1154 OPTICS LETTERS / Vol. 20, No. 10 / May 15, 1995

Broadband multilayer antireflection coatingfor semiconductor laser facets

David M. Braun and Roger L. Jungerman

Microwave Technology Division, Hewlett-Packard Company, Mail Stop 1USK,1412 Fountaingrove Parkway, Santa Rosa, California 95403

Received January 10, 1995

Using a triple-layer antireflection coating of Al2O3, Si, and SiO2, we have achieved a minimum facet reflectivity of1 3 1026 and a bandwidth of 90 nm for a reflectivity of 5 3 1025 or less for 1550-nm center-wavelength InGaAsPsemiconductor lasers. A facet reflectivity of 3 3 1026 and a bandwidth of 30 nm for a reflectivity of 5 3 1025

were achieved for 1310-nm InGaAsP lasers. This coating is applicable to broadband external-cavity-tuned lasersources, edge-emitting light-emitting diodes, and semiconductor laser amplifiers.

Semiconductor lasers used as the gain media inexternal-cavity-tuned laser sources are required tohave low cavity–side facet reflectivity across the en-tire wavelength range of the source, which can be inexcess of 140 nm. Residual reflections from the cavi-ty–side facet introduce ripple in the output powerspectrum and multimoding.1 Other applications forlow-broadband facet reflectivity are edge-emittinglight-emitting diodes used in optical low-coherencereflectometry2 and semiconductor laser amplifiers.

Some methods of achieving low facet reflectivity aretilting the laser stripe relative to the facet,3 antireflec-tion (AR) coating the facet,4,5 or both.6 One drawbackof the tilted-stripe approach is that in cases in whichthe tilting is achieved by cleaving, usually the outputfacet also will be tilted, requiring an external reflec-tor in a linear tunable laser source and increasingthe complexity of the tunable laser source. The useof a curved waveguide to achieve a tilted stripe atonly a single facet7 would result in some increasein the complexity of the device. Single-layer ARcoatings have been calculated to have a minimumreflectivity of 1 3 1026,4 and by use of SiOx a facetreflectivity of 2 3 1025 has been demonstrated at 1290nm.5 However, the bandwidth of a single-layer coat-ing for a reflectivity less than 5 3 1024 is only 65 nmcalculated with a plane-wave model, which may limitthe tuning range of tunable laser sources. In thisLetter we present broadband low-reflectivity triple-layer AR coating designs for 1310- and 1550-nmcenter-wavelength semiconductor lasers. Theoreti-cal calculations and measured results from a largenumber of coating runs are presented.

The three materials used in our AR coating wereAl2O3, Si, and SiO2. The use of these three materi-als in an antireflection coating has been reported byO’Mahoney and Devlin,8 and laser facet reflectivitiesof 3 3 1024 to 5 3 1024 have been claimed,9,10 but fewdetails were published. Our design that used thesethree materials was a two-layer AR coating designwith a broadband single minimum, in which equiva-lent layers of Al2O3 and Si were used in place of thebottom layer, making the overall design a three-layerstructure. Reflectivities of 3 3 1026 and less wereachieved.

0146-9592/95/101154-03$6.00/0

An angular spectrum design approach has been de-veloped for single-layer4,11 and double-layer coatings,12

but the device structure for the lasers used in our studywere not well known. Our design approach beginswith a plane-wave model and incorporates the beamdivergence of the laser in an effective substrate refrac-tive index. For a given substrate in air with a planewave incident at normal incidence, an AR coating thatconsists of two layers, each with a quarter-wave opticalthickness, will have a single minimum at l0 if

n2 ­ nm1/2n1, (1)

where n1 is the refractive index of the top layer, n2

is the refractive index of the bottom layer, and nm isthe refractive index of the substrate. The solution toEq. (1) that gives the broadest result is

n2 ­ nm3/4, (2)

n1 ­ nm1/4. (3)

This model requires a substrate refractive index, whichwill depend on the refractive indices of the core and thecladding and on the thickness of the core.4 We callthis the substrate effective index. The substrate effec-tive indices listed in Table 1 are for InGaAsP buried-crescent lasers13 and for InGaAsP quantum-well ridgelasers used in this study and are the square of theempirically determined SiOx refractive index requiredfor optimum low reflectivity. Equations (2) and (3)show that layers with n2 ­ 2.49 and n1 ­ 1.36 willgive a maximum bandwidth with a single minimum atl ­ 1550 nm. MgF2 has a refractive index near therequired n1, but because we are more familiar withSiO2, which has a relatively close refractive index of1.43, we chose it instead for the top layer. With SiO2

as layer 1, Eq. (1) requires n2 ­ 2.63 to maintain asingle minimum at l0. However, no material withthis refractive index is readily available. Titaniumoxide14,15 (TiO2) and hafnium oxide (HfO2)16 have beenused as first layers, but both materials have refractiveindices lower than 2.63, and although low reflectivitieswere achieved by using less than quarter-wave optical

1995 Optical Society of America

May 15, 1995 / Vol. 20, No. 10 / OPTICS LETTERS 1155

Table 1. Antireflection Coating Models

l ­ 1310 nm l ­ 1550 nm

Layer n t n t

Air 1.0 1.0SiO2 1.44 0.27729l 1.43 0.27885l

Si 3.56 0.09943l 3.54 0.10005l

Al2O3 1.58 0.05445l 1.57 0.05506l

Laser 3.39 3.38

Fig. 1. Calculated facet reflectivities for the 1310- and1550-nm AR coating designs given in Table 1.

thicknesses, the coating bandwidth was less than thatfor an optimal two-layer design.

To provide a quarter-wave optically thick bot-tom layer with n2 ­ 2.63, we used the conceptof equivalent layers whereby this layer was re-placed by a two-layer structure. The two materialschosen were Al2O3 (n ­ 1.57), which has goodadhesion to InP, and Si (n ­ 3.54). The opticallayer thicknesses, calculated by an equivalent laseranalysis,17 were found to be 0.05128l for Al2O3

and 0.08550l for Si. This gave a zero mini-mum at l0, but the broadband minimum wasasymmetric and W shaped, and it shifted to lowerwavelengths. The broadband minimum was centeredby increasing each layer thickness by 9% and wasminimized by use of the damped least-squares re-finement algorithm FILM*STAR.18 The parametersderived for our AR coatings centered at 1310 and1550 nm are listed in Table 1, and the calculationsof reflectivities made with a plane-wave model areshown in Fig. 1. Dispersion of the refractive indiceswas assumed negligible.

The angular-spectrum design approach predicts agreater than quarter-wave optical thickness for anoptimum single-layer AR coating thickness, whereasa plane-wave design requires a quarter-wave opti-cal thickness. This thickness difference was observedwith single-layer coatings on the lasers used in thisstudy and may mean that our plane-wave multilayercoating design should be centered at a wavelengthgreater than the desired center wavelength. The mul-tilayer coating design, however, was too broadband forus to determine the wavelength offset empirically.

All three layers were electron-beam evaporated ina dedicated Balzers BAK640 box coater. An oxygenbackground pressure was used during the Al2O3 andSiO2 evaporations to ensure complete oxidation of

the metal, replacing any oxygen evaporated from thesource that may be pumped out of the chamber. Thelasers and witness wafers were placed directly abovethe source e-gun on a rotating platen. A crystal moni-tor was used for thickness control. The witness waferwas removed after each layer was deposited, and therefractive index and the thickness of the deposited filmwere measured by ellipsometry. On completion of allthree layers, laser facet reflectivities were calculatedby the Fabry–Perot modulation index method19 frommeasurements of the spontaneous emission spectrabefore and after coating.

Measured reflectivity for AR-coated lasers at eachwavelength, as well as the modeled reflectivity fromFig. 1, are shown in Figs. 2 and 3. Low facet re-flectivities of 3 3 1026 at 1310 nm and 1 3 1026 at1550 nm were achieved. The measured bandwidthfrom Figs. 2 and 3 for a reflectivity level of 5 3 1025

was 90 nm for the 1550-nm laser and 30 nm for the1310-nm laser. The measured bandwidth was limitedby the laser subthreshold gain spectrum. These re-sults are more broadband than those for an optimumsingle-layer coating.

Repetition of the results shown in Figs. 2 and 3 waslimited by the layer thickness and refractive-index con-trol. The standard deviation (s) of these parametersfor each layer is listed in Table 2. The cumulativehistory for 28 runs of 1550-nm lasers and 18 runs of1310-nm lasers was tabulated. A histogram of the re-

Fig. 2. Measured facet reflectivity from a 1550-nm laserantireflection coated with layers of Al2O3, Si, and SiO2(solid curve) and the calculated modeled reflectivity(dashed curve).

Fig. 3. Measured facet reflectivity from a 1310-nm laserantireflection coated with layers of AlO3, Si, and SiO2 (solidcurve) and the calculated modeled reflectivity (dashedcurve).

1156 OPTICS LETTERS / Vol. 20, No. 10 / May 15, 1995

Table 2. Layer Control Standard Deviation

Layer sna st (nm)b

Al2O3 0.008 1.3Si 0.023 0.7SiO2 0.002 5.0

The refractive-index standard deviation.The thickness standard deviation.

Fig. 4. Histogram of the measured laser facet reflectivityfor a series of 28 coating runs of 1550-nm lasers and 18coating runs of 1310-nm lasers. The reflectivity resultgraphed for each 1550-nm coating run is the average ofthe reflectivity at 1550-nm of each measured laser in thatrun. For the 1310-nm runs the average was measured at1310 nm.

flectivity for the runs is shown in Fig. 4. At 1550 nmthe mean facet reflectivity was 2 3 1024, with 32%of the runs having reflectivities below 1 3 1024. At1310 nm the mean facet reflectivity was 3.5 3 1024,with 28% of the runs having reflectivities below 1 31024. Layer control might be improved by use of real-time in situ monitoring.20

Also, we have coated up to 35 lasers in a single runand have routinely found their measured reflectivityprofiles to be within 61 3 1024 of the run average.This indicates that the three-layer AR coating is rela-tively insensitive to small changes in the laser effectiveindex caused by laser processing differences. This in-sensitivity also is apparent from examination of themodel.

It is well known that ambient conditions affect theoptical parameters of electron-beam evaporated SiO2.21

After a 60-min bath in boiling water, the refractive in-dex of our SiO2 layer decreased by 0.006 and the thick-ness increased by 0.1 nm. This change is sufficient toaffect the coating reflectivity. Therefore care must beused in packaging the laser for an application.

In summary, we have described a broadband triple-layer antireflection coating that consists of layersof Al2O3, Si, and SiO2. A minimum reflectivity of3 3 1026 has been achieved at center wavelengthsof 1310 and 1550 nm. Bandwidths of 30 and 90 nmwere achieved for a reflectivity of 5 3 1025 or less

for 1310- and 1550-nm lasers, respectively, limitedby the measurement technique. Results from a largenumber of coating runs show a mean facet reflectiv-ity of 2 3 1024 for 1550-nm lasers and 3.5 3 1024 for1310-nm lasers.

The authors thank G. Rankin for the single-layer an-tireflection coating results on lasers similar to thoseused in this study; T. Bagwell for reflectivity mea-surements of triple-layer AR-coated lasers; R. Bray forhelpful discussions; and N. Andring, T. Vincent, andS. LaFrancois for technical assistance.

References

1. P. Zorabedian, IEEE J. Quantum Electron. 30, 1542(1994).

2. J. Fouquet, W. Sorin, G. Trott, M. Ludowise, andD. Braun, IEEE Photon. Technol. Lett. 5, 509 (1993).

3. W. Rideout, R. Holmstrom, J. LaCourse, E. Meland, andW. Powazinik, Electron. Lett. 26, 36 (1990).

4. T. Saitoh, T. Mukai, and O. Mikami, J. LightwaveTechnol. LT-3, 288 (1985).

5. B. Luo, L. Wu, J. Chen, and Y. Lu, IEEE Photon.Technol. Lett. 5, 1279 (1993).

6. P. E. Barnsley, J. J. Isaac, and D. J. Elton, Electron.Lett. 26, 825 (1990).

7. R. Helkey, W. Zou, A. Mar, D. B. Young, and J. Bowers,IEEE Trans. Electron. Devices 40, 2107 (1993).

8. M. J. O’Mahony and W. J. Devlin, in Optical Amplifiersand Their Applications, Vol. 13 of 1990 OSA TechnicalDigest Series (Optical Society of America, Washington,D.C., 1990), p. 26.

9. W. J. Devlin, D. J. Elton, I. Reid, G. Sherlock, A. D.Ellis, D. M. Cooper, J. J. Isaac, P. C. Spurdens, A. J.Cockburn, and W. A. Stallard, in Proceedings of IOOC(Business Center for Academic Societies Japan, Tokyo,1989), paper 20C2-1.

10. W. J. Devlin, D. M. Cooper, P. C. Spurdens, G. Sherlock,M. Bagley, J. C. Regnault, and D. J. Elton, in Proceed-ings of InP and Related Materials (Institute of Electri-cal and Electronics Engineers, New York, 1990), paperWB.1.

11. G. Eisenstein, AT&T Bell Lab. Tech. J. 63, 357 (1984).12. C. Vassallo, Electron. Lett. 24, 61 (1988).13. Y. Sakakibara, H. Higuchi, E. Oomura, Y. Nakajima,

K. Goto, H. Namizaki, K. Ikeda, and W. Susaki, J.Lightwave Technol. LT-3, 978 (1985).

14. P. Doussiere, B. Mersali, A. Accard, P. Garabedian,G. Gelly, J. L. Lafragette, F. Leblond, G. Michaud, M.Moonot, and B. Fernier, in Optical Amplifiers andTheir Applications, Vol. 13 of 1991 OSA TechnicalDigest Series (Optical Society of America, Washing-ton, D.C., 1991), p. 68.

15. B. D. Patterson, J. E. Epler, B. Graf, H. W. Lehmann,and H. C. Sigg, IEEE J. Quantum Electron. 30, 703(1994).

16. M. C. Farries, J. Buus, and M. Q. Kearley, Electron.Lett. 26, 1626 (1990).

17. H. A. Macleod, Thin-Film Optical Filters, 2nd ed.(McGraw-Hill, New York, 1989), pp. 80, 118–119.

18. Trademark of FTG Software Associates, Princeton, N.J.19. I. P. Kaminow, G. Eisenstein, and L. W. Stulz, IEEE J.

Quantum Electron. QE-19, 493 (1983).20. I. Wu, I. Riant, J. M. Verdiell, and M. Dagenais, IEEE

Photon. Technol. Lett. 4, 991 (1992).21. L. I. Maissel and R. Glang, Handbook of Thin Film

Technology (McGraw-Hill, New York, 1970), pp. 11–48.