and MLA.
Fig. 12. Relative intensity distribution (A.U.) at the give location (mm) for (a) cylindrical lens only.
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
the laser beam propagates through the cylindrical lens and converges on a focusing point. The intensity of the laser beam is very high in the center and relatively low at the edges. To improve the homogeneity of the propagating beam, an MLA was placed in the front of a cylindrical focusing lens, as shown in Fig. 11(b) and (c). The intensity of the laser beam after passing through the MLA was more randomized in comparison with the incoming beam. To quantify the homogeneity of the laser beam, the intensity was measured along the transverse direction and compared with that obtained from an optical element without an MLA. The simulation times related to beam propagation and measurement location were the same in both cases, and intensity data were extracted as shown in Fig. 12. To quauntify the effectiveness of MLA, the intensity of incoming beam was measured at the full width half max location and all measured intensity was relatively scaled to this value. The plane wave and cylindrical lens assembly produced an intensity 2.1 times that of the incoming laser beam, whereas the cylindrical lens with the MLA produced an intensity only 1.6 times that of the incoming beam. The measured width of homogeneity at FWHM was 20 for a cylindrical lens only and 40 for a cylindrical lens and MLA combination. A detailed comparsion is presented in Table 3. 3.2. Numerical analysis of temperature distribution Temperature distribution is an important factor in the design of the bonding process, as estimation of bonding temperature and energy is one of the keys to successful reflowing. However, when a laser is used as a heating source, optical properties of the materials (including reflection and absorption) must be considered, unlike regular contact heat sources such as hot bars and hot plates. The energy transfer rate via radiation is related to the material surface properties. In particular, it is a function of temperature. Moreover, the energy distribution in the laser spot is also non-homogeneous and usually has a Gaussian distribution in a single-mode laser beam. Thus, in order to accurately estimate the temperature distribution, the factors mentioned above must be considered carefully. In addition, the FPCB, solder paste, and PCB have complicated geometries, including multilayer materials, and thus the temperature distribution for the process should be obtained by numerical analysis. The irradiation time for laser reflowing is more than a few ms, which is much larger than the relaxation time scale of the materials, and thus local thermodynamic equilibrium can be assumed. In this case, the temperature distribution in the materials can be estimated using the Fourier diffusion equation, which is given by
rcp
@T ¼ rUðkrTÞ þ Q ðx,z,tÞ, @t
ð8Þ
where T denotes temperature, t denotes time, r is the material density, Cp is the specific heat of the materials, k is the heat conductivity, and Q is the absorbed power density on the surface. In our experiment, a Gaussian beam was passed through a cylindrical lens and an MLA, and the final energy distribution was Table 3 Summary of EM simulation data (Relative intensity of laser beam). Description
Cylindrical lens only
Cylindrical lens þMLA
Maximum Half maximum Rate of above half maximum (%)
2.1 1.05 37%
1.6 0.8 69%
*The unit in this table is relative intensity of incoming beam (full width half max).
1649
investigated using the above equation. The exact energy distribution is elliptical in shape, but we can roughly model the crosssectional energy distribution as follows: I 2x2 Q ðx,z,tÞ ¼ ð1RÞ exp 2 expðazÞ, ð9Þ 2Ux0 Ul x0 where R is reflectivity, I is the laser intensity, x0 is the laser spot radius, x is the radial distance from the center of the laser spot, l is the length of the beam spot, a is a distinct constant, and z is the coordinate of the depth direction. The above equation does not represent the exact energy distribution of the modified laser beam because the unmodified beam has a Gaussian distribution, and the effect of the MLA is not considered. However, it is impractical to use an exact analytical expression to describe the beam, since the presence of both an MLA and a cylindrical lens makes the beam shape rather complicated. We used a discrete energy distribution in the numerical simulations instead of the exact analytical equation. Boundary conditions must be imposed to solve the partial differential equation, and an adiabatic condition is applied to the surface exposed to the air due to its high thermal resistance: @T ¼0 ð10Þ @x x ¼ 0 Moreover, axial symmetry is assumed in the radial direction, and is expressed as follows: @T ¼0 ð11Þ @z z ¼ 0 In addition, the FPCB substrate is relatively large compared to the solder paste, and thus a constant temperature condition at the boundary of the FPCB can be used: Tðx-1Þ ¼ 300 K:
ð12Þ
The initial temperature is taken to be room temperature and can be written: Tðt ¼ 0Þ ¼ 300 K:
ð13Þ
After laser irradiation, the following governing equation can be applied to determine the temperature distribution with similar boundary conditions:
rcp
@T ¼ rUðkrTÞ: @t
ð14Þ
An exact analytical solution for a real geometry is difficult, and an approximate temperature profile can be obtained via a numerical approach. The temperature distribution in the materials is important when the temperature of the solder paste reaches the reflowing point. A numerical simulation of the transient temperature profile was performed using the commercial software FLUENT. The laser beam spot diameter was set at 3 mm, and the laser energy input was applied to the top of the FPCB. The energy distribution of the top layer was approximated via the previous numerical results, and the optical and material properties were assumed to be constant with respect to temperature. The temperature range from room temperature to bonding temperature was less than 500 K. Under the above conditions (Fig. 13(a)), the resulting transient temperature distributions of the materials are shown in Fig. 13(b). The temperature of the copper layer increased to the melting point of the solder paste in 100 ms, with a steady laser power input of 5 W/mm2, and the heat diffused to the solder layer in about the same time. The FPCB layer was heated by conduction, mainly due to its low laser absorption rate. When an MLA is used, a more homogenized laser input and temperature distribution can be obtained. This was confirmed by the simulation results (Fig. 13(c)), and the bonding quality was improved by using an
1650
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
Fig. 13. Temperature distribution: (a) schematics for numerical modeling (b) Gaussian beam input (c) Gaussian beam input with MLA (d) photo of fractured sample.
MLA. Comparing temperature gradients at the interfaces of solder and the connector, we can indirectly expect a possible location of fracture induced thermal strain as shown in Fig. 13. As shown in Fig. 13(b) and (c), there is a rapid temperature change at the interface without the MLA combination setup (Fig. 13b) which may result in possible fracture.
4. Experimental results and discussion 4.1. Bonding results An experimentally measured temperature profile of the FPCB and solder paste provides information on energy transfer during the process. Thermocouples were attached on the surface of FPCB layer which locations (A,B, and C) are described in Fig. 14, and transient and spatial temperature profiles were obtained. Fig. 14 shows the temperature history with respect to time for a laser intensity of 2–5 W/mm2 and a scanning speed of 200 mm/s. The temperature of the solder paste increased to 220 1C (or 493 K, the melting temperature of the solder) in 200 ms. This result indicates that the ramp rate of the laser process exceeds 1100 K/s, which is faster than conventional processes using contact heating (such as hot bar and hot plate processes). Similar measurements were obtained with an MLA, except that the maximum temperature at the center of the FPCB connector was lower due to energy loss from the modified laser beam. Moreover, it is believed that the high heat-conduction rates of copper and solder paste led to a
Fig. 14. Temperature profile with respect to time during laser irradiation.
quick response to energy transfer via laser irradiation. The effects of the diffuser were confirmed by the spatial temperature profile shown in Fig. 15. The results are in good agreement with previous numerical results, and the diffuser induced a fairly homogeneous profile in the materials. The temperatures at the left and right ends differed because one end was exposed to the PI layer, in
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
which different boundary conditions must be applied. On the basis of these results, we may conclude that this type of beam shaping can improve the quality of the bonding between the materials. The cross-sectional views of the samples bonded via the laser process are shown in Fig. 16. When beam shaping with an MLA was carried out at an appropriate energy level, a uniform bonding layer was achieved between the FPCB and the substrate (Fig. 16(a)). Bonding failure caused by an excessive energy level is shown in Fig. 16(b). Microscopic views of the reflowing solder
Fig. 15. Temperature profiles with respect to the X-cooridnates of the FPCB during laser irradiation.
1651
paste are shown in Fig. 16(c) and (d). The melted solder paste flowed onto the copper connector and formed a cover layer with a thickness of 1 2 mm. It is thought that this kind of reflowing behavior improves the mechanical and electrical reliability of the connecting parts. 4.2. Mechanical bonding strength measurement A shear test was performed to confirm the mechanical bonding quality of the laser process using a micro-shear tester (Instron 5800). A bonding sample was taken with 1 mm wide and 5 mm long and was mounted on a micro-jig, and shear force was applied until a maximum load was reached and the sample fractured with a speed of 2 mm/min. We used the same rate of elongation of the
Fig. 17. MLA effect on shear strength.
Fig. 16. Photographs of the bonded FPBCs (a) a good bonding result of the copper interconnections of the substrate (b) a bad bonding result of the copper interconnections of the substrate (c) SEM of joined part (d) SEM photographs of the copper interconnections of the substrate.
1652
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
sample to synchronize the starting point and the maximum load, and the elongation was measured at this point. The bonding strength (P) is defined as follows: P¼
L , Aef f
ð15Þ
where L is the applied load, and Aeff is the effective bonding area, which is around 0.4 mm2. The bonding strengths of the samples with and without an MLA are shown in Fig. 17. When the same amount of laser energy was applied for both Gaussian shape and MLA combination, the sample that was bonded with a homogenized laser beam exhibited higher shear strength during elongation, indicating that a reshaped laser beam with a MLA can broaden the effective bonding area of an FPCB and thereby improve the bonding quality. It should be noted that the comparison between the bonding strengths with and without beam shaping was made under the assumption that the total bonding energy was the same in both cases, as energy loss can be expected when the laser beam is reshaped using an MLA, and the resulting energy could be less than what is required for the bonding process. In this case, the comparison would be meaningless. Thus, when an MLA was used, the amount of laser beam energy lost due to the MLA was compensated for the sake of the comparison. Appropriate energy input is important for successful bonding. Too much energy can cause the polymer layer of the FPCB to overheat, which may harden the polymer layer and lead to thermal damage of the FPCB, resulting in a defective bond at the interface. Moreover, too little energy may fail to raise the temperature of the solder paste to the melting point. As it is rather difficult to examine the bonding quality directly, shear strengths were measured with respect to various laser power densities, as shown in Fig. 18. When the power density of the laser beam was 10% higher or lower than the optimized bonding power, the bonding sample exhibited a low bonding strength over the range of elongation, which implies that the laser power density must fall within that window.
a precise amount of energy to a local bonding area so that thermal damage to adjacent materials can be minimized. The temperature ramp rate exceeds 1100 K/s, and thus bonding can be accomplished quickly. In order to improve the laser reflowing quality, we modified the Gaussian beam to a somewhat flattened shape with an MLA and a cylindrical lens, resulting in a more homogenized temperature profile on the FPCB and solder paste. By electromagnetic field modeling, the modified laser beam was homogenized 30% more than the original beam. A concentrated laser beam may cause damage to a polymer layer by overheating, and this kind of beam shaping can improve the bonding strength. A thermal diffusion model and simulation were presented to confirm the reflowing effect in the material, and it was determined that the solder paste reaches the melting point in roughly 100–200 ms with a laser power density of 5 W/cm2. The temperature profile on the FPCB was measured experimentally, and the measured results were shown to be in good agreement with the calculated results. The effects of beam shaping on laser reflowing were verified with mechanical shear tests. The maximum shear strength increased when beam shaping was used, and it was inferred that the mechanical reliability of bonding layers can be improved by beam shaping. The mechanical strength of the bonding layer diminished when the laser bonding energy increased or decreased beyond a certain level, clearly implying that the laser power density must fall within a definite window to produce a good bond. Based on these results, it can be concluded that the laser reflowing procedure is an effective process for FPCB bonding. The principal challenge in applying this process will be determining how to control the laser power density on the FCPB.
Acknowledgments This work was supported by the National Research Foundation of Korea (no. 2011-0004266) and the authors thank Mr. B. H. Park for technical supports. References
5. Conclusions A laser reflowing process for bonding FPCBs was presented. This procedure offers fast, precise, and reliable bonding characteristics compared with conventional contact methods. The process transfers
Fig. 18. Laser power density effect on shear strength.
[1] Morrissey A, Kelly G, Alderman J, Barrett J, Lyden C, O’Rourke L. Some issues for microsystem packaging in plastic and 3D. Microelectr J 1998;29:645–50. [2] Van Brussel H, Peirs J, Reynaerts D, Delchambre A. Reinhart, Roth A, Weck M, Zussman E. Assembly of Microsystems. CIRP Ann—Manuf Technol 2000;49:451–72. [3] Shen K, Chen X, Guo M, Cheng J. A microchip-based PCR device using flexible printed circuit technology. Sens Actuators, B 2005;105:251–8. [4] Berenyi R. Prototyping of a reliable 3D flexible IC cube package by laser micromachining. Microelectron Rel 2009;49:800–5. [5] Giovangrandi L, Gilchrist K, Whittington R, Kovacs G. Low-cost microelectrode array with integrated heater for extracellular recording of cardiomyocyte cultures using commercial flexible printed circuit technology. Sens Actuators, B 2006;113:545–54. [6] Svedberg M, Nikolajeff F, Thornell G. On the integration of flexible circuit boards with hot embossed thermoplastic structures for actuator purposes. Sens Actuators, A 2006;125:534–47. [7] Zhou J, Zhuang L, He Z, Herscovici D. Practical design, simulation, and prototype of a hot bar blade for printed circuit board soldering. J Manuf Syst 2001;20:177–87. [8] Darwish S, Al-Habdan S, Al-Tamimi A. A knowledge-base for electronics soldering. J Mater Process Technol 2000;97:1–9. [9] Kim K, Jung J, Zhou Y. Ultrasonic bonding of flexible PCB to rigid PCB using an Sn interlayer. Solder Surf Mt Technol 2009;21:4–10. [10] Lee J, Koo J, Hong S, Shin S, Moon Y, Jung J, et al. Longitudinal Ultrasonic Bonding of Electrodes between Rigid and Flexible Printed Circuit Boards. Jpn J Appl Phys 2008;47:4300–4. [11] Ille´s B, Harsa´nyi G. Investigating direction characteristics of the heat transfer coefficient in forced convection reflow oven. Exp Therm Fluid Sci 2009;33:642–50. [12] Illyefalvi-Vite´z Z. Laser processing for microelectronics packaging applications. Microelectron Rel 2001;41:563–70. [13] Chaminade C, Fogarassy E, Boisselier D. Diode laser soldering using a leadfree filler material for electronic packaging structures. Appl Surf Sci 2006;252:4406–10.
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
[14] Tian Y, Wang C, Ge X, Liu P, Liu D. Intermetallic compounds formation at interface between PBGA solder ball and Au/Ni/Cu/BT PCB substrate after laser reflow processes. Mater Sci Eng B 2002;95:254–62. ¨ ¨ [15] Stauffer L, Wursch A, Gachter B, Siercks K, Verettas I, Rossopoulos S, et al. A surface-mounted device assembly technique for small optics based on laser reflow soldering. Opt Laser Eng 2005;43:365–72. [16] Lee J, Lee Y, Kim Y. Fluxless laser reflow bumping of Sn–Pb eutectic solder. Scripta Mater 2000;42:789–93. [17] Jian Q, Morita N, Shi D, Nakamura O, Liu S. FDTD Simulation of Nonlinear Ultrasonic Pulse Propagation in ESWL. IEEE-EMBS. In: 27th Annual international conference of medicine and biology society; 2005. p. 1806–9.
1653
[18] Sakai A, Baba T. FDTD simulation of photonic devices and circuits based on circular and fan-shaped micro disks. J Light wave Technol 1999;17:1493–9. [19] Slavcheva M, Arnold M, Ziolkowski W. FDTD simulation of the nonlinear gain dynamics in active optical waveguides and semiconductor micro-activities. IEEE J Quantum Electron 2004;10:1052–62. [20] Tian D, Wang C, Tian Y. Effect of solidification on solder bump formation in solder jet process:Simulation and experiment. Trans Nonferrous Met Soc China 2008;18:1201–8.
Contents lists available at SciVerse ScienceDirect
Optics and Lasers in Engineering journal homepage: www.elsevier.com/locate/optlaseng
Flexible printed circuit boards laser bonding using a laser beam homogenization process Joohan Kim a, Haewoon Choi b,n a b
Dept. of Mechanical and Automotive Engineering, Seoul National University of Science and Technology, Nowon-Gu Gongneung-2Dong 172, 139-743, Seoul, South Korea Dept. of Mechanical and Automotive Engineering, Keimyung University, 704-701, Daegu, South Korea
a r t i c l e i n f o
abstract
Article history: Received 23 May 2011 Received in revised form 19 May 2012 Accepted 21 May 2012 Available online 27 June 2012
A laser micro-bonding process using laser beam shaping is successfully demonstrated for flexible printed circuit boards. A CW Ytterbium fiber laser with a wavelength of 1070 nm and a laser power density of 1–7 W/mm2 is employed as a local heat source for bonding flexible printed circuit boards to rigid printed circuit boards. To improve the bonding quality, a micro-lens array is used to modify the Gaussian laser beam for the bonding process. An electromagnetic modeling and heat transfer simulation is conducted to verify the effect of the micro-lens array on the laser bonding process. The optimal bonding parameters are found experimentally. As the measured temperature ramp rate of the boards exceeds 1100 K/s, bonding occurs within 100–200 ms at a laser power density of 5 W/mm2. The bonding quality of the FPCB is verified with a shear strength test. Process characteristics are also discussed. & 2012 Elsevier Ltd. All rights reserved.
Keywords: Flexible printed circuit boards FPCB Laser welding Micro-bonding
1. Introduction As multifunctional portable electrical appliances are developed and commercialized, advanced micro-packaging technology for highdensity electronic devices becomes necessary, and related research has been carried out [1,2]. In particular, flexible printed circuit boards (FPCB) have been widely adopted in portable electrical devices such as cell phones and laptops, in which folding or sliding mechanisms are required [3–6]. This FPCB is usually connected to other electrical components, such as rigid printed circuit boards (PCB) or transparent substrates of other display units, and the bonding or reflowing process can be accomplished with various heating sources, including hot bars, soldering irons, ultrasonic sources, and convection ovens [7–11]. A hot bar or soldering iron offers a broad heating source, but imperfect or irregular surface contact may lead to bonding or reflowing failure, and tool wear is also a concern. Ultrasonic bonding reduces the local negative thermal effect on electrical boards, but precise contact for vibration transfer can be a problem. A convection oven can be used to heat the entire product for reflowing in non-contact mode, and high throughput can be expected. However, because heat is applied to all parts of the product, other temperature-sensitive components can also be affected. In sum, although each of these processes offers distinct advantages, each is limited to high-precision bonding of
n
Corresponding author. Fax: þ82 53 580 6725. E-mail address: [email protected] (H. Choi).
0143-8166/$ - see front matter & 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.optlaseng.2012.05.005
ultrafine pitch electrical boards in high-density micro-system packaging. Moreover, it is generally difficult to control the bonding or reflowing temperature of the components to achieve consistent bonding results, as the boundary conditions for the parts being heated (including the heat sources) cannot be defined precisely. Heat loss to the surroundings by radiation or convection is not easily expressed, and irregular contact between the heat source and the FPCB can lead to inconsistent heat resistance. Consequently, imperfect bonding can occur in any of these processes, resulting in a reliability issue for the overall packaging procedure. A focused laser beam offers unique advantages as a heat source, including high spatial resolution, absence of contact, precision control, and rapid energy delivery. Accordingly, micro-system packaging using lasers has been developed as an advanced manufacturing process [12,13]. Laser irradiation transfers heat to a material and thus provides the mechanism for bonding to a substrate. A laser beam can heat metal inter-connectors, and the heat can be transferred by conduction to a solder paste for bonding. In addition, a laser beam with an appropriate wavelength can penetrate a polymer layer of a PCB, and direct heating can then be applied to a solder paste beneath it. In this case, the upper layer is bonded to the substrate via the melted solder paste. This type of laser soldering is also known as a laser reflow soldering process [14–16]. Laser reflow soldering provides precise control of energy to the reflow solder paste, and hence a precise soldering temperature of the solder paste can be achieved. In this type of laser process, reliability and repeatability can be improved in comparison with conventional procedures. Moreover, the processing time can be
1644
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
considerably reduced because of the rapid non-contact heat transfer of radiation. The spot size of a focused laser beam can easily be held to less than 100 mm, and thus local heating can be achieved with minimal heating of the surroundings. Because the laser beam path can be also controlled with micro-motion stages or a galvano scanner, not only a 1-D heat source, but also a 2-D or 3-D heat source path can be applied. This enables the generation of an optimized heat-source pattern for controlling the temperature of the solder paste with respect to various boundary conditions and material properties. In this paper, we propose a laser reflow process for bonding an FPCB to a PCB. An FPCB film consists of polymer and metal layers. Solder paste is inserted between the FPCB and the PCB. Therefore, a careful process design must be adopted to achieve successful bonding. A focused laser beam with a Gaussian energy distribution has a higher energy density in the center and can produce an unintended local temperature increase in the material, leading to FPCB damage. Moreover, large spatial and temporal temperature differences can induce internal stress in the boundary, causing micro-cracks. We use a CW IR laser system to reflow the solder paste between the FPCB and the PCB. Laser process parameters to achieve successful bonding are investigated. The laser beam is homogenized with a micro-lens array, and the effect is examined via a finite difference analysis of electromagnetic theory. The mechanical bonding strength is tested to evaluate the bonding quality, and related issues are also discussed.
2. Experimental setup and procedures 2.1. Experimental materials A commercial FPCB, of the type used in cell phones, was prepared for laser reflowing, and its configuration is shown in Fig. 1. The FPCB consists of a polymer stiffener, polymer insulators, connectors, and conducting lines. Electrical signals are transferred through the connectors of the PCB and FPCB, which are bonded with lead-free solder. This connecting layer is usually thinner than the other layers, making the component flexible and enabling physical movement of the FPCB. It consists of a polymer stiffener (polyimide: PI) and Cu. Micro-holes are fabricated to connect the top and bottom of the Cu layer when the solder paste
Fig. 1. Photographs of the FPCB and PCB.
is melted. The bonding substrate is a PCB, and solder paste is placed at each connecting point (Cu layer). The solder paste is Sn–3.0Ag–0.5Cu (wt%), and the melting temperature is 217–220 1C (or 490–493 K). The alignment of the FPCB and PCB connectors is achieved with a bonding jig, which is used to install other optical delivery devices such as lenses. The material properties of the solder paste and the FPCB composites are given in Table 1. Since bonding is accomplished using a laser beam source, material interaction with the laser (such as laser absorption) is important to the design and evaluation of the reflowing process. As mentioned above, the main components of the FPCB are a PI layer and a Cu layer with an Au coating. When the laser beam irradiates the FPCB layer, the Cu layer and Au coating directly interact with the laser beam. As the laser beam spot moves across the FPCB, the polymer layer is also exposed to the beam. The absorption characteristics of these two materials (PI and Cu) as well as Au with respect to wavelength were measured. The optical properties can be dependent on the surface topography. So, we measured the optical properties of the material used in the experiment by using spectroscopy. The measured data are shown in Fig. 2. Generally speaking, a high absorption rate is obtained for Cu in the UV wavelength range. However, the absorption decreases rapidly for relatively long wavelengths (in the vicinity of 1 mm). Moreover, the Cu layer is coated with a thin gold film with prior Nickel diffusion barrier, and thus the absorption rate is assumed to be slightly lower than that of a bulk Cu layer. Because the absorption rate is also dependent on the surface roughness, the actual rate will be somewhat different. For polyimide (the base material for the FPCB stiffener), the absorption also decreases rapidly as the wavelength increases. At a wavelength of 1.07 mm (the wavelength used herein), most of the laser energy will be transmitted to the solder paste or the PCB layers. In terms of the optical properties, the metal layer is mainly heated by the laser beam, and the heat is diffused to the contacting layers (solder paste and polyimide). When an IR laser beam is irradiated, it is to
Fig. 2. The optical absorption spectra of PI, Au and Cu.
Table 1 Thermal properties of the solder paste and the FPCB [20]. Material
Thermal Conductivity (W/(m K))
Specific Heat (J/g K)
Melting Point (1C or K)
Polyimide Cu Solder paste (Sn–3.0Ag–0.5Cu)
0.174–0.317 401 57
1.00–1.42 0.39 0.22
None 1083 (or 1356 K) 217–220 (or 490–493 K)
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
be expected that more than 80% of the energy will be reflected. However, the minimum energy required to melt the solder paste is relatively small compared to the power produced by commercial material-processing lasers (10 100 W). In our work, we use a 50 W ytterbium CW IR fiber laser (wavelength¼1070 nm) as a heat source.
2.2. Laser processes The experimental setup for the laser reflowing process is shown in Fig. 3. The laser system was combined with a scanner system (Scanlab) and computer-controlled micro-motion stages for the experiment. A bonding sample is aligned with the motion stages, and the laser beam is scanned by a galvanometer scanner with a computer-programmed path. The maximum power of the laser is 50 W, but the material could be damaged by a focused laser beam spot with a diameter of around 20 mm. As the power density needed to increase the temperature of the solder paste is less than 5 W/mm2 at a scanning speed of approximately 100 mm/s, the laser beam is defocused to reduce the power density on the target, and the spot diameter is held at around 3 mm. The laser parameters used in the experiment are given in Table 2. Because the connector assembly is longitudinal in shape, a linear or elliptical laser spot is preferable for scanning the reflowing area, instead of the circular shape obtained from a Gaussian beam. The elliptical or linear spot cannot cover the whole connector assembly in a single path, and hence a multiple scanning scheme is introduced. As previously noted, the irradiated laser beam is absorbed in the Cu and PI layers of the FPCB, raising the temperature of the layers. The heat is diffused into the solder paste, and the temperature of the solder paste rises to the melting point. After laser irradiation, the components are cooled, and bonding is completed. Prior to the
1645
bonding process, a flux is applied to prevent oxidation and improve the wetting of the solder by eliminating any contaminants. Actual bonding occurs between the FPCB and PCB when the solder material melts. However, pressure must be applied to the assembly to obtain a reliable and homogeneous bonding quality. The cylindrical lens is used not only to reshape the laser beam spot, but also to exert pressure on the FPCB and solder layer. The laser beam from the scanner is transmitted through the focusing lens, and an elliptical-shaped beam is created. This cylindrical lens is mounted on the jig holding the samples. A clamping pressure is also applied on the bonding materials on the jig (see Fig. 4). To improve the homogeneity of the heat source on the target, a microlens diffuser is adopted. The exact coordinates of the bonding targets are aligned with the motion stages before the scanner
Fig. 4. Schematics of the sample jig.
Fig. 3. Schematic of the experimental setup.
Table 2 Experimental laser parameters. Laser type
Wavelength (nm)
Beam mode
Power density (W/mm2)
Scanning speed of the laser beam (mm/s)
Target materials
Ytterbium fiber laser (CW)
1070
TEM00, CW
1–7
100–500
FPCB, PCB, and solder ball
1646
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
is used. The controllable process parameters are the scanning speed, the laser power density, and the number of laser scans. The laser beam is chopped with a signal from a function generator to control the energy and prevent any damage by
overheating. The configuration of each layer and specifications including thickness are shown in Fig. 5.
2.3. Beam shaping The laser beam used in the experiment has a Gaussian energy profile, which is generally expressed as follows: Iðr,zÞ ¼ ð1RÞI0
Fig. 5. Configuration of the layers.
Fig. 6. Effects of beam profiles on the laser bonding process.
w0 wz
2
2r 2 , exp w2z
ð1Þ
where R is the reflectivity of the target material, I0 is the intensity at the center of the beam, w0 is the beam waist, wz is the laser beam width at which the intensity is 1/e2 of the axial value, and r is the radial distance from the center. The maximum power density occurs at the center of the laser beam. In a laser process, this type of beam distribution can produce a large temperature difference in the absorbing layer, and the material exposed to the center of the beam can become overheated. The glass transition temperature of the PI attached to the Cu layer is roughly 360– 410 1C (or 633–683 K). Therefore, damage can be expected when the Cu layer is overheated. The heat conductivities of both Cu and the solder paste are high, and thus heat can readily be transferred through the contacting solder layer. Hence a relatively uniform temperature in these layers can be also assumed. However, PI has a low thermal conductivity, and thus is vulnerable to overheating. Consequently, the Gaussian profile of the laser beam (which is desirable for the fine cutting and drilling of materials) may not be appropriate in a bonding process, which requires uniform heat transfer to the sample. A relatively smooth laser beam profile with a low peak at the center seems to be more appropriate for a bonding process (see Fig. 6). When a homogenized heat source is formed on the surface, it produces a fairly uniform temperature distribution on the absorbing layer. To achieve this effect, we employ a diffuser with a micro-lens array (Thorlabs; ED1). The material of diffuser is injected molded ZEONORTM and its index of refraction is 1.53. The size of the diffuser is 1.5 mm thick and 25 mm diameter. The diffusion angle can be limited in 20 degree and the transmission was optimized for 380–1100 nm. As shown in Fig. 7 (source Thorlabs Inc.), the micro-lens array (MLA) is formed on the diffuser surface to convert a Gaussian laser beam input into a square-shaped profile. This reshaped laser spot is used in the reflowing process to improve the bonding quality. The smoothing beam is condensed by a cylindrical lens, so that an elliptical-shaped beam is finally applied to the FPCB surface.
Fig. 7. Photos of the MLA: (a) the magnified lens array and (b) the He–Ne laser beam profile with the lens array.
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
3. Results and discussion 3.1. Numerical analysis of beam homogenization The modified laser energy distribution from the MLA is obtained by numerical analysis. Finite difference time domain (FDTD) is a computational method that can be used to analyze electromagnetic field propagation in a medium, and laser beam propagation can be estimated by this technique [17–19]. The theory of FDTD is basically derived from Maxwell’s equations, given by Eqs. (2) and (3): !
r E ¼ m !
r H ¼e
! @H @t
1647
beam propagation was closer to collimated or parallel beam propagation due to the focusing effect of the lens. To analyze the beam propagation through the combined optical elements (focusing lens þMLA), an additional simulation was conducted, and the results are shown in Fig. 10. As can be seen, the laser homogeneity was much improved by the combined optical elements. Another type of simulation was also used to analyze the beam propagation through the combined optical elements (focusing lens þMLA). This simulation was based on plane waves and a combination of a cylindrical lens and an MLA. As Fig. 11(a) shows,
ð2Þ
! ! @E þs E , @t
ð3Þ
where E is the electric field, H is the magnetic field, e is the dielectric constant, and m is the magnetic permeability. From Maxwell’s equations, the differential form of FDTD can be expressed as Eqs. (4) and (5): F n ði þ 12 ,j,kÞF n ði12,j,kÞ @F n ði,j,kÞ ¼ þOðDx2 Þ @x Dx 1
ð4Þ
1
@F n ði,j,kÞ F n þ 2 ði,j,kÞF n2 ði,j,kÞ ¼ þ OðDt 2 Þ: @t Dx
ð5Þ
Eqs. (4) and (5) can be expanded along the y-axis and z-axis, and the final form of FDTD is then given by Eqs. (6) and (7) 1 nþ2
Hx
ði,j þ 12 ,kþ 12Þ 1 n2
¼ Hx
ði,j þ 12 ,kþ 12 Þ
Dt
m
1 n ðEx ði,j þ 1,k þ 12ÞÞ ðEnx ði,j,kþ 12ÞÞ Dy
1 n ðE ði,j þ 12 ,kþ 1ÞðEny ði,j þ 12,kÞÞ Dz y 1 nþ2
Ex
ðiþ 12 ,j,kþ 12Þ
ðiþ 12 ,j,kÞ ¼ En1 x
1
Dz
1 n2
ðHy
Dt
e
Fig. 8. (a) simulation model and (b) laser beam propagating through air.
ð6Þ
1 1 1 n n ðHx 2 ði þ 12 ,j þ 12,kÞÞ ðHx 2 ði þ 12 ,j12,kÞÞ Dy 1 n2
ði þ 12 ,j,k þ 12 ÞHy
ði þ 12 ,j,k12ÞÞ:
ð7Þ
To analyze the homogeneity of the MLA and inverse Kepler optical device, this FDTD was input to the commercial simulation software EM Explorer (EM Explore Inc, 2007) for comparison of the 3-D lens assembly. Assuming that the software has already been proven for such an application, only a simple beam propagation simulation was used as the validation procedure in this study. The boundary conditions used in the simulation were perfectly matched boundary layer (PML) along the sidewalls of beam propagation and full absorption at the free end of beam propagation. As shown in Fig. 8(a), a laser beam with a wavelength of 1070 nm was irradiated through an excitation port or slit to the air domain (free space). It propagated in spherical form, with weaker intensity along the propagation direction, as shown in Fig. 8(b). As expected, the laser beam propagates through open air domain. Under the same conditions depicted in Fig. 8, a focusing lens with an index of refraction of 1.4 was used to converge the laser beam to the air domain, as shown in Fig. 9. As expected, the divergence of the laser decreased, and the laser beam was focused along the center line in the direction of propagation. Compared to the free-space propagation experiment shown in Fig. 8, the laser
Fig. 9. (a) simulation model with focusing lens and (b) laser beam propagating through air.
Fig. 10. (a) simulation model with focused lens and MLA and (b) laser beam propagating through air.
1648
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
Fig. 11. Plane laser beam propagation through air with (a) cylindrical lens only and (b) (c) cylindrical lens and MLA.
Fig. 12. Relative intensity distribution (A.U.) at the give location (mm) for (a) cylindrical lens only.
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
the laser beam propagates through the cylindrical lens and converges on a focusing point. The intensity of the laser beam is very high in the center and relatively low at the edges. To improve the homogeneity of the propagating beam, an MLA was placed in the front of a cylindrical focusing lens, as shown in Fig. 11(b) and (c). The intensity of the laser beam after passing through the MLA was more randomized in comparison with the incoming beam. To quantify the homogeneity of the laser beam, the intensity was measured along the transverse direction and compared with that obtained from an optical element without an MLA. The simulation times related to beam propagation and measurement location were the same in both cases, and intensity data were extracted as shown in Fig. 12. To quauntify the effectiveness of MLA, the intensity of incoming beam was measured at the full width half max location and all measured intensity was relatively scaled to this value. The plane wave and cylindrical lens assembly produced an intensity 2.1 times that of the incoming laser beam, whereas the cylindrical lens with the MLA produced an intensity only 1.6 times that of the incoming beam. The measured width of homogeneity at FWHM was 20 for a cylindrical lens only and 40 for a cylindrical lens and MLA combination. A detailed comparsion is presented in Table 3. 3.2. Numerical analysis of temperature distribution Temperature distribution is an important factor in the design of the bonding process, as estimation of bonding temperature and energy is one of the keys to successful reflowing. However, when a laser is used as a heating source, optical properties of the materials (including reflection and absorption) must be considered, unlike regular contact heat sources such as hot bars and hot plates. The energy transfer rate via radiation is related to the material surface properties. In particular, it is a function of temperature. Moreover, the energy distribution in the laser spot is also non-homogeneous and usually has a Gaussian distribution in a single-mode laser beam. Thus, in order to accurately estimate the temperature distribution, the factors mentioned above must be considered carefully. In addition, the FPCB, solder paste, and PCB have complicated geometries, including multilayer materials, and thus the temperature distribution for the process should be obtained by numerical analysis. The irradiation time for laser reflowing is more than a few ms, which is much larger than the relaxation time scale of the materials, and thus local thermodynamic equilibrium can be assumed. In this case, the temperature distribution in the materials can be estimated using the Fourier diffusion equation, which is given by
rcp
@T ¼ rUðkrTÞ þ Q ðx,z,tÞ, @t
ð8Þ
where T denotes temperature, t denotes time, r is the material density, Cp is the specific heat of the materials, k is the heat conductivity, and Q is the absorbed power density on the surface. In our experiment, a Gaussian beam was passed through a cylindrical lens and an MLA, and the final energy distribution was Table 3 Summary of EM simulation data (Relative intensity of laser beam). Description
Cylindrical lens only
Cylindrical lens þMLA
Maximum Half maximum Rate of above half maximum (%)
2.1 1.05 37%
1.6 0.8 69%
*The unit in this table is relative intensity of incoming beam (full width half max).
1649
investigated using the above equation. The exact energy distribution is elliptical in shape, but we can roughly model the crosssectional energy distribution as follows: I 2x2 Q ðx,z,tÞ ¼ ð1RÞ exp 2 expðazÞ, ð9Þ 2Ux0 Ul x0 where R is reflectivity, I is the laser intensity, x0 is the laser spot radius, x is the radial distance from the center of the laser spot, l is the length of the beam spot, a is a distinct constant, and z is the coordinate of the depth direction. The above equation does not represent the exact energy distribution of the modified laser beam because the unmodified beam has a Gaussian distribution, and the effect of the MLA is not considered. However, it is impractical to use an exact analytical expression to describe the beam, since the presence of both an MLA and a cylindrical lens makes the beam shape rather complicated. We used a discrete energy distribution in the numerical simulations instead of the exact analytical equation. Boundary conditions must be imposed to solve the partial differential equation, and an adiabatic condition is applied to the surface exposed to the air due to its high thermal resistance: @T ¼0 ð10Þ @x x ¼ 0 Moreover, axial symmetry is assumed in the radial direction, and is expressed as follows: @T ¼0 ð11Þ @z z ¼ 0 In addition, the FPCB substrate is relatively large compared to the solder paste, and thus a constant temperature condition at the boundary of the FPCB can be used: Tðx-1Þ ¼ 300 K:
ð12Þ
The initial temperature is taken to be room temperature and can be written: Tðt ¼ 0Þ ¼ 300 K:
ð13Þ
After laser irradiation, the following governing equation can be applied to determine the temperature distribution with similar boundary conditions:
rcp
@T ¼ rUðkrTÞ: @t
ð14Þ
An exact analytical solution for a real geometry is difficult, and an approximate temperature profile can be obtained via a numerical approach. The temperature distribution in the materials is important when the temperature of the solder paste reaches the reflowing point. A numerical simulation of the transient temperature profile was performed using the commercial software FLUENT. The laser beam spot diameter was set at 3 mm, and the laser energy input was applied to the top of the FPCB. The energy distribution of the top layer was approximated via the previous numerical results, and the optical and material properties were assumed to be constant with respect to temperature. The temperature range from room temperature to bonding temperature was less than 500 K. Under the above conditions (Fig. 13(a)), the resulting transient temperature distributions of the materials are shown in Fig. 13(b). The temperature of the copper layer increased to the melting point of the solder paste in 100 ms, with a steady laser power input of 5 W/mm2, and the heat diffused to the solder layer in about the same time. The FPCB layer was heated by conduction, mainly due to its low laser absorption rate. When an MLA is used, a more homogenized laser input and temperature distribution can be obtained. This was confirmed by the simulation results (Fig. 13(c)), and the bonding quality was improved by using an
1650
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
Fig. 13. Temperature distribution: (a) schematics for numerical modeling (b) Gaussian beam input (c) Gaussian beam input with MLA (d) photo of fractured sample.
MLA. Comparing temperature gradients at the interfaces of solder and the connector, we can indirectly expect a possible location of fracture induced thermal strain as shown in Fig. 13. As shown in Fig. 13(b) and (c), there is a rapid temperature change at the interface without the MLA combination setup (Fig. 13b) which may result in possible fracture.
4. Experimental results and discussion 4.1. Bonding results An experimentally measured temperature profile of the FPCB and solder paste provides information on energy transfer during the process. Thermocouples were attached on the surface of FPCB layer which locations (A,B, and C) are described in Fig. 14, and transient and spatial temperature profiles were obtained. Fig. 14 shows the temperature history with respect to time for a laser intensity of 2–5 W/mm2 and a scanning speed of 200 mm/s. The temperature of the solder paste increased to 220 1C (or 493 K, the melting temperature of the solder) in 200 ms. This result indicates that the ramp rate of the laser process exceeds 1100 K/s, which is faster than conventional processes using contact heating (such as hot bar and hot plate processes). Similar measurements were obtained with an MLA, except that the maximum temperature at the center of the FPCB connector was lower due to energy loss from the modified laser beam. Moreover, it is believed that the high heat-conduction rates of copper and solder paste led to a
Fig. 14. Temperature profile with respect to time during laser irradiation.
quick response to energy transfer via laser irradiation. The effects of the diffuser were confirmed by the spatial temperature profile shown in Fig. 15. The results are in good agreement with previous numerical results, and the diffuser induced a fairly homogeneous profile in the materials. The temperatures at the left and right ends differed because one end was exposed to the PI layer, in
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
which different boundary conditions must be applied. On the basis of these results, we may conclude that this type of beam shaping can improve the quality of the bonding between the materials. The cross-sectional views of the samples bonded via the laser process are shown in Fig. 16. When beam shaping with an MLA was carried out at an appropriate energy level, a uniform bonding layer was achieved between the FPCB and the substrate (Fig. 16(a)). Bonding failure caused by an excessive energy level is shown in Fig. 16(b). Microscopic views of the reflowing solder
Fig. 15. Temperature profiles with respect to the X-cooridnates of the FPCB during laser irradiation.
1651
paste are shown in Fig. 16(c) and (d). The melted solder paste flowed onto the copper connector and formed a cover layer with a thickness of 1 2 mm. It is thought that this kind of reflowing behavior improves the mechanical and electrical reliability of the connecting parts. 4.2. Mechanical bonding strength measurement A shear test was performed to confirm the mechanical bonding quality of the laser process using a micro-shear tester (Instron 5800). A bonding sample was taken with 1 mm wide and 5 mm long and was mounted on a micro-jig, and shear force was applied until a maximum load was reached and the sample fractured with a speed of 2 mm/min. We used the same rate of elongation of the
Fig. 17. MLA effect on shear strength.
Fig. 16. Photographs of the bonded FPBCs (a) a good bonding result of the copper interconnections of the substrate (b) a bad bonding result of the copper interconnections of the substrate (c) SEM of joined part (d) SEM photographs of the copper interconnections of the substrate.
1652
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
sample to synchronize the starting point and the maximum load, and the elongation was measured at this point. The bonding strength (P) is defined as follows: P¼
L , Aef f
ð15Þ
where L is the applied load, and Aeff is the effective bonding area, which is around 0.4 mm2. The bonding strengths of the samples with and without an MLA are shown in Fig. 17. When the same amount of laser energy was applied for both Gaussian shape and MLA combination, the sample that was bonded with a homogenized laser beam exhibited higher shear strength during elongation, indicating that a reshaped laser beam with a MLA can broaden the effective bonding area of an FPCB and thereby improve the bonding quality. It should be noted that the comparison between the bonding strengths with and without beam shaping was made under the assumption that the total bonding energy was the same in both cases, as energy loss can be expected when the laser beam is reshaped using an MLA, and the resulting energy could be less than what is required for the bonding process. In this case, the comparison would be meaningless. Thus, when an MLA was used, the amount of laser beam energy lost due to the MLA was compensated for the sake of the comparison. Appropriate energy input is important for successful bonding. Too much energy can cause the polymer layer of the FPCB to overheat, which may harden the polymer layer and lead to thermal damage of the FPCB, resulting in a defective bond at the interface. Moreover, too little energy may fail to raise the temperature of the solder paste to the melting point. As it is rather difficult to examine the bonding quality directly, shear strengths were measured with respect to various laser power densities, as shown in Fig. 18. When the power density of the laser beam was 10% higher or lower than the optimized bonding power, the bonding sample exhibited a low bonding strength over the range of elongation, which implies that the laser power density must fall within that window.
a precise amount of energy to a local bonding area so that thermal damage to adjacent materials can be minimized. The temperature ramp rate exceeds 1100 K/s, and thus bonding can be accomplished quickly. In order to improve the laser reflowing quality, we modified the Gaussian beam to a somewhat flattened shape with an MLA and a cylindrical lens, resulting in a more homogenized temperature profile on the FPCB and solder paste. By electromagnetic field modeling, the modified laser beam was homogenized 30% more than the original beam. A concentrated laser beam may cause damage to a polymer layer by overheating, and this kind of beam shaping can improve the bonding strength. A thermal diffusion model and simulation were presented to confirm the reflowing effect in the material, and it was determined that the solder paste reaches the melting point in roughly 100–200 ms with a laser power density of 5 W/cm2. The temperature profile on the FPCB was measured experimentally, and the measured results were shown to be in good agreement with the calculated results. The effects of beam shaping on laser reflowing were verified with mechanical shear tests. The maximum shear strength increased when beam shaping was used, and it was inferred that the mechanical reliability of bonding layers can be improved by beam shaping. The mechanical strength of the bonding layer diminished when the laser bonding energy increased or decreased beyond a certain level, clearly implying that the laser power density must fall within a definite window to produce a good bond. Based on these results, it can be concluded that the laser reflowing procedure is an effective process for FPCB bonding. The principal challenge in applying this process will be determining how to control the laser power density on the FCPB.
Acknowledgments This work was supported by the National Research Foundation of Korea (no. 2011-0004266) and the authors thank Mr. B. H. Park for technical supports. References
5. Conclusions A laser reflowing process for bonding FPCBs was presented. This procedure offers fast, precise, and reliable bonding characteristics compared with conventional contact methods. The process transfers
Fig. 18. Laser power density effect on shear strength.
[1] Morrissey A, Kelly G, Alderman J, Barrett J, Lyden C, O’Rourke L. Some issues for microsystem packaging in plastic and 3D. Microelectr J 1998;29:645–50. [2] Van Brussel H, Peirs J, Reynaerts D, Delchambre A. Reinhart, Roth A, Weck M, Zussman E. Assembly of Microsystems. CIRP Ann—Manuf Technol 2000;49:451–72. [3] Shen K, Chen X, Guo M, Cheng J. A microchip-based PCR device using flexible printed circuit technology. Sens Actuators, B 2005;105:251–8. [4] Berenyi R. Prototyping of a reliable 3D flexible IC cube package by laser micromachining. Microelectron Rel 2009;49:800–5. [5] Giovangrandi L, Gilchrist K, Whittington R, Kovacs G. Low-cost microelectrode array with integrated heater for extracellular recording of cardiomyocyte cultures using commercial flexible printed circuit technology. Sens Actuators, B 2006;113:545–54. [6] Svedberg M, Nikolajeff F, Thornell G. On the integration of flexible circuit boards with hot embossed thermoplastic structures for actuator purposes. Sens Actuators, A 2006;125:534–47. [7] Zhou J, Zhuang L, He Z, Herscovici D. Practical design, simulation, and prototype of a hot bar blade for printed circuit board soldering. J Manuf Syst 2001;20:177–87. [8] Darwish S, Al-Habdan S, Al-Tamimi A. A knowledge-base for electronics soldering. J Mater Process Technol 2000;97:1–9. [9] Kim K, Jung J, Zhou Y. Ultrasonic bonding of flexible PCB to rigid PCB using an Sn interlayer. Solder Surf Mt Technol 2009;21:4–10. [10] Lee J, Koo J, Hong S, Shin S, Moon Y, Jung J, et al. Longitudinal Ultrasonic Bonding of Electrodes between Rigid and Flexible Printed Circuit Boards. Jpn J Appl Phys 2008;47:4300–4. [11] Ille´s B, Harsa´nyi G. Investigating direction characteristics of the heat transfer coefficient in forced convection reflow oven. Exp Therm Fluid Sci 2009;33:642–50. [12] Illyefalvi-Vite´z Z. Laser processing for microelectronics packaging applications. Microelectron Rel 2001;41:563–70. [13] Chaminade C, Fogarassy E, Boisselier D. Diode laser soldering using a leadfree filler material for electronic packaging structures. Appl Surf Sci 2006;252:4406–10.
J. Kim, H. Choi / Optics and Lasers in Engineering 50 (2012) 1643–1653
[14] Tian Y, Wang C, Ge X, Liu P, Liu D. Intermetallic compounds formation at interface between PBGA solder ball and Au/Ni/Cu/BT PCB substrate after laser reflow processes. Mater Sci Eng B 2002;95:254–62. ¨ ¨ [15] Stauffer L, Wursch A, Gachter B, Siercks K, Verettas I, Rossopoulos S, et al. A surface-mounted device assembly technique for small optics based on laser reflow soldering. Opt Laser Eng 2005;43:365–72. [16] Lee J, Lee Y, Kim Y. Fluxless laser reflow bumping of Sn–Pb eutectic solder. Scripta Mater 2000;42:789–93. [17] Jian Q, Morita N, Shi D, Nakamura O, Liu S. FDTD Simulation of Nonlinear Ultrasonic Pulse Propagation in ESWL. IEEE-EMBS. In: 27th Annual international conference of medicine and biology society; 2005. p. 1806–9.
1653
[18] Sakai A, Baba T. FDTD simulation of photonic devices and circuits based on circular and fan-shaped micro disks. J Light wave Technol 1999;17:1493–9. [19] Slavcheva M, Arnold M, Ziolkowski W. FDTD simulation of the nonlinear gain dynamics in active optical waveguides and semiconductor micro-activities. IEEE J Quantum Electron 2004;10:1052–62. [20] Tian D, Wang C, Tian Y. Effect of solidification on solder bump formation in solder jet process:Simulation and experiment. Trans Nonferrous Met Soc China 2008;18:1201–8.