Active-fiber star coupler that uses arrays ofmicrolenses and liquid-crystal modulators

Kasra Rastani, Chinlon Lin, and Jay S. Patel

An active-fiber star coupler that uses arrays of N x N microlenses and liquid-crystal modulators ispresented. A simplified implementation of this device uses an 8 x 8 array of Fresnel microlenses fan outan incident beam into 64 focused spots (an 18 dB fan-out loss); on-off capability (270:1 extinction ratio) ateach element is provided by liquid-crystal spatial light modulators. A row of focused spots is coupled intoan array of 1 x 8 multimode fibers with a measured excess loss of 12 dB along each path (with anestimated overall loss of 30 dB in an 8 x 8 device). A modification of this device that is capable ofwavelength selection at individual output fibers is proposed for wavelength division multiplexingapplications.

Key words: Star coupler, microlens arrays, VVDM.

1. Introduction

Star couplers are becoming vital for optical signaldistribution in high-density optical communicationand computing networks. Although a common nameexplains their function star couplers come in manydifferent designs and capabilities. For example com-mercially available fused-fiber star couplers are basedon the evanescent coupling of fiber modes to fan outtheir optical signals without switching (passive fan-out). Integrated optical couplers in different materi-als such as lithium niobate (LiNbO3) and glass canalso be used to perform passive star coupling.1 Insuch components input and output fibers are coupledinto their respective waveguides on the substrate,where fan-out is performed. The electro-optic prop-erty of some waveguides, such as LiNbO3 , also allowsthe switching function to be performed in a starcoupler.5 6 An example, which is given in Ref. 6, is a1 x 16 switch with a 50-ns operation speed. Becauseof the planarity of optical waveguides, however, suchcouplers have limitations in number of fan-out ele-ments. The limitations are due to difficulty in fabri-cating and routing electrodes for electro-optic modu-lators and also the coupling of fiber arrays. Anothertype of passive star couple, which is based on micro-lenses and the three-dimensional nature of light

The authors are with Bellcore, NVC 3X-118, 331 NewmanSprings Road, Red Bank, New Jersey 07701-7040.

Received 25 July 1991.0003-6935/92/163046-05$05.00/0.© 1992 Optical Socioty of America

propagation, has been proposed by a few investiga-tors. 7 -9

Here we propose a fiber star-coupler architecture,also based on microlens arrays, that has the addedcapability of on-off switching of light beams in individ-ual fiber elements. The switching is provided by anarray of amplitude-spatial light modulators (A-SLM's). The proposed device has many applicationssuch as in subscriber loop optical networks in whichoptical signal distribution and switching are desired.A simplified implementation of the proposed devicecapable of coupling one fiber to an array of 1 x 8fibers by using arrays of binary-phase Fresnel lenses(8 x 8) and liquid-crystal SLM's (8 x 8) to provideon-off switching is presented in Section III. A modi-fication of the proposed device can also provide wave-length selectivity at output fibers, making the devicesuitable for wavelength division multiplexing (WDM)applications. This modification, which requires anarray of tunable liquid-crystal talon filters, is alsodescribed. A summary of the results is presented inSection IV.

I. Active-Fiber Star-Coupler Device Architecture

In this section we describe our proposed architecture[Fig. 1(a)] that fans out the optical signal output froma single fiber into an array of N x N fibers, with thecapability of switching each light beam path individu-ally. Modification of the proposed device to make itwavelength selective for VVDM applications is thenpresented.

In Fig. 1(a) a diverging beam from the end point ofan input fiber is collimated with a lens. The result-

3046 APPLIED OPTICS / Vol. 31, No. 16 / 1 June 1992

Liquid crystal

Spatial light modulator array

Tunable etalon filter array(NoN)

I

Input

fiber

Collimating lens

Holder block

I

Microlensarray

(No N)

(a)

y

Holder block

Liquid crystal

Spatial light modulator aray/

Tunable ealon 1itr array(N N)

Input fiberI

O output

fiberarray(NuN)

2' tput fiber

2 arrayJ (NuN)

Collimating Microlens arraymicrolens (N N)

(b)

Fig. 1. (a) Schematic of a fiber-star coupler architecture that usesa microlens array to perform fan-out and a liquid-crystal SLMarray to perform switching. To make the device wavelengthselective for WDM applications an array of liquid-crystal tunable6talon filters can be used. (b) A compact hybridized version of thearchitecture.

ing collimated beam is incident upon an array ofmicrolenses (N x N) after passing through a SLMarray. The function of the microlens array is two-fold. It multiplexes the incident beam into N x Nfocused beams, and it enhances the coupling effi-ciency into the output fiber array with better posi-tional tolerances. The fiber array, with separationsthat match those of the microlens array (N x N), issecured in a holder block. For positioning purposes,the holder block can be moved along x, y, z and alsocan be rotated about the x andy axes.

Switching of each of the multiplexed beams isperformed by an individually addressable array ofSLM's [see Fig. 1(a)]. The dimensions of this deviceand the microlens array also match.

In our architecture, since optical power in the inputfiber is multiplexed into N x N beams, a 1 /N2 fan-outloss is inherent in the device.10 Excess losses alsoexist because of Fresnel reflections, diffraction, andpropagation in the device. The throughput effi-ciency of a typical device is discussed in Section III,where we present results of implementing a simpli-fied version of Fig. 1(a).

A modification of the proposed device can make itwavelength selective at output fibers for WDM appli-cations in which the input fiber carries multiplewavelengths. This requires incorporating an array(N x N) of electronically tunable liquid-crystal 6talonfilters, as shown in Fig. 1(a). In this manner se-lected wavelengths can be coupled into individualfibers in the output array.

It should be pointed out that we have assumed thatthe total range of all optical wavelengths in the inputfiber is not larger than 100 nm. Given this condi-tion, the dispersion of the elements in the device isnegligible.

An electronically tunable liquid-crystal 6talon filterhas been reported." In this device a 1- to 5-V drivevoltage is used to continuously tune the 6talon across60 nm with a passband width of 0.5 nm. The use ofan array of this filter in our architecture can providesingle wavelength selection at each of the outputfibers. If more than one wavelength is desired ateach of the output fibers an array of tunable liquid-crystal 6talon filters with multiwavelength capabilitycan be used. A single element of such a devicecapable of filtering many wavelengths in the 1.2-1.6-pm range has been reported.' 2

One can foresee making the device of Fig. 1(a)compact and rugged as shown in Fig. 1(b). In Fig. 1(b)components are hybridized in blocks of a transparentmaterial such as glass. A liquid-crystal modulatoror tunable filter array is planar, which makes hybrid-ization practical. The wires needed to drive theindividual pixels of modulators or filters can beconnected from the side of the device. Since Fresnelmicrolens arrays are also planar, they lend them-selves well for the compact device.

111. Implementation of the Architecture

A simplified version of the architecture of Fig. 1(a)was implemented to examine its practicality andperformance. At this stage we only concentrate onthe use of modulators to perform on-off switching.Wavelength selection with tunable liquid-crystal 6ta-lon filters is left for future work.

We used a visible beam from a He-Ne laser(X = 0.633 [m), instead of beams at 1.3 or 1.5 jlm, asour source at this stage of the work. This choicemade alignment of the elements easier. Our fibersare multimode at this wavelength. This promptedus to use the collimated He-Ne beam for the inputsection of the implementation (in place of an inputfiber) to alleviate the spatial mode distribution. Forthe output stage a linear array of 1 x 8 fibers wasused.

First we dwell on the three major components ofthe implementation, namely, the microlens array, theliquid-crystal SLM array, and the fiber array holder.The experimental results are then presented.

We used a binary-phase Fresnel microlens arraythat was fabricated in house for multiplexing theinput beam. This choice was made because of thearray's planar nature and simplicity of fabrication.The array consists of 8 x 8 microlenses fabricated in aglass substrate through a selective ion-beam millingtechnique (see Ref. 13). An optical micrograph of asection of the array is shown in Fig. 2. Each lenshas an aperture of 1.2 mm with a primary focal lengthof 16 mm (for X = 0.633 jim). These parametersgive a divergence angle (half-angle) of 2° at thefocus. The focal spot size of each element is mea-

1 June 1992 / Vol. 31, No. 16 / APPLIED OPTICS 3047

I - 1.2mm IFig. 2. Optical micrograph of a section of the binary-phaseFresnel microlens array (8 x 8) used in our experiments.

sured at - 15 jim, and the diffraction efficiency intoprimary focus is 30% (a loss of 5.2 dB) (see Ref.13). The efficiency of the microlenses can be im-proved if one uses multiple phase levels at the cost ofincreased fabrication difficulty.'4

Both the divergence angle and the spot size aresmaller than the acceptance angle and the corediameter of the fiber to ensure efficient coupling.We used Corning Corguide multimode fiber, with a48-jim core, 125-jim cladding, and a numerical aper-ture of 0.21 (an acceptance angle of 120).

Modulation of the optical beam was provided bytwisted nematic liquid crystals sandwiched betweencrossed polarizers.15 The liquid-crystal modulatorwas turned into an array (8 x 8) by shaping a thinfilm of indium tin oxide (ITO) transparent electrodesinto circular pixels with appropriate contacts. Thegeometry of the electrodes is shown schematically inFig. 3(a), where for clarity contacts of the lower halfof the array have not been drawn. The photographin Fig. 3(b) shows the full pattern of electrodes andcontacts in our liquid-crystal modulator array. Cir-cles are each 0.7 mm in diameter and are situated on a1.2-mm pitch. The liquid-crystal film is 50 jimthick. Voltages of 1-5 V are applied to individualpixels by ITO thin-film wires routed to the perimeterof the device. The electronics that drive the modula-tor array consist of a supply box and control circuitsthat are addressed by a personal computer. Pixelsare turned off by 5-V applied voltage.

The throughput efficiency of the modulator (withcrossed polarizers) in the on state was measured to be- 30% (a loss of 5.2 dB). The extinction ratio of the

modulators was measured at 270:1 (24.3 dB). Theswitching speed of this type of liquid-crystal modula-tor was 10 ms. Compared with the LiNbO3 switch,which operates at 10 ns, this is very slow. How-ever ease of fabrication and also compatibility withmicrolens arrays make the liquid-crystal modulatorsattractive for our application.

At this stage of our work we opted for coupling into

--....,..... ............................E... i...---------..-.---- .-.---- .

.......,@ @ @,,@ @.,@ @ A~~~~~~~~~~~~~~~~~~...........

........... ..;:-:-:::.:..--------...... .A-----.-.. .......................I-...'''''

............ .... ., .................. .. ...............

............ @ .. ...........

........................................................... ...................................... .......................................

\ ~~~~~~~~~~~1.2 mm

1.2 mm

1 ~~8.4 mm

E _ ~~ITO transparent electrodesEl ~~~~and contact wires

(a)

(b)

Fig. 3. (a) Geometry of the electrodes and the wire contacts forthe liquid-crystal SLM array (8 x 8). Each pixel is individuallyaddressable. (b) Photograph of the liquid-crystal modulator arrayshowing all the electrodes and the appropriate contacts.

a one-dimensional array of 1 x 8 fibers with 1.2-mmspacing. Fibers (125 pim in diameter) were securedon a lucite holder with parallel V grooves that were200 m wide and 30 mm long (end to end on thesurface of the holder). After the fibers were routedthrough the grooves, their ends were individuallycleaved and cleaned, and they were all lined up withone another by pressing them against an optical flat.They were then secured in place on the holder by aclamp that had a rubber washer. A similar holderand similar procedure were used for the exit interfaceon the 1-m-long fibers. At the input end of the fiberarray provisions were made for x, y, and z movementsof the input holder. In addition the holder couldrotate about both the x and the z axes.

3048 APPLIED OPTICS / Vol. 31, No. 16 / 1 June 1992

------- - ............................ .......

To conduct the coupling experiment, we first alignedarrays of microlenses and liquid-crystal SLM's.While the pixels in the modulator array were all on,the fiber array holder was brought into position 16mm away from the microlens array. A CCD camerawas used to monitor the exit end of the fiber array.By using micropositioners and rotators coupling intothe fiber array and their uniformity were optimized.A photograph of the resulting CCD camera image isshown in the upper portion of Fig. 4. Couplingacross the 1 x 8 array was uniform, as evidenced bythe intensity scan shown in the lower part of Fig. 4.

To show the switching action, the liquid-crystalmodulator array was employed to turn off differentcombinations of the pixels, hence extinguishing theoptical beams in their respective fibers, as shown inthe sequence of photographs in Fig. 5.

In the above experiments, coupling efficiency in andout of each fiber was measured to be approximately70% (a loss of 1.5 dB). This efficiency is close to theexpected value determined from the overlap integralof the focused beam and the fiber mode at thiswavelength. Of course this efficiency will becomelower (higher loss) for longer wavelengths such as 1.3and 1.5 Lm.

The overall loss in the configuration of Fig. 1(a) istabulated in Table I to be 30 dB from the fiber inputto the output of fibers. Here a fan-out loss of 1/N2 isestimated at 18 dB for a 64-way split (8 x 8). Inaddition, the excess loss is estimated at 12 dB becauseof the modulator, the microlens, and the fiber cou-pling, based on the above analysis.

In the implementation of Figs. 1(a) and 1(b) diffi-culty in aligning the arrays of Fresnel microlensesand modulators can be eliminated by using electri-cally controlled liquid-crystal Fresnel lens arrays.' 6

An additional advantage of this type of lens-modula-tor array combination is its reduced excess loss. Infact the excess loss of 10.4 dB (because of a separate

IntensityScan

PeakIntensities

____________________....__ S Scale

Fig. 4. Upper photograph, CCD camera image of the output endof a 1 x 8 array of fibers. Lower photograph, Scan of the intensityacross the fiber array revealing uniformity.

Fig. 5. Sequence of photographs from top to bottom show dif-ferent combinations of elements in the 1 x 8 fibers turned off bythe liquid-crystal SLM array.

microlens and modulator) can be reduced to 5.2 dB.The added advantage of the liquid-crystal Fresnellens array reported in Ref. 16 is that it is polarizationindependent.

There are many factors limiting how large N in ourN x N active star coupler can be. One of the mostimportant limitations is due to 1/N2 fan-out. Thedensity of microlenses with adequate aperture andfocal length provides another constraint in this de-vice. One can foresee arrays as large as 32 x 32 inthis device, where fiber amplifiers are used to compen-sate for large fan-out loss.

IV. Summary

We propose an active-fiber star coupler that usesarrays of microlenses and SLM's. This device cou-ples one fiber into an array of N x N fibers with thecapability of switching each light beam path on andoff. A simplified implementation of this device ispresented in which a binary-phase Fresnel microlens

Table 1. Estimated Device Loss

Fan-Out Loss 18 dB(N x N) 10 LogN2 dB (N = 8)

Excess loss Liquid-crystal modulator -> 5.2 dBBinary-phase Fresnel lens -- 5.2 dB 12 dBFiber in-outcoupling -* 1.5 dB

Overall Loss -> 30 dB

1 June 1992 / Vol. 31, No. 16 / APPLIED OPTICS 3049

array (8 x 8) is used to fan out an incident beam into64 focused beams. An array of (8 x 8) liquid-crystalSLM's was used to switch individual elements on oroff (a 270:1 extinction ratio). A row of the focusedbeams is coupled into an array of 1 x 8 multimodefibers where each light beam path could be modu-lated. The overall loss of a complete system with 64(8 x 8) output fibers is estimated at 30 dB, whichincludes a fan-out loss of 18 dB and an excess loss of12 dB. The excess loss can be reduced to approxi-mately 7 dB in an improved device. In a modifica-tion of the proposed architecture, the use of an array(N x N) of electronically tunable liquid-crystal 6talonfilters can provide wavelength selectivity at individ-ual fibers of the output for VVDM applications. Select-ing either single or multiple wavelengths at eachoutput fiber may be possible, depending on the tun-able 6talon filter employed.

We are grateful to A. Marrakchi and J. Meyer ofBellcore for the design and implementation of thesoftware and electronics for controlling the liquid-crystal modulator array.

References1. B. H. Verbeek, C. H. Henry, N. A. Olsson, K. J. Orlowsky, R. F.

Kazarinov, and B. H. Johnson, "Integrated four-channelmulti/demultiplexer fabricated with phosphorous doped SiO2waveguides on Si," IEEE J. Lightwave Technol. 6, 1011-1015(1988).

2. C. Dragone, "Efficiency of a periodic array with nearly idealelement pattern," IEEE Photon. Technol. Lett. 1, 238-240(1989).

3. C. Dragone, C. H. Henry, I. P. Kaminow, and R. C. Kistler,"Efficient multichannel integrated optics star coupler onsilicon," IEEE Photon. Technol. Lett. 1, 241-243 (1989).

4. R. A. Becker and L. M. Johnson, "Low-loss multiple-branchingcircuit in Ti-in-diffused LiNbO3 channel waveguides," Opt.Lett. 9, 246-248 (1984).

5. G. A. Bogert, E. J. Murphy, and R. T. Ku, "Low crosstalk 4 x 4TiLiNbO3 optical switch with permanently attached polariza-tion maintaining fiber array," IEEE J. Lightwave Technol.LT-4, 1542-1545 (1986).

6. J. E. Watson, M. A. Milbrodt, and T. C. Rice, "A polarization-independent 1 x 16 guided-wave optical switch integrated onlithium niobate," IEEE J. Lightwave Technol. LT-4, 1171-1721 (1986).

7. H. Hosokawa and T. Yamashita, "ZnS micro-Fresnel lens andits uses," Appl. Opt. 29, 5106-5111 (1990).

8. K. Hamanaka, H. Nemoto, M. Oikawa, E. Okuda, and T.Kishimoto, "Multiple imaging and multiple Fourier transfor-mation using planar microlens arrays," Appl. Opt. 29, 4064-4070 (1990).

9. A. Akiba and K. Iga, "Image multiplexer using a planarmicrolens array," Appl. Opt. 29, 4092-4097 (1990).

10. J. W. Goodman, "Fan-in and fan-out with optical inter-connections," Opt. Acta 32, 1489-1496 (1985).

11. M. W. Maeda, J. S. Patel, C. Lin, J. Horrobin, and R. Spicer,"Electronically tunable liquid-crystal-6talon filter for high-density WDM systems," IEEE Photon. Technol. Lett. 2,820-822 (1990).

12. J. S. Patel and M. W. Maeda, "Multi-wavelength tunableliquid-crystal talon filter," IEEE Photon. Technol. Lett. 3,643-644 (1991).

13. K. Rastani, A. Marrakchi, S. F. Habiby, W. M. Hubbard, H.Gilchrist, and R. E. Nahory, "Binary phase Fresnel lenses forgeneration of two-dimensional beam arrays," Appl. Opt. 30,1347-1354 (1991).

14. J. Jahns and S. J. Walker, "Two-dimensional array of diffrac-tive microlenses fabricated by thin film deposition," Appl. Opt.29, 931-936 (1990).

15. J. Patel, "Liquid crystals for optical modulation," in SpatialLight Modulators and Applications III. Clinical Reviews, U.Efron, ed., Proc. Soc. Photo-Opt. Instrum. Eng. 1150, 14-26(1990).

16. J. S. Patel and K. Rastani, "Electrically controlled polarization-independent liquid-crystal Fresnel lens arrays," Opt. Lett. 16,532-534 (1991).

3050 APPLIED OPTICS / Vol. 31, No. 16 / 1 June 1992