Coupling characteristics of lens systems for laser diodemodules using single-mode fiber
Kenji Kawano
The characteristics of lens systems for coupling InGaAsP laser diodes to single-mode (SM) fibers are studied.Two groups of lens systems are investigated. One group employs a combination of lenses and a SM fiber in aconventional configuration. The other group utilizes a combination of lenses and a virtual fiber, where thevirtual fiber is formed by attaching a GRIN rod lens to the input endface of the SM fiber. The maximumcoupling efficiency and misalignment tolerances for the optical circuit components in these lens systems arecompared from the viewpoint of fabricating laser diode modules. It is confirmed that lens systems usingvirtual fibers offer better coupling characteristics than does the conventional coupling method.
1. Introduction
To fully utilize the low-loss characteristics of single-mode (SM) fibers in the long-wavelength region, high-ly efficient and stable laser diode modules are indis-pensable. However, because butt joints provide poorcoupling efficiency, numerous methods have been pro-posed for reducing coupling loss at the laser diode-SMfiber joint.
These coupling methods can be divided into fourcategories. The first group employs microlenses ofextremely short focal lengths.'-"1 The second grouputilizes a relatively large lens, such as a hemisphericalGRIN rod lens.12 The third group employs a confocaltwo-lens method,13"4 where two relatively large lensesare positioned in a nearly confocal arrangement. Thefourth group makes use of lenses and the virtual fi-ber,15"16where the virtual fiber is formed by attaching aGRIN rod lens to the input endface of the SM fiber.This results in a virtual fiber spot size, which is largerthan the SM fiber spot size.
Although high coupling efficiency can be obtainedwhen the microlens methods are used in optimal cou-pling conditions, tight misalignment tolerances limitthe fabrication ease, yield rate, reproducibility, andstability of laser diode modules. Also, the difficulty ofensuring air-tightness and suppressing the reflectioneffects limit ease of fabrication. Since the latter three
The author is with NTT Electronics & Mechanics TechnologyLaboratories, Musashino-shi, Tokyo 180, Japan.
groups employ relatively large lenses having long focallengths, these coupling methods appear to offer advan-tages with respect to assembling laser diode modules.However, there are few reports concerning the cou-pling tolerances for methods using relatively largelenses.13-'6 Furthermore, no theoretical or experi-mental results have yet been presented to allow com-parison and evaluation of the coupling characteristicsof the large lens methods from the viewpoint of laserdiode module fabrication.
This paper describes the coupling characteristics ofthe latter three groups, which utilize relatively largelenses. The maximum coupling efficiency and mis-alignment tolerances are compared and discussed forthe same laser diode and SM fiber from the viewpointof laser diode module fabrication. The problems inthe microlens methods are discussed briefly in thispaper.
11. Coupling Schemes
The five methods investigated for coupling laserdiodes to SM fibers are illustrated in Figs. 1(a)-(e). InFig. 1(a) a hemispherical GRIN rod lens is used (meth-od A)12 and in Fig. 1(b) a relatively large spherical lensis combined with a GRIN rod lens (method B).13,14The coupling tolerances for the lens and fiber in meth-od A have not been previously reported. In these twocoupling methods, SM fibers are conventionally em-ployed and the misalignment tolerances for the SMfibers are determined by the small spot size (5.5 ,m).To improve the misalignment tolerances for SM fibers,the authors previously proposed two coupling meth-ods, whereby a virtual fiber was formed by attaching aGRIN rod lens to the SM fiber input endface.15"16 Inone method, the virtual fiber was employed in combi-nation with a relatively large spherical lens in Fig. 1(c)
LD HEMISPHERICAL GRIN ROD LENSGRIN ROD LENS SM FIBER
LENS 1 LENS 2(d)
LD SPHERICAL GRIN ROD LENSESLENS SM FIBER
LENS 1 LENS 2-1 LENS 2-2
(e)
Fig. 1. Sectional view of five coupling methods investigated forlaser diode modules; (a)-(e) correspond to methods A-E, respectively.
(method C).15 A hemispherical GRIN rod lens is uti-lized in Fig. 1(d) (method D). This arrangement hasbeen modified by using a hemispherical GRIN rodlens, which is described for the first time in this paper.In the other method, a relatively large spherical lenswas combined with a GRIN rod lens in Fig. (e) (meth-od E).'6
Since the beam divergence angle of the laser diodeand SM fiber spot size affect the maximum couplingefficiency and misalignment tolerances, the same laserdiode and SM fiber were employed in the couplingexperiments so that an exact comparison of the charac-teristics of each coupling method could be made. Themeasured far-field pattern of the laser diode emittingat 1.29 um was 35.6 X 25.50, perpendicular and parallelto the junction plane, respectively, at full width half-maximum (FWHM) power. The SM fiber has an o.d.of 124.5 m and a core diam of 9.8 um. The refractive-index difference between the core and cladding was0.22%. The spot size was calculated to be 5.6 m.
The spherical lenses used in methods B, C, and Ehad a radius of 0.4 mm, a refractive index of 1.78, and afocal length calculated to be 456 m. The axial refrac-tive index and focusing parameter of the hemispheri-cal and plane-ended GRIN rod lenses were 1.59 and0.32 mm-'.
In method A, the rod diameter, hemispheric radius,and pitch length of the hemispherical GRIN rod lenswere 1.8 mm, 2.0 mm, and 0.25 pitch, respectively.The focal length was calculated to be 2.0 mm. Inmethod B, lens 2 was a 0.2 pitch GRIN rod lens with afocal length of 2.1 mm. The SM fiber was convention-ally employed in both methods. In method C, lens 2
2 .Or
3E
0I-
I-
0
1.5.
1.0
Q5.
0
METHOD E
LD CHIP METHOD BOUT - METHOD C
I /I// METHOD AMETHOD D
SM FIBEROUT
BUTT JOINT
10 20 30 40
DRIVE CURRENT ( mA )
Fig. 2. Output power from a laser diode and SM fiber pigtail as afunction of laser diode drive current.
(2-2) was an 0.18 pitch GRIN rod lens,15 with a focallength calculated to be 2.2 mm. The virtual fiber spotsize was calculated to be 13 Am. In method D, lens 1 isa 0.25 pitch hemispherical GRIN rod lens. In methodE, lens 1 is a relatively large spherical lens. Lenses 2-1and 2-2 were 0.06 and 0.18 pitch GRIN rod lenses,16respectively. The focal length of lens 2-1 was calculat-ed to be 5.4 mm.
Ill. Experimental Results and Considerations
During the experiments, precise manipulators wereemployed to attain various coupling conditions.
A. Maximum Coupling Efficiency
The output power measured at the laser diode chipand at the SM fiber pigtail is shown in Fig. 2 as afunction of the laser diode drive current for couplingmethods A-E as well as for the butt joint method. Themaximum coupling efficiency was determined by mea-suring the differential quantum efficiency.
The maximum coupling efficiency was -4.5, -3.6,-4.1, -5.2, and -3.0 dB for methods A-E, respective-ly. Since the coupling efficiency for the butt jointmethod is -10.3 dB, the improvement in efficiency formethods A-E is 5.8,6.7,6.2,5.1, and 7.3 dB, respective-ly. Method E was the most efficient. The lower effi-ciency for methods A and D is thought to be due to thenumerical aperture and spherical aberration of thehemispherical GRIN rod lens. Only the lens in meth-od A, lens 2 in method B, and lens 1 in method D wereantireflection coated. A higher coupling efficiencycan be expected by applying antireflection coating tothe sides of the other lenses. It should be noted that amismatch of more than several tens of micrometersbetween the centers of the GRIN rod lenses and SMfibers in the virtual fiber had little effect on efficiency.Since no optical adjustment is required during assem-bly, the virtual fiber can be easily fabricated by solder-ing a GRIN rod lens and SM fiber ferrule to a cylindri-
Fig. 3. Normalized measured coupling efficiency as a function of (a) lateral offset, (b) angular misalignment, (c) axial misalignment for SM fi-ber and virtual fiber.
cal holder. Reliable fixation in the virtual fiber hasbeen confirmed with regard to solder creep.'7
To ensure laser diode reliability, it is necessary tohermetically seal the laser diode. Since the spacingbetween the optical circuit components in these cou-pling methods is relatively large, a hermetically sealedwindow can be easily inserted. On the other hand,since the spacing in the microlens methods is verysmall due to extremely short focal lengths, trouble-some fiber sealing is unavoidable.
Two techniques are currently used to suppress re-flection in optical circuit components. In one, an an-tireflection coating is applied to the sides of a lens. Inthe other technique, either an antireflection-coatedglass plate is attached to the input endface of the SMfiber or the input endface of the SM fiber is obliquelypolished. The antireflection coating can then be easi-ly applied, because the lenses are fixed in cylindricalholders. On the other hand, applying a tight antire-flection coating to the fabricated endfaces of the SMfibers in the microlens method is difficult because SMfiber is nylon coated.
B. Coupling Tolerances
In the assembly process, misalignment tolerancesfor the optical circuit components affect the ease withwhich laser diode modules can be fabricated. In actu-al use, the reliability of the modules is directly depen-dent on the degradation in coupling efficiency causedby the displacement of the optical circuit componentsdue to thermal expansion. Therefore, coupling toler-ances are just as important as maximum coupling effi-ciency.
1. Fiber ToleranceIn the following discussions, the measured values
correspond to a 0.5-dB coupling loss increase.Figures 3(a)-(c) represent the normalized measured
coupling efficiency as a function of the lateral offset,angular misalignment, and axial misalignment for theSM fiber and virtual fiber, respectively.
The lateral offset tolerance for the SM fiber is tightfor methods A (3.6 Mm) and B (3.2 Am) due to the SM
fiber's small spot size. On the other hand, the lateralmisalignment tolerance for the virtual fiber is aboutthree times larger for methods C (9 .8 Am), D (12.5 Mm),and E (10.7 Mim). This difference between the degra-dation characteristics of the coupling efficiency iq
caused by the SM fiber and virtual fiber offset can beexpressed by the following well-known formula18:
71 = exp (-9)X (1)
where w is the spot size of the light beam and x is theoffset of the SM fiber or virtual fiber. The spot sizecan be easily calculated utilizing the ray matrix ap-proach.'0 As shown in Eq. (1), a larger spot size canincrease the lateral offset tolerance.
The angular misalignment tolerance for the virtualfiber is 1.00, 1.20, and 1.10 for methods C, D, and E,respectively. Since the virtual fiber spot size is en-larged, these tolerances are smaller than those for theSM fiber, i.e., 3.10 for method A and 3.60 for method B,because the larger spot size reduces the angular mis-alignment tolerance in accordance with the followingequation'8:
(-7r 2wro2)77= exp x2 ) (2)
where 0 and X are the angular misalignment and laserdiode emission wavelength, respectively. However, itis important to note that these angular misalignmenttolerances cause no problem in laser diode fabrication,because the angular misalignment of the light beamcan be compensated by adjusting the other lens duringthe assembling process. Furthermore, it can be seenthat the lateral offset tolerance is more important inlaser diode module fabrication than the angular mis-alignment tolerance.
The axial misalignment tolerance for the virtual fi-ber is 700, 800, and 1000,Mm in methods, C, D, and E,respectively. This is a significant improvement com-pared with the SM fiber misalignment tolerances (106Mim) for both methods A and B). The larger spot size
also relaxes the axial misalignment tolerance, as shownin the following equation18:
* METHOD A" METHOD B" METHOD CC METHOD D* METHOD E
2" * X A
-50 0 50
LATERAL OFFSET ( m )Fig. 4. Normalized measured coupling efficiency as a function oflateral offset for the lens in method A and lens 1 in methods B-E.Positions of lens 2 or lens 2-1 and SM fiber or virtual fiber are
adjusted in the lateral direction.
1= 1 (3)
1+ ( 2)22rw2
where z designates the axial misalignment of the beamwaist.
Since the lateral and axial misalignment tolerancesfor the virtual fiber are relaxed in comparison withthose for the SM fiber, the ease of fabrication andreliability for the laser diode modules are expected tobe dramatically improved.
2. Lens Tolerances(1) When One Lens and Fiber are Adjusted. The
lateral offset of a lens causes a lateral offset and angu-lar misalignment of the light beam. The focal lengthsof the lenses utilized in the coupling methods studiedin this paper are relatively long. Therefore, the angu-lar misalignment of the light beam is not so serious aproblem as it was for the microhis methods. Further-more, the angular misalignment of the focused beam inmethods B and E can be transformed into a beamoffset by adjusting the positions of lens 2 or 2-1 in thelateral direction. This beam lateral offset can be com-pensated by laterally adjusting the SM fiber and virtu-al fiber.
The axial misalignment of the lens causes changes inthe position of the focused beam waist and variationsin spot size. By adjusting the SM fiber or virtual fiberposition in the axial direction, the coupling efficiencydeterioration caused by changes in the position of thelight beam waist can be completely compensated.Thus, the lens tolerances discussed in this subsectioncorrespond to the fabrication ease and reproducibilityof the laser diode module.
Figure 4 shows the normalized measured couplingefficiency as a function of the lateral offset for the lensin method A and lens 1 in methods B-E. Here thefollowing adjustments were made while the lens toler-ance was being measured: in method A, the SM fiber;in method B, lens 2 and SM fiber; in methods C and D,
the virtual fiber; in method E, lens 2-1 and virtualfiber. Permissible misalignments are 108, 118, 12.6,44, and 78 um for coupling methods A-E, respectively.The lateral offset tolerance for method C was the tight-est, because the focal length of lens 1 is shorter thanthat for the hemispherical GRIN rod lens and it isimpossible to compensate for the angular misalign-ment of the beam emitted from lens 1. For methods Aand D, these lens tolerances are large due to the rela-tively long focal length of the hemispherical GRIN rodlens. This is easily understandable from the followingcoupling efficiency equation'8:
7r2w2
X27 = exp - _
where f and x are the lens focal length and lateral offsetfor the lens, respectively. Longer focal length lensesincrease the lenses' lateral offset tolerance. Since thelens focal lengths in the microlens methods are ex-tremely short (- 1 0-20Mm), the lateral offset tolerancefor these lenses is very tight. 9 -1"
The light beam tilt can be easily compensated formethods B and E by adjusting the position of lens 2and the SM fiber or lens 2-1 and the virtual fiberpositions as mentioned earlier. Since a lateral offsetof about ±8 um for lens 1 corresponds to a beam tilt of±10, the permissible tilt of the beam emitted from lens
1 is ±7.40 and ±4.90 for methods B and E, respectively.As lens 1 can be fixed by soldering within a 10 beam tilt,these lateral offset tolerances are believed largeenough for both coupling methods. The lens 1 offsettolerance in method D is also large in the laser diodemodule assembling process.
Figure 5 shows the normalized measured couplingefficiency as a function of the axial misalignment for(a) the lens in method A and lens 1 in method D, (b)lens 1 in methods B and E, and (c) lens 1 in method D.Here, the SM fiber and virtual fiber were adjusted inthe axial direction. The permissible misalignmentsfor each lens are 370, >50, 24.2, 100, and >50 gm formethods A-E, respectively. The axial misalignmenttolerance for lens 1 in method C is tight and so is thelateral offset tolerance. Since the focal length of lens 1in method D is longer than that for method C, the axialmisalignment tolerance for lens 1 in method D is larg-er. The light beam emitted from lens 1 in methods Band E is nearly collimated. Thus, determination ofthe lens 1 position in both lateral and axial directionscan be easily performed during the actual assemblyprocess.
Figure 6 shows the normalized measured couplingefficiency as a function of the lateral offset for lens 2and lens 2-1 in methods B and E. Here the measure-ments were accompanied by lateral adjustments forthe SM fiber and virtual fiber. The permissible mis-alignments are 117 and 101 um for methods B and E,respectively. These are large enough to easily fabri-cate laser diode modules.
Figure 7 shows the normalized measured couplingefficiency as a function of the axial misalignment forlenses 2 and 2-1 in methods B and E. When adjusting
Normalized measured coupling efficiency as a function of axial misalignment for (a) lens in method A, (b) lens 1 in methods B, C, andE, and (c) lens 1 in method D. Positions of SM fiber or virtual fiber are adjusted in the axial direction.
Co0moz
c] UW W -2
o 0 -4Z 11
zzD
0L0
"METHOD B*METHOD E
* * * af ftf
-100
0
-2
IL
z z-J
0 4
00 100
LATERAL OFFSET (m )
Fig. 6. Normalized measured coupling efficiency as a function oflateral offset for lens 2 in method B and lens 2-1 in method E.Positions of SM fiber or virtual fiber are adjusted in the later
direction.
"METHOD B-METHOD E
. f t e . * .
-2000 0 2000
AXIAL MISALIGNMENT (m )
Fig. 7. Normalized measured coupling efficiency as a function ofaxial offset for lens 2 in method B and lens 2-1 in method E.Positions of SM fiber and virtual fiber are adjusted in the axial
direction.
the SM fiber and virtual fiber in the axial direction, thetolerable misalignments are 3480 and 4320 m, respec-tively. These extremely large misalignment toler-ances make laser diode module design and fabricationquite easy.
Since the focal lengths of the lenses in methods A-Eare relatively long, the lateral and axial misalignmenttolerances for the lenses offer a drastic improvementwhen compared with the microlens methods.9-1"
(2) When Lens and Fiber Are Not Adjusted.When no adjustments are made to correct lens mis-alignments, the lens tolerances exert a dominant effecton the reliability of the fabricated laser diode modules.
Figure 8 shows the normalized coupling efficiency asa function of the lens in method A and lens 1 in meth-ods B-E. Since the lateral offset for lens 1 is magni-fied by the image magnification factor at the focusingplane, the offset tolerance for lens 1 is extremely tight(-1 ,m) as is the case for the microlens methods.'-"Therefore, rigid fixation of lens 1 is essential for ob-taining highly reliable laser diode modules even inmethods employing a virtual fiber. Since a small hold-er for lens 1 can be directly soldered to the laser diodeheat sink in methods B and E, these methods offeradvantages for obtaining high reliability. The lateraland axial misalignments caused in the soldering pro-
c] 0WJZ_
N 2J 0
< 11ILL
o oD-4ZZ-JEL0
* METHOD A" METHOD BXMETHOD C* METH D* METHOD E
I= fe
-1 0 1
LATERAL OFFSET (m)Fig. 8. Normalized measured coupling efficiency as a function oflateral offset for the lens in method A and lense 1 in methods B-E.
No adjustment has been made.
cess can be completely compensated for in the subse-quent assembly process.
Figure 9 shows the normalized measured couplingefficiency as a function of the lateral offset for lens 2 inmethod B and lens 2-1 in method E. The permissiblemisalignment in method E (8.0 Mm) is larger than thatfor lens 2 in method B (3.0 ,m). It is noted that themeasured results for methods A and D in Fig. 8 corre-
Fig. 9. Normalized measured coupling efficiency as a function oflateral offset for lens 2 in method B and lens 2-1 in method E. No
adjustment has been made.
spond to the results shown in Fig. 9. This results fromthe structural similarity of the laser diode modules.Therefore, fabricated laser diode modules employingmethod E are expected to be highly reliable.
IV. Conclusion
The coupling characteristics of lens systems for laserdiode modules used in SM fiber transmission systemsare discussed from the viewpoint of laser diode modulefabrication. These coupling methods conventionallyemployed a SM fiber or of late utilized a virtual fiber.
The misalignment tolerance for the virtual fiber,formed by attaching a GRIN rod lens to the inputendface of the SM fiber, was improved by factors of 3and 9 in the lateral and axial directions, respectively,compared with the misalignment tolerance for the SMfiber. When the GRIN rod lens and SM fiber weredisplaced by more than several tens of micrometers, nodeterioration resulted in coupling efficiency. There-fore virtual fiber can be easily assembled without theneed for optical adjustment. Although the lateral off-set tolerance for lens 2-1 in method E was similar tothat for lens 2 in method B prior to assembling the laserdiode module, the tolerance in method E is better thanthat in method B after the modules were fabricated.Furthermore, since the distances between lens 1, lens2-1, and the virtual fiber were shorter than those forthe confocal conditions, the total axial length of thecoupling scheme E is compact.
The author believes that method E is superior to theother methods with respect to the maximum couplingefficiency and misalignment tolerances. Laser diodemodules with high coupling efficiency, good reproduc-ibility, and high reliability should be fabricated byemploying coupling method E.
The author wishes to thank Hiromichi Jumonji, Sec-tion Chief, and Toshinori Nozawa, Staff Engineer, fortheir constant encouragement and thoughtful discus-sions throughout this project. He appreciates OsamuMitomi, Staff Engineer, for the time and useful theo-retical and experimental advice he contributed.Thanks are also due to Masatoshi Saruwatari, StaffEngineer, for his stimulating technical discussions.
The author is also indebted to Hiroshi Miyazawa,Engineer, for his technical assistance in the experi-ments and detailed discussions.
References1. L. G. Cohen and M. V. Schneider, "Microlenses for Coupling
Junction Lasers to Optical Fibers," Appl. Opt. 13, 89 (1974).2. E. Weidel, "Light Coupling from a Junction Laser into a Mono-
mode Fibre with a Glass Cylindrical Lens on the Fibre End,"Opt. Commun. 12, 93 (1974).
3. H. Kuwahara, M. Sasaki, and N. Tokoyo, "Efficient Couplingfrom Semiconductor Lasers into Single-Mode Fibers with Ta-pered Hemispherical Ends," Appl. Opt 19, 2578 (1980).
4. J. Sakai and T. Kimura, "Design of a Miniature Lens for Semi-conductor Laser to Single-Mode Fiber Coupling," IEEE J.Quantum Electron, QE-16, 1059 (1980).
5. J. Yamada, Y. Murakami, J. Sakai, and T. Kimura, "Character-istics of a Hemispherical Microlens for Coupling Between aSemiconductor Laser and Single-Mode Fiber," IEEE J. Quan-tum Electron. QE-16, 1067 (1980).
6. H. Sakaguchi, N. Seki, and S. Yamamoto, "High EfficiencyCoupling from Laser Diodes into Single-Mode Fibers with Qua-drangular Pyramid-Shaped Hemielliptical Ends," in TechnicalDigest, Third International Conference on Integrated Opticsand Optical Fiber Communication (Optical Society of America,Washington, DC, 1981), paper TULL.
7. G. Eisenstein and D. Vitello, "Chemically Etched Conical Mi-crolenses for Coupling Single-Mode Lasers into Single-ModeFibers," Appl. Opt. 21, 3470 (1982).
8. G. D. Khoe, J. Poulissen, and H. M. de Vrieze, "Efficient Cou-pling of Laser Diodes to Tapered Monomode Fibers with HighIndex End," Electron. Lett. 19, 205 (1983).
9. E. Weidel, "New Coupling Method for GaAs-Laser-Fiber Cou-pling," Electron. Lett. 11, 436 (1975).
10. M. Saruwatari and K. Nawata, "Semiconductor Laser to Single-Mode Fiber Coupler," Appl. Opt. 18, 1847 (1979).
11. Y. Odagiri, M. Shikada, and K. Kobayashi, "High-EfficiencyLaser-to-Fiber Coupling Circuit Using a Combination of a Cy-lindrical Lens and a Selfoc Lens," Electron. Lett. 13,395 (1977).
12. J-I. Minowa, M. Saruwatari, and N. Suzuki, "Optical Compon-entry Utilized in Field Trial of Single-Mode Fiber Long-HaulTransmission," IEEE J. Quantum Electron. QE-18, 705 (1982).
13. M. Saruwatari and T. Sugie, "Efficient Laser Diode to Single-Mode Fiber Coupling Using a Combination of Two Lenses inConfocal Condition," IEEE J. Quantum Electron. QE-17, 1021(1981).
14. T. Sugie and M. Saruwatari, "Semiconductor Laser Module forSingle-Mode Fiber Transmission System Using a Combinationof Confocal Two Lenses," Trans. IECE Jp. J65-B, 374 (1982), inJapanese.
15. K. Kawano, 0. Mitomi, and M. Saruwatari, "Combination LensMethod for Coupling a Laser Diode to a Single-Mode Fiber,"Appl. Opt. 24, 984 (1985).
16. K. Kawano, M. Saruwatari, and 0. Mitomi, "A New ConfocalCombination Lens Method for a Laser Diode Module Using aSingle-Mode Fibre," IEEE/OSA J. Lightwave Technol. LT-3,739 (1985).
17. 0. Mitomi, T. Nozawa, and K. Kawano, "Effect of Solder Creepson Optical Component Reliability," in Proceedings, IEEECHMT Symposium, Tokyo (1984), p. 198.
18. H. Kogelnik, "Coupling and Conversion Coefficients for OpticalModes," in Microwave Research Institute Symposia Series,Vol. 14, J. Fox, Ed. (Polytechnic Press, Brooklyn, 1964), p. 333.
