Optical interconnections that utilize inherent fea-tures of free-space light propagation are promisingtechnologies for providing massively parallel andhigh-speed computing.1,2 For developing such anoptical computing system consisting of many opticaland optoelectronic ~OE! components, such as lenses,holograms, laser diode ~LD! arrays, and photodetec-tor ~PD! arrays, a feasible packaging technique isrequired for assembling the components efficiently.
We have proposed reflective block optics ~REBOP!as an approach to acquiring rigid and compact pack-aging of three-dimensional ~3-D! optical systems.3An optical system based on REBOP is constructed bythe assembly of optical block units consisting of glassblocks together with cube beam splitters, reflectivelenses, and other optical components, as shown inFig. 1. These optical components are adhered toeach other by optical cement. Lenses and deflectingprisms can be implemented as reflective optical com-ponents and affixed to the optical block. The reflec-tive lenses are plano–convex lenses with a reflectivecoating. Thus, in the REBOP system, the lightbeam propagates in only the transparent solid media.
The authors are with the Department of Electrical Engineering,Faculty of Engineering, Osaka City University, 3-3-138, Sugimoto,Sumiyoshi-ku, Osaka, 556, Japan.
Received 7 March 1997; revised manuscript received 30 June1997.
228 APPLIED OPTICS y Vol. 37, No. 2 y 10 January 1998
Because an optical system based on REBOP com-bines discrete optical components, it requires an ac-curate and effective alignment technique. When theoptical block units are set in contact with each other,the components are located in the proper position inthe longitudinal direction parallel to the optical axisand at the proper angle to the optical axis. How-ever, an additional alignment mechanism should beused for alignment in directions that are perpendic-ular ~lateral directions! to the optical axis.
The alignment of components is a common problemfor free-space optical interconnections. A free-spaceoptical interconnection that uses a macro-optical sys-tem often requires higher alignment accuracy in lat-eral directions than in longitudinal directions. Analignment accuracy of a few micrometers or submi-crometers is required when high-density OE devices,in which each element is ;10 mm in size, are used.By using 3-D integrated micro-optics such as planaroptics,4 we can avoid the alignment problem. How-ever, when several optical modules fabricated withthe 3-D optical integration technique are combined toconstruct a complicated optical computing system,the alignment of the optical modules becomes a sig-nificant problem.
To mount OE devices accurately, we can adjust theposition of the devices with a micropositioner by mon-itoring the light passing through the optical system.5However, this procedure needs complicated equip-ment and takes a long time to complete the align-ment. Several researchers have proposed passivealignment techniques for assembling optical compo-nents by using guide grooves6 or solder-bump bond-
ing.7 However fabrication errors in the sizes ofoptical blocks and lenses cause the shift of the opticalaxis, even though every component is aligned withthe guide-groove and the solder-bump methods. Inaddition, angle errors of reflective surfaces shift theoutput image in lateral directions. Thus the align-ment accuracy achievable between the input and theoutput planes with these passive alignment tech-niques is worse than the fabrication accuracy of op-tical blocks, which is from 50 to 100 mm.
In this paper we propose a novel alignment tech-nique based on optical microconnectors for 3-D opti-cal systems. The use of optical microconnectors wasoriginally proposed as a method of connecting an op-tical fiber to a microlens.8 The self-alignment tech-nique described in Section 2 has features of both thepassive method and the monitoring method, which isuseful for accurate alignment between input and out-put planes in an imaging system. The reliability ofthe optical microconnector is verified from prelimi-nary experiments with a simple imaging system inSection 3, and problems are discussed in Section 4.
Fig. 1. Examples of REBOP: ~a! 4f system, ~b! combination ofoptical blocks in a complicated optical computing system.
2. Self-Alignment with Optical Microconnectors
The self-alignment technique with optical microcon-nectors can achieve accurate alignment between anoptical fiber and a microlens without monitoring ofthe traveling light and without micropositioningequipment. A bump called an optical plug, which ismade of a photosensitive resin exposed to light fo-cused by the microlens, is formed on a planar micro-lens. The optical plug is fit into a hollow concave cutlocated at the end of an optical fiber. Thus the op-tical fiber is automatically guided to the focal point ofthe microlens by the optical plug.
A similar technique can be applied to the align-ment between input and output planes in 3-D imag-ing systems. Figure 2 illustrates the self-alignmentprocedure in an imaging system based on REBOP.We consider an optical interconnection system inwhich the light from an OE device containing a two-dimensional ~2-D! LD array ~or a light modulatorarray! is transferred to an OE device containing a 2-DPD array through the imaging system. Optical com-ponents such as reflective lenses, cube polarizingbeam splitters, and quarter-wave plates are assem-bled by use of a passive alignment method for con-struction of the optical system. Owing tomisalignment of the components and the OE devicesin longitudinal directions, which are parallel to anoptical axis, defocus and magnification errors are in-troduced. These errors decrease as the f-numberand the focal length increase. In this optical system,it is assumed that the location of every componentand device is satisfactorily accurate in the longitudi-nal direction, so that defocus and magnification er-rors of the output image are negligible. Thus wemust adjust the position of the input and the outputOE devices in only lateral directions for alignment.
Initially the output plane is coated with a photo-sensitive resin. Light sources are located on the in-put plane to expose the photosensitive resin. Thephotosensitive resin is exposed to focused light pass-ing through the optical system. Unexposed resin isremoved through development, and the exposedparts form optical plugs. For mounting an OE de-vice chip on the output plane, the optical plugs on theoutput plane fit into sockets, which are hollow open-ings on the chip surface. The sockets are fabricatedin advance by etching of the substrate of the OE
Fig. 2. Self-alignment by optical plugs. The optical plugs are made of a photosensitive resin exposed to light passing through the opticalsystem.
10 January 1998 y Vol. 37, No. 2 y APPLIED OPTICS 229
device or by a similar procedure such as plug forma-tion, in which the substrate around the active area ofthe OE device is coated with the photosensitive resinand a negative photomask is used for exposure.
Even though the optical image on the output planeis slightly shifted by fabrication errors of the opticalcomponents, the position of the optical plugs corre-sponds exactly to the position of the light sourceslocated on the input plane. Thus we can use theoptical plugs as reference marks to align the devicesmounted on the input and the output planes. Inaddition, an output image may rotate because of an-gle errors of reflective surfaces. Misalignmentcaused by this rotation can be corrected by opticalplugs, because the arrangement of the optical plugs isalso rotated according to the rotation of the outputimage and an OE device chip is guided to the positionof the output image.
As the alignment method is based on the actualimaging configuration in which location errors of opti-cal components and angle errors of the reflective sur-faces are involved, the accuracy of alignment must behigher than the conventional passive alignment meth-ods. In addition, each optical system is not equippedwith monitoring and micropositioning devices. Thusthe proposed method has higher mass productivitythan the conventional active alignment methods.
3. Experimental Results
We formed optical plugs in a simple reflective imag-ing system, as shown in Fig. 3. A square plano–convex lens with an aluminum-coated sphericalsurface was used as a reflective lens. The lens wasa 20 mm 3 20 mm square with a focal length of 30mm. The input plane and the output plane eachequally shared the surface of the optical block oppo-site the reflective lens. In Fig. 3 the upper half of theleft side plane is the input plane, and the lower halfis the output plane. A patterned metal mask wasplaced on the input plane and was illuminated by anAr laser with a wavelength of 458 nm. A glass platespin coated with a photosensitive polyimide ~Pho-toneece, Toray Co. Ltd.! was placed on the outputplane. The polyimide used has good mechanicalproperties ~tensile strength, 20 kgymm2! and adheresstrongly to glass. The resin was exposed to the laserlight, forming an optical image of the patterned metal
Fig. 3. Experimental setup for forming optical plugs.
230 APPLIED OPTICS y Vol. 37, No. 2 y 10 January 1998
mask placed on the input plane. After the develop-ment procedure, exposed parts of the resin formedbumps or optical plugs.
We formed two types of optical plugs: a circulartype and a long strip type. The diameter of thecircular-type plug was 100 mm, and the width of thestrip type was 100 mm. An example of the strip-type optical plug is shown in Fig. 4. The height ofthe optical plug was ;25 mm, as shown in Fig. 4~b!.The optical energy density for the exposure was
;1.2 Jycm2. We used a glass plate coated with thephotosensitive polyimide to substitute for an OEdevice. Hollow openings or sockets were formed onthe glass plate when the photosensitive polyimidewas exposed to light. The depth of the sockets was;20 mm. The strip-type optical plug was set intothe socket, as shown in Fig. 5. The exactness ofshapes between the optical plugs and the sockets isthe primary factor that determines alignment ac-curacy. In our experiment, the difference in widthbetween the optical plugs and the sockets was10–20 mm. Thus alignment accuracy was esti-mated to be ;20 mm.
This alignment method with an accuracy of 20 mmattained in the experiment is available for a free-space optical interconnection system when PD’s arelarger than 20 mm 3 20 mm in area. Even thoughthe PD’s are smaller, microlenses can be used in frontof the PD’s to increase the alignment tolerance.9
4. Discussion
As described in Section 3, an alignment accuracy of20 mm was attained in our experiment. However,we believe that this accuracy is not a limit of thismethod. For achieving higher alignment accuracy,the shape of the optical plugs must correspond ex-actly to that of the sockets. As the size and theshape of the optical plugs depend on the thickness of
Fig. 5. Optical plug and socket.
the resin, exposure, and development time, we shouldoptimize and stabilize these conditions.
The shape of the optical microconnectors also af-fected the ease of connecting. In the experiments, thelong strip-type optical plug could be connected moreeasily than the small circular type. For easy connec-tion, a thick resin layer is preferable because tall plugs,which fit into sockets easily, can be formed. However,for the precise patterning of a photosensitive resin, theresin layer should be as thin as possible. Thus thereis a trade-off between accuracy and ease of connection.It is necessary to investigate further the shape and thesize of the optical microconnectors for easy and preciseconnection of optical plugs.
The accuracy of forming the optical plugs is essen-tially restricted by the resolution of the photosensi-tive resin and the optical system. The resolution ofthe photosensitive resin is of the order of submi-crometers. The optical system used in the experi-ment has a resolving power of ;120 line pairsymm;then the minimum controllable size of the opticalplugs is ;8 mm. For the formation of more preciseplugs, the optical system should be improved to havehigher resolving power.
Another problem is wavelength mismatching of theinformation-transmitting light and optical-plug-forming light. The wavelength region of the photo-sensitive resin used in the experiment was less than500 nm, which is shorter than the usual wavelengthsof LD’s used in typical optical interconnections.Some optical passive devices, such as polarizing beamsplitters and diffractive optical elements, have char-acteristics sensitive to changing wavelengths. Ifsuch optical devices are used in an optical system andoptimized to match the LD wavelength, the light forforming optical plugs is not transmitted properly.The development of a new resin that is sensitive tolonger wavelengths or light sources with shorterwavelengths is required.
5. Conclusions
To achieve easy and accurate construction of a 3-Doptical system, we have proposed the self-alignmentmethod with optical plugs. Optical plugs and sock-ets were fabricated experimentally, and an alignmentaccuracy of approximately 20 mm was attained. Theoptical microconnector method is promising for accu-rate alignment of the order of micrometers withoutprecise positioning and monitoring equipment.
The authors thank Y. Ichioka and J. Tanida ofOsaka University for useful discussions. This studywas supported by Grant-in-Aid for Encouragement ofYoung Scientists from the Ministry of Education, Sci-ence, Sports, and Culture.
References1. D. Miyazaki, J. Tanida, and Y. Ichioka, “Optical implementa-
tion of the banyan network using a Sagnac inverter with apatterned mirror,” Opt. Commun. 93, 283–288 ~1992!.
2. Y. Awatsuji, D. Miyazaki, J. Tanida, and Y. Ichioka, “Overalloperations of the hybrid optical parallel array logic system ~H-OPALS162!,” Opt. Rev. 1, 163–165 ~1994!.
10 January 1998 y Vol. 37, No. 2 y APPLIED OPTICS 231
3. D. Miyazaki, J. Tanida, and Y. Ichioka, “Reflective block opticsfor packaging of optical computing systems,” Opt. Lett. 19,1281–1283 ~1994!.
4. J. Jahns and A. Huang, “Planar integration of free-space opticalcomponents,” Appl. Opt. 28, 1602–1605 ~1989!.
5. G. C. Boisset, B. Robertson, and H. S. Hinton, “An approach toactive alignment of free-space optical interconnects,” in OpticalComputing, B. S. Wherrett and P. Chavel, eds., Vol. 139 of theInstitute of Physics Conference Series ~Institute of Physics,London, 1994!, pp. 227–230.
6. F. B. McCormick, F. A. P. Tooley, J. L. Brubaker, J. M. Sasian,T. J. Cloonan, A. L. Lentin, R. L. Morrison, R. J. Crisci, S. L.
232 APPLIED OPTICS y Vol. 37, No. 2 y 10 January 1998
Walker, S. J. Hinterlong, and M. J. Herron, “Design and toler-ance comparisons for S-SEED-based free-space switching fab-rics,” Opt. Eng. 31, 2697–2711 ~1992!.
7. M. J. Goodwin, A. J. Moseley, M. Q. Kearley, R. C. Morris, C. J. G.Kirkby, J. Thompson, R. C. Goodfellow, and I. Bennion, “Opto-electronic component arrays for optical interconnection of circuitsand subsystems,” J. Lightwave Technol. 9, 1639–1645 ~1991!.
8. A. Sasaki, T. Baba, and K. Iga, “Put-in microconnectors foralignment-free coupling of optical fiber arrays,” IEEE Photon.Technol. Lett. 4, 908–911 ~1992!.
9. S. Kawai, “Free-space optical crossbar-switches for self-routingand strictly nonblocking networks,” Opt. Rev. 4, 390–394 ~1997!.
