Packaging of fiber collimators

A novel automation process for photonic devices


Packaging and assembly still dominate the overall cost of any fiber optic devices,1 largely because fiber optic alignment and attachment are difficult and time consuming.2 Attachment with UV adhesives offers advantages in terms of cost and ease of use compared to other attachment technologies, such as laser welding and soldering.2 However, the adhesive curing and its subsequent cooling processes may cause large displacements leading to fiber optic misalignment. Many studies on epoxy bonding have been undertaken with optical fibers to waveguide devices and lasers to fibers.3-5 This work presents the development of automated packaging using UV adhesives for alignment and attachment of 8-channel fiber collimators for photonic devices.

Challenges for Fiber Collimators

Among optical components, optical switches, isolators, attenuators and multiplexer/demultiplexers are the key devices for the next generation wave division multiplexing (WDM) lightwave network.6-9 These photonic devices consist primarily of arrays of optical elements, such as vertical cavity surface-emitting lasers (VCSELs), photodetectors, lenses and fibers.10-11 Integration of array-to-array through optical interconnect technology is the most promising field to be explored to exploit high-speed CMOS technology. For interconnect technology, one of the most promising ones emerging is free space optical interconnection12-13 where signals are sent through space to interconnect two optical elements. In free space based fiber optic components, single mode or multi-mode fiber collimators have been widely used because the coupling between two fiber collimators has a large allowable separation distance with a low loss, which is critical for a practical free space interconnected fiber optic module or subsystem.

Figure 1. Schematic of a 1 x 8 collimator array, showing (a) the top and (b) side views.
Click here to enlarge image

Thus, the packaging of free space based fiber collimators for photonic devices using adhesives provides a possible solution for the tremendous demand of optical components. However, free space interconnection suffers from dif-ferent misalignment loss as well as cou-pling loss because of scattering and diverging as the signal passes long distances in space. In addition, the shrink- age of the adhesive during curing may cause large movements, which can affect the alignment.14 Also, there can be long-term reliability issues because of a stress build-up from uneven distribution of the cross-linking of the monomer of the adhesive, high adhesive modulus, and the thermal expansion mismatch between the adhesives and the components. It is, therefore, important to solve these important engineering issues of yield and reliability with a proper optical design and an optimal joint design with large misalignment tolerances, high coupling efficiency and reliable joints. The emphasis of this work is on the comparison of three UV-curing adhesives (identified as A, B and C) in light of these requirements.

Automated Fiber-to-Lens Array Alignment

A 1 x 8 collimator array composed of a 1 x 8 fiber array and a 1 x 8 lens array is schematically shown in Figure 1, where the pitch between two channels is up = 1.25 mm, the width of the array is W = 11 mm, and the height of the array is 2 mm.

Figure 2. (a) Angular and (b) lateral misalignments of the collimator.
Click here to enlarge image

The tolerance is the maximum value of distance and angle at which good coupling efficiency can be achieved. For example, during the assembly process of the collimator array, a 1 x 8 single mode fiber array (FA) is aligned to a 1 x 8 lens array (LA) made of quarter-pitch gradient-index lenses, and the lateral and angular misalignment will cause the loss of coupling power. A larger alignment tolerance results in easier assembly of the collimator array. Therefore, it is important to know the tolerance distribution for the collimator. The misalignments between fiber array and lens array include lateral offset (d) and angular misalignment (θ), as shown in Figure 2.

The expression for coupling efficiency of such an optical scheme has been previously derived.15 From this calculation, for a collimator length of 4.8 mm and a fiber spot size of 4.5 µm, the maximum coupling efficiency results when the focusing distance (the spacing between the lens and the tip of the fiber) is 15 µm, and the collimating distance (the spacing between the lens arrays of the two collimators being aligned) is 20 mm. This optimized focusing and collimating distance is thus adopted in our joint design and assembly of the collimator arrays. The misalignment tolerance analysis for such a configuration is summarized in Table 1, where pitch, roll and yaw are the rotation of the fiber array around x, y and z directions, respectively. It is seen from the table that the roll misalignment in the present array arrangement is the most significant. In fact, in the case of roll misalignment, the fiber array rotates around an axis along the fiber direction. Therefore, the fiber end and lens end remain parallel to each other, and no angular misalignment but a large lateral offset is introduced between them. The magnitude depends on the position in the array.

Table 1. Summary of misalignment tolerance.
Click here to enlarge image

To perform the fiber-optic alignment automation, an auto alignment workstation system was used. This automated alignment, characterization and bonding system is a modular process development and production automation platform designed to measure, develop, and automate photonics device measurement and assembly processes.

Table 2. Material properties of the adhesives used in these evaluations. (Not all data was available.)
Click here to enlarge image

An active alignment of the fiber array to lens array was select-ed for the assembling process. A mirror working as a reflector was placed half the collimating distance from the collimated fiber array. The reflected light was then back-coupled into the fiber array. Both the input light and the output light power were measured by a built-in power meter, and were recorded for power loss analysis. The input light power was from a well-stabilized power source with a wavelength of 1,310 nm.

Fiber and Lens Array Attachment Using Adhesives

Adhesive joint design: After the lens-fiber array is aligned, the lens-fiber array must be fixed. UV curing adhesive bonding is a low-cost and convenient method. However, the shrinkage of the adhesive materials during the curing and the thermal expansion mismatch of various materials used in the structure can cause deformation of the structure. This will lead to displacements between the lenses and fibers, resulting in the coupling loss of the lens-fiber array. A proper adhesive joint design and, regardless of curing shrinkage, control of the bondline thickness can minimize the coupling loss.

Table 3. Epoxy dispensing and curing settings.
Click here to enlarge image

Displacement tolerance: The finite element method (FEM) was used to predict the displacements induced among the optical elements during the UV curing process. As shown in Table 1, it was found that the coupling loss is very sensitive to the roll misalignment, because a small roll angle will lead to a large relative displacement at the edges of the fiber and lens arrays. The coupling loss caused by the displacement tolerance in the lens-fiber assembly is shown in Figure 3.

Figure 3. Coupling loss as a function of displacement tolerance.
Click here to enlarge image

Bondline thickness design: The thickness of the adhesive layer in the structure has a significant influence on the displacement between the lens array and the fiber array, and therefore on the coupling loss of the lens-fiber assembly. The relationship between the displacements induced during the UV curing process and the adhesive thickness for three commercially avail-able adhesives (A, B and C, whose properties are summarized in Table 2) is depicted in Figure 4. The displacement reaches its maximum value at an adhesive thickness that results in the maximal coupling loss in the optical assembly. The maximum value of the coupling loss induced during the curing process may be as high as 0.5 dB. On the other hand, the coupling loss induced by the curing for all three adhesives of different shrinkage is always less than the optical specification of 2 dB. Therefore, if the lens-fiber arrays are aligned with high coupling efficiency, (e.g., a coupling loss of 1 dB or less), the attachment structure with different adhesive thickness and materials will still have a good coupling efficiency.

Figure 4. Curing induced displacement vs. adhesive thickness.
Click here to enlarge image

Epoxy dispensing and bonding: The adhesives were dis-pensed after the full alignment of all eight channels, and the dis- pensing and bonding conditions are summarized in Table 3. The epoxy dispensing system was included in the auto align-ment workstation system. The intensity and duration of the UV light was controlled. With the bondline thickness controlled in the range of 80 to 100 µm, the adhesives were applied to a glass plate, which was then put on the fiber and lens arrays. The adhesive was then cured.

Collimator Assembling

The process flow for assembling the collimator is:

  • Load 1 x 8 fiber array and lens array
  • Align fiber array and lens array to optimal coupling position
  • Measure the coupling loss of each channel
  • Attach a plate with adhesive between the fiber and lens arrays
  • UV cure the adhesive
  • Measure the coupling loss of each channel after UV cure.

Figure 5. Schematic set-up of the 1 x 8 collimator array assembly using UV curable adhesives.
Click here to enlarge image

The structure during UV curing is shown in Figure 5. By executing the alignment sequences, the fiber and lens arrays were aligned and adjusted in six directions (x, y, z, pitch, yaw and roll) until a final optimal coupling position was reached. After UV curing, the curing-induced shrinkage will affect the alignment between the fiber array and the lens array. The new insertion loss was measured, and the difference of the insertion loss before and after the attachment was calculated.

Results and Analysis

The coupling loss with the optimal optical and joint structure design was less than 1.5 dB for all channels, with a variation of less than 0.5 dB among all eight channels.

Figure 6. A comparison of the coupling loss due to the attachment process using three different adhesives. It is noted that the loss is within the 2 dB specification.
Click here to enlarge image

Experimental results of insertion loss for eight channels after the alignment (before the attachment) and after attachment for all three adhesives are summarized in Figure 6. It is evident that for all of the adhesives, the coupling loss after the alignment and attachment, which is a characterization of the insertion loss caused by the adhesives shrinkage after UV cur-ing, is within the 2 dB specification. The optical performance does not show much difference, so it can be concluded that the three types of adhesives are suitable for packaging the present collimator.


The development of automated alignment and the attachment processes for the packaging of 1 x 8 collimator arrays, which are critical for many photonic devices (including MEMS-based optical switches), is summarized in this work. The feasibility using three UV-curing adhesives for bonding both fiber and lens arrays is investigated. It is concluded that regardless of the different degrees of curing shrinkage, a proper adhesive joint design can alleviate the optical loss effect of the displacement induced during UV curing and the subsequent cooling processes. AP


  1. M. Shaw, M. Marazzi and S. Bonino, ECTC Proceedings 2000, pp. 742-748.
  2. J. Liu et al, ECTC Proceedings 2000, pp. 975-980.
  3. D. Wu and Y. Lee, ECTC Proceedings 1998, pp. 1186-1191.
  4. T. Kokubun and Y. Katsuyama, Journal of Lightwave Technology, 8(11), 1990, pp. 1654-1658.
  5. H. Nagata et al., Optical Fiber Technology, 1995, pp. 283-288.
  6. S. Yuan and N. A. Riza, Applied Optics, vol. 38, 1999, p. 3214.
  7. T. Sakamoto, Applied Optics, vol. 31, 1992, p. 5184.
  8. W.J. Tomlinson, Applied Optics, vol. 19, 1980, p. 1127.
  9. P. Chanclou et al., Journal of Lightwave Technology, vol. 17, 1999, p. 924.
  10. D.V. Plant and A.G. Kirk, Proceedings of IEEE, 2000, p. 806.
  11. S.H. Wang, C.J. Tay et al., Optics and Lasers in Engineering, 34, 2000, p. 121.
  12. A.K. Ghosh, ECTC Proceedings 2000, p. 1493.
  13. N.A. Riza and S. Sumriddetchkajorn, Optics Communications, 185, 2000, pp. 103-108.
  14. N. Murata, ECTC Proceedings 1998, pp. 1178-1185.
  15. S.K. Mondal, R. Zhang and F.G. Shi, unpublished.

H. Zhou, Z. Tang, Y. Lin, W. Liu, S. Mondal and F.G. Shi can be contacted at The Henry Samueli School of Engineering, Optoelectronics Packaging and Automation Laboratory, 916 Engineering Tower, University of California, Irvine, CA 92697-2575; 949-824-5362; Fax: 949-824-2541; E-mail: [email protected].


Easily post a comment below using your Linkedin, Twitter, Google or Facebook account. Comments won't automatically be posted to your social media accounts unless you select to share.