IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 29, NO. 3, AUGUST 2006 463 Fiber Pigtailed Multimode Laser Module Based on Passive Device Alignment on an LTCC Substrate Kimmo Keränen, Jukka-Tapani Mäkinen, Kari T. Kautio, Jyrki Ollila, Jarno Petäjä, Veli Heikkinen, Juhani Heilala, and Pentti Karioja Abstract—A concept that utilizes structured planar substrates based on low-temperature cofired ceramics (LTCC) as a precision platform for a miniature passive alignment multimode laser module is demonstrated. The three-dimensional shape of the laminated and fired ceramic substrate provides the necessary alignment structures including holes, grooves, and cavities for the laser-to-fiber coupling. The achieved passive alignment accuracy allows high coupling efficiency realizations of multimode fiber pigtailed laser modules. Thick-film printing and via punching can be incorporated in order to integrate electronic assemblies directly on the optomechanical platform. The platform is scalable, and it allows embedding of subsystems, such as silicon optical bench (SiOB), but it also provides the interface for larger optical systems. Temperature management of high-power laser diodes is achieved by realizing heat dissipation structures and a cooling channel into the LTCC substrate. The measured maximum laser metallization temperature was 70 C when a thermal power of 0.5 W was applied at the laser active area using a liquid cooling of 50 mL/min. The measured maximum temperature of the laser surface was about three times higher without liquid cooling. Optical coupling efficiency of the multimode laser systems was simulated using optical systems simulation software. The nominal coupling 1 m stripe laser and 62.5/125- m efficiency between 100 graded index fiber 0.275 was 0.37. The simulated coupling efficiency and alignment tolerances were verified by prototype realization and characterization. The measured alignment tolerance values between laser and fiber in AT prototype series 7.7 m, 7.6 m, and 10.8 m (SD were values). The corresponding values in A2 prototype series were 3.1 m, 9.1 m, and 10.2 m. The measured average coupling efficiency was 0.28 in AT series and 0.31 in A2 series. The coupling efficiencies of all operational prototypes varied from 0.05 to 0.43. (NA = 1 = 1 = 1 = 1 = ) 1 = 1 = Index Terms—Hybrid integration, low-temperature cofired ceramics (LTCC), passive alignment, photonic module. I. INTRODUCTION HOTONIC module manufacturers pursue miniature, long-term stable, and precise module realization and assembly technologies in order to achieve cost-effective solutions for market demands. The integration of photonic, electrical, and mechanical functionalities into one system can greatly improve the cost efficiency of systems. Monolithic integration is seen as the most cost-effective method to produce photonic modules due to the fact that packaging cost, which normally represents P Manuscript received October 21, 2004; revised April 22, 2005. K. Keränen, J.-T. Mäkinen, K. T. Kautio, J. Ollila, J. Petäjä, V. Heikkinen, and P. Karioja are with the VTT Electronics, 90570 Oulu, Finland. J. Heilala is with the VTT Industrial Systems, 02044 VTT, Finland. Digital Object Identifier 10.1109/TADVP.2006.872995 the largest part of the photonic module production cost, is minimized [1], [2]. The cost-effective monolithic integration of photonic systems still faces extensive challenges [3]. Hybrid integration of InGaAsP/InP sources and silica fibers using passive alignment silicon waferboard is suggested and demonstrated for producing optical communication modules [4], [5]. Research work has led to the demonstration and utilization of planar lightwave circuits (PLCs) [6]–[8]. However, hybrid integration also seems to offer a very competitive solution for optical communication modules [9]–[11]. In some module realizations, such as high-power optical transmitter modules needed in sensor and tooling applications, hybrid integration seems to be the only realistic method [12]. The use of passive alignment in the production of hybrid integrated modules is especially advantageous in volume production, due to the fact that the concept is simple and fast [13], [14]. The silicon micromachining technologies that are used in the microelectronics industry are widely applied to tool silicon precision substrates for passive alignment fiber optic subassemblies. The accuracy of silicon substrates is adequate for the passive alignment of single mode photonic devices and components [4], [15], [16]. Another possible technology for producing high-precision structures with high aspect ratios and great structural heights with tight tolerances is lithography, electroplating, and molding (LIGA) [17]. The LIGA process has been utilized to produce microoptical benches for the passive alignment of devices [18], [19] and mould inserts for fiber-optic passive alignment substrate replication [20]. However, the high cost of the mask production for the LIGA master limits the application of LIGA to devices that are not cost-sensitive or are produced in a very large volume [21]. Medium- and low-volume production pursues methods with high flexibility and low processing cost. In order to meet the processing cost requirement for mediumand low-volume production of alignment substrates, a concept in which alignment grooves are embossed into metallic substrates has been suggested [21]. We have studied the possibility to create passive alignment substrates for the hybrid integration of high-power laser diodes with microoptical components and silica fibers using low-temperature cofired ceramic (LTCC) technology. Low conductor resistance and dielectric loss, multilayer structures with fine-line capability, compatibility with hermetic sealing and the ability to integrate passive electrical components into the substrate make LTCC a useful technology for optical MEMS and communication applications [21], [24], [25]. The thermal conductivity of LTCC is quite low, about 3 W/m K. However, the thermal conductivity of LTCC can be increased 1521-3323/$20.00 © 2006 IEEE 464 IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 29, NO. 3, AUGUST 2006 Fig. 1. Coordination in laser-to-fiber coupling simulations. locally by processing heat spreaders, thermal vias, and cooling channels on the LTCC substrate. This enables the use of LTCC substrate with a high-power laser diode [26], [27]. The cost efficiency of the LTCC technology-based photonic transmitter module in volume production is advantageous due to the fact that LTCC panel manufacture and component assembly are realized using standard high-volume applicable production methods, such as screen printing, flip-chip assembly of devices, and surface-mount assembly and reflow soldering of discrete components. The cost efficiency of LTCC technology, however, is not limited to high-volume applications, but is also applicable to low- and-medium volume applications, because the process related initial and mask costs are typically moderate compared to the silicon and LIGA processes. Fig. 2. Example of modeling laser-to-multimode fiber coupling utilizing a ball lens. TABLE I SENSITIVITY ANALYSIS OF 1 100 m LASER 62.5/125-m BUTT-COUPLING 2 AND II. LASER-TO-FIBER COUPLING SIMULATIONS In order to get information about achievable coupling efficiency of laser-to-fiber systems and to evaluate the effect of coupling tolerances, coupling simulation and tolerance analysis systems were built using ASAP (Breault Research Organization) optical design and simulation software. The simulation of the passive laser-to-fiber coupling utilizing LTCC test substrates was started by performing an optical coupling efficiency analysis between 100 1 m emitting area laser and 62.5/125- m fiber, 0.275. The laser beam full width divergence was intensity values were applied. 10 56 degrees, when the The emitted power center wavelength was 800 nm. The coordination used in the laser-to-fiber coupling simulations is shown in Fig. 1. The multimode laser model used in the simulations obeyed Gaussian intensity angular distribution in the far-field -direction and top hat in the -direction. The laser model had top hat spatial irradiance distribution in both directions. The simulation model was fitted to the manufacturer’s data. An example of modeling of an edge-emitting stripe laser optical power coupling to a multimode fiber using a ball lens is shown in Fig. 2. The nominal coupling efficiency in butt coupling achieved in the simulation was 0.37, when a nominal distance of 30 m between laser and fiber was used. A ball lens was not finally used in the system demonstration because butt coupling was estimated to be the most cost-efficient method of module implementation in this case. The tolerance analysis of the coupling system was performed in two steps. First, the sensitivity analysis of the system was performed. During the sensitivity analysis, each tolerance variable using a 3 value is simulated separately and the most critical tolerance variables are found, which enables system optimization (see Table I). As we can see from Table I, the fiber decenter along the -axis is the most significant tolerance variable with the used tolerance values according to the simulation. Slight differences in impact values between variable min and max impact values in symmetrical cases are due to simulation noise. Second, optical system Monte-Carlo tolerancing was performed (Fig. 3). Monte-Carlo tolerance simulation is simple and fast method for system performance analysis. We want to get information about performance distribution of large manufacturing lot of modules, and the Monte-Carlo method is very well capable of producing that information. Other tolerance analysis methods, such as worst case analysis seemed to be unsuitable in this case. In Monte-Carlo tolerancing, all tolerance variables are simulated simultaneously, and statistical KERÄNEN et al.: FIBER PIGTAILED MULTIMODE LASER MODULE Fig. 3. Relative coupled power of systems by Monte-Carlo tolerance analysis. information about the system performance is obtained. The variables are represented as distributions in the simulations. Only two variable distributions had non-Gaussian distribution, namely source tilt and fiber tilt . All the other variable distributions had Gaussian distribution. As we can see from Fig. 3, the maximum coupling efficiency peak is around 0.36 and the maximum value is about 0.4. This simulation tool can be used as an assisting tool for setting a totally new optical system performance specification limit. In addition, this tool can also utilize other system optical performance criteria, such as the modulation transfer function (MTF) value. In the case of the fiber pigtailed multimode laser module, the performance criterion is coupling efficiency. III. PROCESSING OF LTCC SUBSTRATES The fabrication of a multilayer ceramic substrate using LTCC technology is shown in Fig. 4. First, glass ceramic tape sheets are blanked to the specified panel size. Second, the sheets are punched in order to form via holes. Via holes are metallized to create electrical interconnects between layers. Cavities and grooves can be processed by via punching. The next step is patterning of electrical conductors and passive circuits onto each layer using screen-printing or photo imaging. The final steps are layer lamination, firing below 980 C temperatures, processing of photo imaged grooves, and circuit dicing. Relatively low sintering temperature allows the use of noble metal conductor materials, such as silver and gold. The assembly of discrete devices onto the substrate finishes the system on package (SOP). The passive alignment of edge emitting laser-to-multimode fiber alignment was studied using LTCC substrates. The purpose of these experiments was mainly to evaluate the passively aligned fiber position accuracy along -axis. Two methods for the manufacture of the fiber grooves were tested, namely punching and photo imaging. The cavities for the fiber grooves were punched to the LTCC tape sheet to the size of 0.15 9 mm, using a 150- m round tool expected to provide a suitable final groove width for the 62.5/125- m fiber. Du Pont 951-AT tape with a green thickness of 114 m was used. This was laminated on top of three blank layers of thicker LTCC tape to obtain enough mechanical strength for the substrate. After lamination, the parts were 465 further cofired, and the groove was diced to 7 mm length. Before lamination, the edges of the punched fiber groove are quite sharp, and the groove reaches its final shape during the isostatic lamination step. The lamination parameters, e.g. pressure and the use of different lamination foils, can be used to adjust the shape of the groove and, consequently, the height of the fiber. The groove reaches its final dimensions during the cofiring process, when substantial shrinkage occurs. The lamination pressure for the experimental fiber grooves was either 1000 or 1500 psi. The shape of the groove was further affected by using different combinations of lamination foils, e.g., Tedlar film (25 m), polyethylene foil (60 m), latex rubber foil (300 m), and steel foil (50 m). With the use of flexible foils, a V-shaped groove can be laminated [see Fig. 5(a)]. Steel, being a rigid material, reduces the lamination force to the cavity edges and, therefore, produces a fairly orthogonal shape for the groove, as seen in Fig. 5(b). To evaluate the fiber passive alignment accuracy to the manufactured grooves, 10/125- m single-mode fiber without jacket was pressed against the groove and attached to the substrate using a UV-curable epoxy. Photoimageable thick-film conductor materials have been previously demonstrated to produce very accurate features and only a few micrometer edge resolution. The processing steps include screen-printing of UV-sensitive paste on a fired substrate, exposure through a photomask, spray development, and finally firing at about 850 C. Similarly, photoimageable glass paste can be used to manufacture alignment structures on a fired substrate. The benefit of using glass instead of conductive paste is that a thicker layer can be exposed and imaged using only a minimal amount of UV energy. The cost of glass material is also lower than that of conductive paste [28]. To manufacture the experimental fiber grooves, either three or four layers of photoimageable glass were screen-printed and dried on alumina or LTCC substrates. The exposure was done on a regular screen exposure unit. The chromium glass mask used had a track width of 70 m and a track spacing of 80 m. The exposed pattern was spray-developed using 0.8% sodium carbonate, followed by water rinsing and spin drying. The patterned glass tracks were fired in a belt furnace, using a standard 850 C thick-film profile. The firing shrinkage typically reduced the track width to 50 m, corresponding to a groove width of 100 m. The fired thickness of the glass track for three and four printed layers was 40 and 48 m, respectively. Pieces of 62.5/125- m multimode fiber were attached to the grooves using UV-curable epoxy. A typical shape of the photoimaged groove pattern and the assembled fibers are shown in Fig. 6. The most important features of the developed module substrate are the realized fiber groove and the liquid cooling channel structure, as depicted in Fig. 7. The fiber groove was punched to the outer tape layer and laminated, as described in Fig. 4. The end of the slightly deformed groove was diced off. The biggest challenge was to develop the manufacturing methods for a buried liquid cooling channel without excessive deformation of the substrate surface, which would obviously deteriorate the passive alignment accuracy and fiber-to-laser coupling efficiency. The channel structure, 1.65 mm wide 466 Fig. 4. IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 29, NO. 3, AUGUST 2006 Standard LTCC process. Fig. 6. Photoimaged fiber grooves on LTCC. Fig. 7. Fig. 5. Punched and laminated fiber grooves on LTCC (a) Flexible foil. (b) Rigid foil. and with a total length of about 30 mm, was punched on two tape layers that were buried two layers deep. To prevent the Schematics of the demonstrator module. sagging of the LTCC stack during the lamination step, the buried channel was filled with an organic precision-machined filler insert, which burns off efficiently during the cofiring step. The final dimensions of the channel cross section were 1.4 0.4 mm. The planarity of the module surface on top of the buried channel was measured on eight samples, showing a m, and 7 m in the worst case. The typical warpage of expected warpage can be noticed in the system design in order KERÄNEN et al.: FIBER PIGTAILED MULTIMODE LASER MODULE 467 Fig. 9. Fig. 8. Cooling channel cross section, 1.4 Demonstrator module. 2 0.4 mm. to decrease the offset along the -axis in the laser-to-fiber coupling. In the module packaging process, the warping tolerance effect is added to the laser diode height tolerance chain. Obviously, to introduce minimum amount of warping, the thickness of the organic insert must be controlled very accurately. To enhance the thermal flow to the cooling channel, thermal vias and heat spreader layers were processed. A photo of the cooling channel cross section is shown in Fig. 8. IV. MODULE PACKAGING A fiber pigtailed laser module series consisting of 40 modules in two different series was realized using a six-layer LTCC substrate, with a size of 22 25 mm, to evaluate the multimode laser-to-fiber passive alignment accuracy and the thermal characteristics of LTCC for the packaging of high-power lasers. The demonstrator assembly started with laser chip alignment and attachment to the LTCC substrate gold metallization. The 200 m W laser chip dimensions were 117 m H 1000 m L . The laser diode was aligned with a flip-chip bonder and attached to the LTCC substrate by a 50%In/50%Pb solder preform. The solder preform size was 200 200 m and a thickness of 20 m was used for the attachment. The upper contact was wire-bonded by a wedge-bonder using gold wire with a 25- m diameter. To measure the effectiveness of the liquid cooling, a thermistor (Shibaura Electronics PT7-312) was attached on the LTCC substrate in the vicinity of the laser diode, using thermally conductive silicone (Dow Corning Q-9226). The thermistor contact wires were microwelded to the LTCC substrate metallization. 0.275 fiber A 62.5/125- m multimode graded-index (Spectran) was aligned manually to the LTCC groove under microscope. The separation between the laser and the fiber was adjusted to about 40 m. The fiber was pressed to the groove using a small weight at the fiber center and epoxy bonded using three separate drops of Loctite 3525 UV-epoxy. The strain relief was realized by bonding the fiber buffer to the widened groove using the same epoxy. Fig. 10. Closeup of laser chip and fiber. The water inlet and outlet tubes, made of nickel-coated copper, were solder attached to the substrate using eutectic SnPb-solder. Silicone pipes were attached to the inlet and outlet tubes in order to enable water injection to the cooling channel. The realized demonstrator module is shown in Fig. 9 and a close up of the laser-to-fiber coupling in Fig. 10. V. MODULE CHARACTERIZATION The accuracy of laser diode-to-fiber passive alignment on the LTCC substrate was characterized by measuring alignment errors of assembled devices from the realized modules. A Veeco white light interferometer, the Wyko NT-3300 model, and an optical three-dimensional (3-D) coordinate measurement system, the Smartscope 200 model, were used in the measurements. Table II shows the measured alignment tolerances. LTCC substrate surface was used as a reference surface when measuring laser diode and fiber absolute assembly accuracy in the -axis direction, which was seen as the most critical tolerance in the sensivity analysis. The fiber -axis average offset was 17.1 m in the AT series and 10.4 m in the A2 series. The laser diode -axis average offset was 0.3 m in AT series and 2.2 m in A2 series. We could not, however, measure and directions due to the absolute assembly accuracy in lack of proper reference surfaces/marks. For the same reason, we could not define offset of components along the -axis. The 468 IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 29, NO. 3, AUGUST 2006 TABLE II MEASURED LASER-TO-FIBER ALIGNMENT TOLERANCES TABLE III AVERAGE COUPLING EFFICIENCYS AND DISTRIBUTIONS assembly accuracy in the - and Z-axis directions was characterized by measuring the relative alignment error between laser and fiber. The difference between series was that AT series precision structure tape thickness after sintering was 90 m and A2 series tape thickness was 130 m. As one can see from Table II, the transverse alignment errors in both series are under 9.1 m, which suggests good coupling efficiency between laser diode and fiber. The coupling efficiency of the modules was measured so that the fiber-coupled power was first measured after module assembly by an optical power meter applying 10-mm square silicon detector. Second, the substrate was cut near the laser diode, and the total emitted power produced by the laser diode was measured using the same meter and detector. The coupling efficiency was achieved by simply dividing fiber-coupled power by total power. The average coupling efficiency and distribution are shown in Table III. The total number of operational modules was 13 in the AT series and 16 in A2 series. The total number of manufactured modules was 40. A prestudy found poor laser output with 11 modules, and fibers were not assembled on those modules. As one can see from Table III, the average coupling efficiency in the A2 series is better than in the AT series. This is very consistent with the fact that the measured alignment errors are smaller in the A2 series than in the AT series. The distribution of coupling efficiency in both series is wider than expected based on the simulations. Laser near and far-field intensity distributions were measured in order to see the accuracy of the laser model used. Near-field intensity distribution was measured by imaging laser facet to a CCD camera detector using a microscope objective. In Fig. 11, an example of the measured spatial intensity distribution is shown. We can see from Fig. 11 that the laser near-field irradiance distribution is clearly a multimode and not a top hat as used in the initial simulations. There is an intensity drop in the middle of the laser facet facing the fiber core in butt coupling at nominal alignment. This causes a clear reduction in the coupling efficiency. In the opposite case with very bright laser center area, the coupling efficiency would be higher than nominally possible with a top hat spatial irradiance laser model. Therefore, a more accurate laser model is needed in order to achieve truly predictive simulations. Fig. 11. Example of a measured laser spatial near-field intensity. Fig. 12. Effect of laser near-field spatial characteristics extremes on coupling efficiency. Fig. 13. Temperature near the laser diode measured by a thermistor. KERÄNEN et al.: FIBER PIGTAILED MULTIMODE LASER MODULE Fig. 14. 469 (left) Simulated and (right) measured laser surface temperatures. (Color version available online at http://ieeexplore.ieee.org.) After the laser modules were characterized, a new set of emulative simulations was performed. The effect of laser characteristics with realized tolerances is seen in Fig. 12. Two laser models were created according to two characterized lasers. Two distributions were simulated with the characterized tolerances. In these simulations the assembly tolerance value distributions used were identical. The distribution showing lower coupling efficiency values was obtained with a laser model with a central intensity drop in the near-field and 21 28 deg divergence values. This model represents an example of a low-performance laser like that seen in Fig. 11. The second distribution showing higher coupling efficiency values was obtained by using a laser model with 7 28 deg divergence and a top hat near-field intensity distribution. This represents an example of a high-performance laser, such as our nominal laser used in the initial simulations. The characterized coupling efficiencies of series A2 are also shown in Fig. 12. Fig. 12 shows that the performance distribution of manufactured modules was clearly wider than the initial simulations shown in Fig. 2 suggested. There are two narrow distributions covering almost the entire range of measured coupling efficiency values. This means that the effect of assembly tolerances is much smaller than the effect of variation on laser diode characteristics. The high-performance laser shows slightly better performance than the initial laser, although its divergence value is larger. The performance discrepancy is caused by different tolerance distributions and shape of intensity distributions. The sensitivity to assembly tolerances is larger with the high-performance laser than with the low-performance laser. This can be seen from the wider distribution obtained with the high-performance laser. Cooling system efficiency was tested by running a highpower laser diode (50 mL/min) with and without water-cooling and by measuring surface temperature near the laser diode using a thermistor (Fig. 13). TABLE IV SIMULATED AND MEASURED MAXIMUM LASER SURFACE TEMPERATURES AT LTCC SUBSTRATE However, the thermistor measurement does not measure the laser surface temperature. In order to get a better view of laser surface temperature, a thermal camera model SC3000 manufactured by FLIR Systems was used. The ambient temperature was 24 C, cooling water temperature 22.5 C, and flow rate 50 mL/min during the measurements. Simulated (Flotherm) and measured temperature distributions are shown in Fig. 14. As one can see from Fig. 14, a fairly good correspondence between simulated and measured temperature values was achieved. The temperature maximum value in the simulation was achieved at the laser active layer near the front mirror surface. The simulated and measured laser upper surface maximum temperature values are shown in Table IV. The water-cooling decreased the laser surface temperature by a factor of 3.4 compared to the noncooled case. The factor was 2.8 when compared to a simulated system with the same thermal management structures except for the cooling channel, the volume of which was replaced with ceramic LTCC material in the simulation. The laser chip temperature can be further decreased using a higher flow rate and/or cooler water. The cooling structure optimization based on the experiments and simulations is also possible. 470 IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 29, NO. 3, AUGUST 2006 VI. DISCUSSION The possibility to evaluate module concepts and optimize detailed structures and tolerances through simulations is a cost-effective approach to photonic module integration. The realization and characterization of the prototype devices and modules verifies module design and offers information to enhance design and simulation models. Passive alignment is the most cost-effective method to produce photonic modules in mass production. The costeffectiveness of the photonic module is further improved by fully utilizing the 3-D integration possibilities of an LTCC substrate. 3-D integration enables high circuit density and versatile technology designs, including RF, analog, digital, and optical either independently or in combination. It is also possible to integrate high-speed laser driving electronics onto the LTCC substrate [29]. The typically defined performance limit attenuation value for optical coupling is 1 dB compared to the maximum coupling value. The absolute accuracy of the alignment structures combined with the realized tolerances should provide coupling efficiency within this limit. Achieved accuracy can be judged as deviation of the measured average value compared to the design value and tolerance value defines the distribution. Typical multimode transverse alignment tolerances allowed for m. Typical 1-dB coupling attenuation are roughly single-mode transverse alignment tolerances allowed for 1-dB 1 m attenuation for laser-to-fiber couplings are roughly 0.5 and for couplings between single-mode fibers 1 3 m. We achieved 3 10 m transverse alignment tolerances in the laser-to-fiber couplings in the modules. We can say that achieved alignment tolerances are adequate for multimode couplings but inadequate for single-mode couplings. In order to meet transverse alignment tolerance requirements for single-mode couplings, roughly five to tenfold improvement in the laser-to-fiber coupling tolerances has to be achieved. At this moment, it seems that the use of new photoimageable materials offers the greatest opportunity for improvement to the alignment structure accuracy. Another benefit would be that the passive alignment structures or alignment fiducials for the photonic devices could be manufactured in the same process. These advantages will be utilized when high-precision alignment structures for single-mode applications are pursued. temperature of the laser surface decreased by a factor 3.4 using water-cooling compared to the noncooled case. Manufactured module series coupling efficiency was prestudied through simulations performing tolerance analysis in two steps. Sensitivity analysis indicated that the most significant variable in optical coupling was fiber decenter along the -axis. Monte-Carlo tolerancing revealed the expected performance distribution with the presumable manufacturing tolerances and showing quite narrow performance distribution. The measured alignment tolerance values between laser 7.7 m, and fiber in the AT prototype series were 7.6 m, and 10.8 m (SD values). The corre3.1 m, sponding values in the A2 prototype series were 9.1 m, and 10.2 m. The measured average coupling efficiency was 0.28 in the AT series and 0.31 in the A2 series. The coupling efficiency of all operational prototypes varied from 0.05 to 0.43. The characterization showed that the achieved alignment tolerances are adequate for high-efficiency coupling between the used multimode laser and that fiber and narrow performance distribution should result. The performance distribution of modules, however, was clearly wider than the simulations suggested, although the amount of modules was quite small to define the performance distribution accurately. The measured near-field and far-field intensity distributions of lasers suggested that the large variations between individual lasers actually cause the widening of module performance distribution. A more accurate laser model is needed in order to improve the accuracy of predictive performance simulations. This can be achieved by performing device characterizations and emulative simulations. The most important factor in manufacturing process improvement of the fiber pigtailed multimode laser modules is narrowing the performance distribution of the lasers. A highquality manufacturing process produces a narrow performance distribution. Obviously, the high-performance laser module product should also have a high coupling efficiency. In this paper, we have described a method for simultaneous evaluation of both process and product performance. ACKNOWLEDGMENT The authors would like to thank R. Lehtiniemi from Nokia Research Center (NRC) for performing laser surface temperature measurements with a SC3000 thermal camera. VII. CONCLUSION A concept for producing a precision platform for a passive alignment multimode laser module was demonstrated. The platform was a structured planar substrate based on LTCC technology. The 3-D structure of the substrate was achieved by traditional via punching in order to create the necessary alignment structures including holes, grooves, and cavities. Alignment structures can also be created using new photoimageable materials. In this paper, a passive alignment fiber pigtailed laser module on LTCC substrate was designed, manufactured, and characterized. The characterization showed that the maximum laser chip surface temperature was 70 C, when 0.5-W thermal power was applied at the active area of the laser. The maximum REFERENCES [1] T. Koch and U. Koren, “Semiconductor photonic integrated circuits,” IEEE J. Quantum Electron., vol. 27, no. 3, pp. 641–653, Mar. 1991. [2] R. Soref, “Silicon based optoelectronics,” Proc. IEEE, vol. 81, no. 12, pp. 1687–1706, Dec. 1993. [3] R. Kaiser, M. Hamacher, H. Heidrich, P. Albrecht, W. Ebert, R. Gibis, H. Kunzel, R. Löffler, S. Malchow, M. Möhrle, W. Rehbein, and H. Schroeter-Jansen, “Monolithically integrated transceivers on InP: The development nad generic integration concept and its technological challenges,” in Proc. 10th Int. Conf. Indium Phosphide Related Mater., May 1998, pp. 431–434. [4] C. Armiento, M. Tabasky, C. Jagannath, T. Fitzgerald, C. Shieh, V. Barry, M. Rothmann, A. Negri, P. Haugsjaa, and H. Lockwood, “Passive coupling of InGaAsP/InP laser array and singlemode fibers using silicon waferrboard,” Electron. Lett., vol. 27, pp. 1109–1110, Jun. 1991. KERÄNEN et al.: FIBER PIGTAILED MULTIMODE LASER MODULE [5] B. Corbett, K. Rodgers, F. Stam, D. O’Connell, P. Kelly, and G. Crean, “Low-stress hybridization of emitters, detectors and driver circuitry on a silicon motherboard for optoelectronic interconnect architecture,” Mat. Sci. Semicond. Process., pp. 449–453, 2000. [6] S. Mino, K. Yoshino, Y. Yamada, T. Terui, M. Yasu, and K. Moriwaki, “Planar lightwave circuit platform with coplanar waveguide for optoelectronic hybrid-integration,” J. Lightw. Technol., vol. 13, no. 12, pp. 2320–2326, Dec. 1995. [7] T. Hashimoto, Y. Nakasuga, Y. Yamada, H. Terui, M. Yanagisawa, K. Moriwaki, Y. Susaki, Y. Tohmori, Y. Sakai, and H. Okamoto, “Hybridintegration of spot-size converted laser diode on planar lightwave circuit platform by passive alignment technique,” IEEE Photon. Technol. Lett., vol. 8, no. 11, pp. 1504–1506, Nov. 1996. [8] D. Vernooy, J. Paslaski, H. Blauvelt, R. Lee, and K. Vahala, “Alignment-insensitive coupling for PLC-based surface mount photonics,” IEEE Photon. Technol. Lett., vol. 16, no. 1, pp. 269–271, Jan. 2004. [9] K. Kurata, “Mass production techniques for optical modules,” in Proc. ECTC Conf., May 1998, pp. 572–580. [10] H. Althaus, W. Gramann, and K. Panzer, “Microsystems and wafer processes for volume production of highly reliable fiber-optic components for telecom- and datacom-application,” IEEE Trans. Compon., Packag., Manuf. Technol. B, vol. 21, no. 2, pp. 147–156, May 1998. [11] E. Higurashi, R. Sawada, and T. Ito, “An integrated laser blood flowmeter,” J. Lightw. Technol., vol. 21, no. 3, pp. 591–595, Mar. 2003. [12] W. Pittroff, G. Erbert, G. Beister, F. Bugge, A. Klein, A. Knauer, J. Maege, P. Ressel, J. Sebastian, R. Staske, and G. Traenkle, “Mounting of high power laser diodes on boron nitride heat sinks using optimized Au/Sn metallurgy,” in Proc. ECTC Conf., May 2000, pp. 119–124. [13] W. Hunziker, W. Vogt, H. Melchior, P. Buchmann, and P. Vettiger, “Passive self-aligned low-cost packaging of semiconductor laser arrays on Si motherboard,” IEEE Photon. Technol. Lett., vol. 7, no. 11, pp. 1324–1326, Nov. 1995. [14] J. Goodrich, “A silicon optical bench for approach to low-cost high speed tranceivers,” in Proc. ECTC Conf., May–Jun. 2001, pp. 238–241. [15] Y. Sunaga, R. Takahashi, T. Tokoro, and M. Kobayashi, “2-Gbit/s small form factor fiber-optic transceiver for single mode optical fiber,” IEEE Trans. Adv. Packag., vol. 23, no. 2, pp. 176–181, May 2000. [16] M. Uekawa, H. Sasaki, D. Shimura, K. Kotani, Y. Maeno, and T. Takamori, “Surface-mountable silicon microlens for low-cost laser modules,” IEEE Photon. Technol. Lett., vol. 15, no. 7, pp. 945–947, Jul. 2003. [17] C. Malek and V. Saile, “Applications of LIGA technology to precision manufacturing of high-aspect-ratio micro-components and systems: a review,” Microelectron. J., vol. 24, pp. 131–143, 2004. [18] A. Muller, J. Göttert, J. Mohr, and A. Rogner, “Fabrication of stepped microoptical benches for fiber and free space applications,” Microsyst. Tech., vol. 2, pp. 40–45, 1996. [19] H. Schift and J. Söchtig, “LIGA-technology for the fabrication of positioned planar surfaces,” Microsyst. Technol., vol. 4, pp. 132–134, 1998. [20] P. Karioja, J. Ollila, V.-P. Putila, K. Keränen, J. Häkkilä, and H. Kopola, “Comparison of active and passive fiber alignment techniques in laser pigtailing,” in Proc. ECTC Conf., May 2000, pp. 244–249. [21] R. Keyssean, J. Sosnowski, M. Doyle, D. Amey, and S. Horowitz, “A new approach to opto-electronic/MEMS packaging,” in Proc. ECTC Conf., May 2002, pp. 259–262. [22] U. Schmid, “A robust flow sensor for high-pressure automotive applications,” Sens. Actuators A, vol. 97–98, pp. 253–263, 2002. [23] X. Gong, W. Chappell, and L. Katehi, “Multifunctional substrates for high-frequency applications,” IEEE Microw. Wireless Compon. Lett., vol. 13, no. 10, pp. 428–430, Oct. 2003. [24] A. Sutono, A.-V. Pham, J. Laskar, and W. Smith, “RF/microwave characterization of multilayer ceramic-based MCM technology,” IEEE Trans. Adv. Packag., vol. 22, no. 4, pp. 326–331, Aug. 1999. [25] C. Zhang, D. Liu, S. Mathews, J. Graves, T. Schaefer, B. Gilbert, R. Modi, H.-D. Wu, and D. Chrisey, “Laser direct-write and its application in low-temperature cofired ceramic (LTCC) technology,” Microelectron. Eng., vol. 70, pp. 41–49, 2003. [26] M. Gongora-Rubio, P. Espinoza-Vallejos, L. Sola-Laguna, and J. Santiago-Aviles, “Overview of low temperature cofired ceramics tape technology for meso-system technology (MsST),” Sens. Actuators A, vol. 89, pp. 222–241, 2001. [27] V. Chiriac and T.-J. Lee, “Thermal assessment of RF integrated LTCC front end module (FEM),” in Proc. Int. Soc. Conf. Thermal Phen., 2002, pp. 520–527. 471 [28] K. Kautio, “Properties of high definition photoimaged conductors in LTCC,” in Proc. IMAPS Nordic Conf., Sep.–Oct. 2002, pp. 227–232. [29] M. Karppinen, J.-T. Mäkinen, K. Kataja, A. Tanskanen, T. Alajoki, P. Karioja, M. Immonen, and J. Kivilahti, “Embedded optical interconnect on printed wiring board,” Proc. SPIE, vol. 5453, pp. 150–164, Apr. 2004. Kimmo Keränen was born in Hyrynsalmi, Finland, in 1966. He received the M.Sc. (Tech.) and Lic.Sc. (Tech.) degrees in electrical engineering from the University of Oulu, Oulu, Finland, in 1992 and 2002, respectively. He is a Senior Research Scientist with the MicroModules Group, VTT Electronics, Oulu. His research activities include microoptics and module integration of optoelectronic devices. Mr. Keränen is a member of Finnish Optical Society and European Optical Society. Jukka-Tapani Mäkinen received the M.Sc. degree in physics from the University of Oulu, Oulu, Finland, in 1998. He is currently working toward his Ph.D. degree within the optoelectronics research area of VTT Electronics, Oulu. His research interests are optoelectronic systems design, simulation, and prototyping. Kari T. Kautio received the M.Sc. degree in electrical engineering from the University of Oulu, Oulu, Finland in 1983. In 1989, he joined VTT Electronics, Oulu, where he is a Senior Research Scientist. From 1987 to 1989, he was a Thick-Film Process Engineer at Aspo Microelectronics. From 1983 to 1987, he was a Research Scientist at the University of Oulu Microelectronics Laboratory, working on thick-film hybrid applications. His research interests are LTCC processing technology, ceramic-based module packaging, and optoelectronic packaging. Mr. Kautio is a member of IMAPS. Jyrki Ollila received the M.Sc. (Tech.) degree in process engineering from the University of Oulu, Oulu, Finland, in 1992. In 1993, he joined VTT Electronics, Oulu, Finland, where he is a Research Scientist. His research interests are 3-D design of optoelectronic devices, hermetic sealing, glass-to-metal seals, advanced electronics, and optoelectronics package manufacturing by wire-bonding, die-bonding, and flip-chip technologies. Jarno Petäjä was born in Oulu, Finland, in 1976. He received the M.Sc. degree in physics from the University of Oulu, Oulu, Finland, in 2003. He is a Research Scientist with the MicroModules Group, VTT Electronics, Oulu. His latest research activities have been in the fields of thermal modeling, accurate dimensional measurements, and printable electronics. 472 IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 29, NO. 3, AUGUST 2006 Veli Heikkinen (M’97) was born in Hyrynsalmi, Finland, in 1960. He received the M.Sc. (Tech.), Lic.Sc. (Tech.), and Dr.Sc. (Tech.) degrees in electrical engineering from the University of Oulu, Oulu, Finland, in 1986, 1999, and 2004, respectively. He is a Senior Research Scientist with the MicroModules Group, VTT Electronics, Oulu. His professional interests lie in the research and development of packaging technologies for optoelectronic devices. Dr. Heikkinen is a member of the European Optical Society, Finnish Optical Society, and SPIE. Juhani Heilala is a Senior Research Scientist with VTT Industrial Systems, the Technical Research Centre of Finland, Espoo, with 20 years of experience in robotics, modular final assembly system development, system simulation, and virtual environments. Research topics includes micromechanical precision assembly, assembly process analysis, design for assembly, and manufacturability. Pentti Karioja received the MSc., Licentiate in Technology, and Dr. Tech degrees in electrical engineering from the University of Oulu, Oulu, Finland, in 1981, 1993, and 1997, respectively, and the . He became a Research Scientist with VTT Electronics, Oulu, Finland, in 1985. Initially, he was a Scientist on applied optoelectronics research; and since 1986, he has worked as a Project Manager and Program Manager. In 1993, he joined the Optical Sciences Center, University of Arizona, Tucson, working as a Visiting Scholar for a year. In 1996, he spent five months in the Optical Sciences Center finishing his thesis. Since 1998, he has worked as Group Manager, and since 2000, Chief Research Scientist at VTT Electronics, Optoelectronics Research Area.