We present an optimal laser welding assembly sequence for butterfly laser packages: 1) initial shift, 2) front welding, 3) rear welding, 4) joint gripper releasing, 5) mechanical fine tuning of horizontal misalignment. This sequence has been optimized significantly by modeling the initial shift and experimental investigations of three assembly sequences. Our results show that misalignment from the Post-Weld-Shift (PWS) can be compensated by accurately estimating the initial shift in the vertical direction. Furthermore, the laser hammering procedure, to compensate for misalignment of the vertical direction, can be eliminated by proper package design. Using only final mechanical tuning for horizontal misalignments, optical coupling efficiencies of 73-99% have been achieved for lasers packaged in butterfly modules.
©2009 Optical Society of America
A Nd:YAG laser welding technique is commonly used for fiber coupling conventional semiconductor lasers inside butterfly packages . During the welding process, rapid solidification of the welded region and the associated material shrinkage causes a Post-Weld-Shift (PWS) between the metal weld clip and the ferrule. If the PWS is even a few micrometers, then up to 100% of the coupled power can be lost. The misalignment induced by the PWS is typically in the range of a few microns to over 10μm . The PWS can be compensated by a laser hammering technique using a seesaw effect. Typically, the Nd:YAG laser welding sequences are 1) front welding, 2) rear welding, and 3) laser hammering. After front and rear welding, misalignment from the PWS can be compensated by a few implementations of the laser hammering procedure. The laser hammering technique is commonly used for compensating vertical (Y) direction misalignment. The horizontal (X) direction misalignment relies on symmetrical laser welding and the PWS is smaller than Y-directional misalignment because of the same strength and opposite direction of shifting stresses. However, laser hammering procedures make production costs higher, assembly time longer, production yield lower, and reduces reliability. Previously, we proposed a highly efficient sequence  for the laser welding technique used to assemble fiber coupled semiconductor laser packages. The misalignment from the PWS caused when joining a nickel clip to the metal ferrule can be compensated by an optimal initial shift prior to front Nd:YAG laser welding and a minor mechanical adjustment prior to the laser hammering process. As a result, the number of laser hammering sequences required to compensate for misalignment have been significantly reduced. In this paper we present the optimized laser welding sequences for butterfly laser module packages, designed to remove the laser hammering procedures for vertical misalignment compensation, by analyzing laser welding sequences and optimizing the package structure design.
2. Lensed fiber design and experimental procedure
High coupling efficiencies and large misalignment tolerances are required between a lensed fiber and a semiconductor laser for fiber coupling. For efficient fiber coupling to the semiconductor laser, 11µm radius of curvature lensed fibers with anti-reflecting (AR) coatings were used in the experiments. It was previously proposed that this choice of lensed fiber is an optimized and universal solution  when packaging semiconductor lasers with divergence angles ranging from 5° to 30°, as shown in Fig. 1 . The tip of the lensed fibers has been designed from the overlap integral method  and this method can be used to estimate the alignment tolerance between a lensed fiber and semiconductor laser. Figure 2 shows the relationship between curvature radii and the lateral and longitudinal alignment tolerances. Large curvature radii lead to larger lateral tolerances, however coupling efficiency is reduced as shown in Fig. 3 . The coupling efficiency as a function of various curvature radii of lensed fibers was calculated with a divergence angle of 10° × 15° from the lasers used in the experiment. Experimental alignment tolerances using 11µm-curvature radius lensed fibers and 10° × 15°-divergence angle lasers are shown in Fig. 4 .
Table 1 shows the three kinds of laser welding sequences (S1, S2, and S3) used in our experiments. Using these sequences, we have assembled 13 semiconductor laser modules. The butterfly laser packages were assembled by joining a nickel weld clip and a metal ferrule using an Nd:YAG laser (Newport Corporation laser welding workstation). The two welding beams are rotated by 180 degrees from each other and 30 degrees off vertical. A total of four welds are placed between the weld clip and the metal ferrule in pairs, two welds in the front (near the laser diode) and two in the rear, as shown in Fig. 5 . The base holder, which can move in the lateral, longitudinal, and angular directions, holds the semiconductor laser in the butterfly package and a joint gripper, which can move in the vertical direction, holds the end part of the metal ferrule for accurate positioning of the optical fiber relative to the semiconductor laser.
3. Modeling of initial shift in laser welding assembly sequences
To determine the optimum values of the initial shifts in the sequence, we modeled the misalignment induced by the PWS. The PWS is defined as  where ∆X, ∆Y and ∆Z are the values of the shift in each direction, where X, Y, and Z are the lateral, transverse and longitudinal directions, respectively. We can model the horizontal (X) and vertical (Y) misalignments using the configuration of structures as shown in Fig. 5. In this model, the longitudinal (Z) direction can be ignored because of large tolerances, as shown in Fig. 4. The misalignments in both directions can be expressed by
There is a symmetric ~2 µm gap between weld clip and metal ferrule as shown in Fig. 5. The optical loss due to intrinsic angular misalignment can be ignored when less then ± 2°, as this angular offset does not affect the coupling efficiency, as shown in Fig. 4. For gaps of 2µm and L of approximately 3mm, the angular offset is less then 0.2°. During the laser welding process, the metal ferrule can move a maximum lateral distance of 2µm and a maximum distance of 7µm due to angular misalignment [ = Ltan(∆θx)]. ∆x and ∆xi are the lateral offset and intrinsic distortion from laser welding, respectively. Here, ∆xi≈0.11µm . ∆θx is an angular offset. ∆yf_net, and ∆yr_net are the net movement values by front and rear welding without stress, respectively. ∆yi is the intrinsic distortion from the welding process. ∆yf_weld, ∆yr_weld and ∆yrelease are vertical misalignments when angular misalignments are induced by front welding, rear welding and joint gripper release, when holding the metal ferrule by the joint gripper. They can be expressed by following equations.
Modeling of vertical misalignment is important for choosing the value of the initial shift. From this model, we can determine the value of initial shift of ∆yinitial using the following equation.
To solve Eq. (9), we assume that as the metal ferrule will shift downward due to the associated material shrinkage and the direction of is opposite to that of due to the seesaw effect . ∆yf_weld, ∆yr_weld and ∆yrelease have been measured to be approximatelty 2, 0, and 5µm, respectively. Using typical misalignment values from our previous work [2,3], we assume that values of ∆yf_net are in the range of 10-20µm. Therefore, ∆yr_net is in the range of 13-17µm. The values of ∆θy_gripper, ∆θy_pivot1 and ∆θy_pivot2 are in the range 0.07-0.15°, 0.29-0.39° and 0.11°, respectively using Eq. (4), (6)-(7). Here, the values of Lgripper and Lpivot are approximately 8mm and 2.5mm, respectively. The intrinsic distortion (∆yi) value of ~0.11µm of laser welding was taken from . We chose the maximum misalignment cases of ∆θy_gripper, ∆θy_pivot1 and ∆θy_pivot2 under the assumed conditions to calculate proper initial shift values. Figure 6 shows the vertical misalignments as a function of the initial shifts. The optimum values of the initial shifts are from −3µm to + 13µm under these assumptions. To reduce the range of calculated initial shift values, the status of weld clip and metal ferrule was assumed to be . This means that the initial shift should be upward. Therefore, the range of initial shifts will have a restricted range from 0 to + 13µm.
4. Experiments and analysis
We assembled three modules with initial offset values of + 5µm, + 8µm and + 10µm and assembly sequence S1 [2,3]. The sequence S1 has no horizontal misalignment compensation under the assumption of a small misalignment of ∆X because of symmetric welding as discussed in the Introduction. We used the initial shift of + 5µm, + 8µm, and + 10µm for laser modules L1, L2, and L3, respectively. Table 2 summarizes the results of the three laser modules. The results show that horizontal misalignment compensation was required, as indicated by the coupling efficiencies listed in the Table 2. Moreover, the initial shifts of + 5µm, + 8µm and + 10µm were optimum values as predicted by modeling of initial shifts in the vertical direction. Here, the value of coupling efficiency refers to the optical power after performing the final sequence with respect to the maximum coupling power at step #0.
Using the data obtained from the S1 assembly sequence and the initial shift of 5µm, five modules were assembled by using the S2 sequence, as summarized in Table 3 . The laser hammering procedure for vertical misalignment compensation was significantly reduced by performing fine mechanical tuning for horizontal misalignment compensation prior to laser hammering. However, the S2 sequence still requires one or two repetitions of the laser hammering procedure to compensate for vertical misalignment.
Using the results of obtained from the S1 and S2 assembly sequences, we analyzed the laser package design in order to optimize the assembly sequence. The process of joining the nickel weld clip to metal ferrule in the packaged modules assembled by S1 and S2 is described as Type A, as shown in Fig. 7 . The structure of Type A has a gap of approximately 100µm between the center of the metal ferrule and the wall of the weld clip. The assembly tolerance between the wall of weld clip and metal ferrule is ± 210µm in height. The diameter of the focused laser welding spot is approximately 500µm which can cover the assembly tolerance of ± 210µm. Therefore, the gap of approximately 100µm is not critical. However, in order to eliminate the laser hammering procedure, the gap between the weld clip and metal ferrule should be as small as possible. Type B is designed such that the weld joint (between weld-clip and metal ferrule) is positioned at the same height as the center-line of the metal ferrule as shown in Fig. 7. Using a Type B structure, we assembled five laser modules and the results are summarized in Table 4 . Using this optimized structure, the laser hammering procedure is not required for vertical misalignment compensation and only a minor mechanical adjustment was required to tune for horizontal misalignments. As a result, the laser welding assembling sequence can be optimized by S3: 1) initial shift ( + 5µm), 2) front welding, 3) rear welding, 4) joint gripper releasing, 5) mechanical fine tuning for horizontal misalignment. Figure 8 shows the misalignment analysis of using S1, S2 and S3 after the procedure #4. The misalignment can be estimated using the optical coupling powers. The misalignment direction can be observed by applying a small amount of pressure on the metal ferrule - an increase or decrease in optical power will determine the direction of misalignment. After performing procedure #4, it was observed that the coupled optical power in the vertical direction was constant when pressure is applied to the metal ferrule. The coupled optical powers were further increased by performing procedure #5 in the horizontal direction. The coupled optical powers for each procedure (S1, S2, and S3) are summarized in Table 5 . All lasers used in the experiments have similar divergence angles (~10° × 15°) as indicated in the normalized powers. After horizontal misalignment compensating by fine mechanical tuning, the coupling efficiencies of the packaged modules were in the range 73-99%. Figure 9 shows a micrograph of the assembled laser module using the optimal S3 assembly sequence.
An optimal laser welding assembly sequence for butterfly laser packages has been proposed. This is based on modeling of the initial shift, analysis of previous work, and optimization of package design. This optimal assembly sequence consists of the following steps: 1) initial shift, 2) front welding, 3) rear welding, 4) joint gripper releasing, and 5) mechanical fine tuning. The optimal sequence is a major improvement over existing packaging procedures, as it does not require vertical misalignment compensation. As a result, laser hammering has been completely eliminated from the laser welding process. Using the optimal sequence, five butterfly laser packages were assembled without the laser hammering process and coupling efficiencies of in the range 73-99% were measured. This process simplification should lead to a more time-efficient solution for optoelectronic components assembled using the laser welding technique.
This work was supported by the Science Foundation Ireland under Grant 07/SRC/I1173.
References and links
1. S. Jang, “Packaging of photonic devices using laser welding,” SPIE 2610, 138–149 (1996). [CrossRef]
2. J. H. Song, H. N. J. Fernando, B. Roycroft, B. Corbett, and F. H. Peters, “Practical design of lensed fibers for semiconductor laser packaging using laser welding technique,” J. Lightwave Technol. 27(11), 1533–1539 (2009). [CrossRef]
3. J. H. Song, B. Roycroft, B. Corbett, and F. H. Peters, Tyndall National Institute, Cork, Ireland, are preparing a manuscript to be called “Experimental investigation of laser welding assembling sequences for butterfly laser module packages.”
5. Y.-C. Hsu, Y.-C. Tsai, J.-H. Kuang, M.-T. Sheen, P.-H. Hsu, and W.-H. Cheng, “A notch-saddle-compensation technique in butterfly-type laser module packages,” J. Lightwave Technol. 25(6), 1594–1601 (2007). [CrossRef]
6. Y. Lin, W. Liu, and F. G. Shi, “Laser welding induced alignment distortion in butterfly laser module packages: effect of welding sequence,” IEEE Trans. Adv. Packag. 25(1), 73–78 (2002). [CrossRef]
7. Y. Lin, C. Eichele, and F. G. Shi, “Effect of welding sequence on welding-induced-alignment-distortion in packaging of butterfly laser diode modules: simulation and experiment,” J. Lightwave Technol. 23(2), 615–623 (2005). [CrossRef]