High-Performance Visible Semiconductor Lasers ... - IEEE Xplore

High-Performance Visible Semiconductor Lasers Operating at 630 nm Volume 2, Number 4, August 2010 Bocang Qiu O. P. Kowalski Stewart McDougall, Member, IEEE B. Schmidt John H. Marsh, Fellow, IEEE

DOI: 10.1109/JPHOT.2010.2051022 1943-0655/$26.00 ©2010 IEEE

IEEE Photonics Journal

High-Performance Visible Semiconductor Laser

High-Performance Visible Semiconductor Lasers Operating at 630 nm Bocang Qiu, 1 O. P. Kowalski, 1 Stewart McDougall,1 Member, IEEE, B. Schmidt, 1 and John H. Marsh,1;2 Fellow, IEEE 2

1 Intense Ltd., Glasgow G72 0BN, U.K.GlasgowU.K. Department of Electronics and Electrical Engineering, University of Glasgow, Glasgow G12 8LT, U.K. GlasgowU.K.

DOI: 10.1109/JPHOT.2010.2051022 1943-0655/$26.00 Ó2010 IEEE

Manuscript received April 2, 2010; revised May 14, 2010; accepted May 16, 2010. Date of publication May 18, 2010; date of current version June 18, 2010. Corresponding author: B. Qiu (e-mail: [email protected]).

Abstract: A tensile-strained InGaP quantum-well laser structure operating around 630 nm has been designed. The structure offers low threshold current and beam divergence (19 fullwidth at half-maximum (FWHM) or 34 at 1=e2 ) simultaneously by incorporating a V-profile layer within the waveguide. Single-mode lasers (630 nm) were fabricated and tested. The threshold current for coated 1-mm-long and 3-m-wide ridge lasers is only 30 mA, and the slope efficiency is 1.04 W/A. This level of performance makes the lasers ideal for display applications. Index Terms: Red lasers, InGaAlP, low beam divergence.

1. Introduction Visible laser diodes emitting in the wavelength range 630–660 nm were first demonstrated in the mid 1980s [1], [2] and are now used in a wide variety of applications, including pumping solid-state lasers, digital printing [3], data-storage systems, photodynamic therapy, and laser projection displays [4]. For display applications, there is a tradeoff between the laser wavelength and power that is determined by the response of the human eye, with efficiency and reliability additional factors that need to be considered in any design. The ideal wavelength of 605 5 nm cannot be reached using practical red semiconductor lasers, and in order to achieve the same apparent brightness at longer wavelengths, higher powers are required: 2.5 times higher at 635 nm than at 605 nm and 4.8 times higher at 645 nm. As the wavelength increases, lasers exhibit improved temperature sensitivity and wall plug efficiency, but because more optical and electrical power is required, both the system efficiency and laser reliability are reduced. Taking these factors into account, emission around 635 nm is widely considered to be an optimum wavelength for display applications, particularly for batterydriven mobile pico-projectors. The system efficiency is also determined by the beam parameters, with a low divergence preferred so that compact and efficient optics with a small numerical aperture (NA) can be used. However, reducing the beam divergence without significantly impacting the threshold current in red lasers is challenging using conventional designs, because low beam divergence is usually achieved by increasing the spot size of the optical mode [5], leading to a decrease in , which is the optical confinement factor with the quantum wells (QWs). Therefore, there is generally a compromise between threshold current and vertical beam divergence.

Vol. 2, No. 4, August 2010

Page 563



Fig. 1. Near-field profiles for SCH waveguide and GRINSCH waveguide.

In red lasers, the situation is exaserbated by reduced material gain in semiconductors at shorter wavelength and poor carrier confinement in the QWs. Material gain is intrinsically inversely proportional to the photon energy; therefore, when the wavelength decreases, the material gain decreases as well. For the ðAly Ga1y Þ1x Inx P materials used in red lasers, the maximum direct bandgap energy is only about 2300 meV when the Al mole fraction is around 0.7, and Ec =Ev , which is the conduction to valence bandgap offset ratio, is about 0.3 [6], depending on the compositions of the QW and barrier materials, leading to very poor confinement of electrons in the QWs. As a result, the threshold current of red lasers is much more sensitive to than their longer wavelength (e.g., 980 nm) counterparts. In order to obtain both low beam divergence and low threshold, a novel approach to designing semiconductor the laser waveguide has been proposed and demonstrated [7], [8]. The approach uses an extra layer (hereafter refer to as far-field reduction layer or FRL in short) within the waveguide claddings to reduce the far-field diffraction angle while maintaining a good value. The approach has been applied previously to 650 nm red lasers [9]. However, the resulting far-field profile is not ideal as it becomes abnormal when the divergence angle is large. While the full-width at half-maximum (FWHM) value of the beam divergence was reduced to 24 from 37 , the beam divergence at 1=e 2 remained similar to that of structures with no FRL. In this paper, we apply the FRL design concept to visible semiconductor red lasers operating around 630 nm. We report an improved design of the epitaxial structure which combines optimization of the QW region with an FRL. Characterization of broad-area lasers confirmed that both a low threshold current and very low beam divergence can be achieved simultaneously. Compared with the reference structures, the structure with an FRL has lowest threshold current as well as the smallest beam divergence with a fast axis divergence of only 19 FWHM, or 34 at 1=e2 . Singlemode lasers had threshold currents as low as 30 mA with a slope efficiency more than 1.0 W/A, which makes the lasers ideal for display applications.

2. Waveguide Design Typically, two types of waveguide design are used in QW lasers i.e., the separate confinement heterostructure (SCH) and the graded refractive index separate confinement heterostructure (GRINSCH), and it seems from published literature that SCH designs are more commonly used for red lasers. As described in [9], by including an FRL in a red SCH laser structure, we can achieve low threshold current and low beam divergence simultaneously. However, it is not ideal as the far-field profile becomes abnormal when the divergence angle is large, and the beam divergence at 1=e2 remains similar to that without FRL. Modeling of the interaction between the waveguide core and the FRL indicated that a GRINSCH-based waveguide structure could lead to an improvement in the far-field profile owing to the slight change in the near field profile and much improved growth tolerance. For identical values of , GRINSCH based designs tend to have a slightly longer tail in the far-field spectrum (see Fig. 1), which helps to improve the interaction


Page 564



Fig. 2. Tolerance of factor over the variation of waveguide material.

Fig. 3. Tolerance of factor over the variation of FRL material.

between the waveguide core and the FRL in terms of reducing the beam divergence while having little effect on the factor. For a laser structure with an FRL design, due to the interaction between the waveguide core and the FRL, and the far field are intrinsically sensitive to the variations in the waveguide core, the FRL material composition, and its layer thickness (see Figs. 2 and 3). As can be seen, a GRINSCH based design has superior growth tolerance. For example for an SCH waveguide core, decreases by more than 14% when the Al mole fraction of ðAly Ga1y Þ1x Inx P at the barrier is increased by 0.01; this compares with only a 6% decrease for a GRINSCH waveguide core. Improved growth tolerance is highly desirable and offers a larger window to optimize the overall laser performance. In addition, it is believed GRINSCH waveguides offer extra carrier confinement, leading to improvements in threshold current and slope efficiency.

3. Far-Field Reduction Layer Modeling As can be seen in Fig. 4, the FRL is incorporated in the n-cladding layer. The FRL is essentially a much weaker waveguide compared with the waveguide core surrounding the active QWs and serves to expand the optical near field slightly in the vertical direction. Theoretical calculation also suggests the FRL can suppress higher order lateral mode lasing [7] because of increased discrimination between the modal gain of the fundamental mode and that of higher order modes. Expansion of the near field leads to a reduction of the corresponding far field through Fourier transformation. The


Page 565



Fig. 4. (a) SCH design. (b) GRINSCH/FRL design.

waveguide modes were found by solving the following Helmholtz equation using the transfer matrix method [10]: h i d 2E 2 2 2 ¼ K nðx Þ n E 0 e dx 2

(1)

where E denotes the electric field, K0 ¼ 2=0 , with 0 being the wavelength in vacuum, is the free space wavenumber, nðx Þ is the refractive indices of layers in the epitaxial structure, and ne is the effective index of a given waveguide mode. The transfer matrix method was implemented straightforwardly using MATLAB, thanks to its extensive support of a variety of matrix operations. The far-field spectrum was then obtained by calculating the Fourier transformation of the correspondent near optical field. The refractive indices of the materials used in the model were found using the following formula proposed by Afromowitz [11]: " !#1=2 2E02 Eg2 Ep2 Ed Ed Ep2 Ep4 ln þ þ (2) n ¼ 1þ E0 Eg2 Ep2 E03 where Ep ¼ 1:24=0 is the photon energy in eV at the wavelength of 0 in microns, and Eg is the material bandgap energy. In the case of material having strain, the degeneracy between the heavy and light hole bands is removed, and Eg therefore refers to the bandgap from the conduction band edge to the heavy or light hole band edge, whichever is smaller. E0 is the energy position parameter in the single-effective-oscillator model. Ed and are parameters which can be expressed as 1=2 Ed ¼ 2E02 Eg2 (3) ¼

2E03

Ed E02 Eg2

:

(4)

For red laser material ðAly Ga1y Þ0:51 In0:49 P, we have found design results agreed very well with test by using following fitting parameters for Ed and E0 : Ed ¼ 35:79 1:16y E0 ¼ 4:17 þ 0:39y :

(5) (6)

The FRL was then optimized to deliver the required beam divergence and . The parameters used for the optimization were the thickness of the FRL, the separation between the GRINSCH and


Page 566



Fig. 5. Vertical far-field versus gamma.

Fig. 6. Vertical far-field versus threshold current.

FRL, and the material composition at the mid-point of the FRL. Fig. 5 shows the relationship between the VFF and for the structure with and without an FRL. It is apparent that, for the structure with an FRL, the beam divergence can be reduced to below 20 FWHM without significantly affecting and, therefore, the threshold current. To evaluate the effect of the FRL on the threshold current, the structures were simulated using a detailed laser model which was validated using measured data from red laser structures. The model starts with the calculation of material gain for a given structure, and the threshold current is then calculated using, as an input, the value of calculated from the mode solver described before. The increase in threshold current for a 75-m-wide, 1-mm-long metal stripe laser is plotted in Fig. 6 as function of beam divergence (VFF) for designs with and without an FRL. It can be seen that the FRL enables the beam divergence to be reduced to below 20 and in the meantime there is a only limited increase in laser threshold current (G 15%). In contrast, the increase in threshold current is much more significant (9 170%) when using the conventional approach of reducing the thickness of SCH layers forming the waveguide core.

4. Fabrication and Test To verify the theoretical modeling, three different structures were grown. Structure A was an SCH with a VFF of around 28 ð 3:22%Þ; structure B had a reduced VFF of about 24 ð 2:82%Þ by reducing the SCH thickness of the structure A; and structure C had a GRINSCH waveguide core and an FRL incorporated in the n-cladding layer giving a VFF of 19 but at a of 3.32%. Wafers were grown using metal–organic vapor phase epitaxy (MOVPE) on an nþ GaAs substrate


Page 567



Fig. 7. Vertical far-field profiles of broad area lasers for structure A, B, and C.

Fig. 8. Percentage of power within a given total beam divergence.

misoriented by 10 toward h111i. The QW was a tensile-strained InGaP layer and the barrier was ðAl0:55 Ga0:4 Þ0:51 In0:49 P. The cladding layers were Al0:51 In0:49 P, and the material composition at the bottom of the FRL was ðAl0:78 Ga0:22 Þ0:51 In0:49 P with the total thickness being 720 nm. The wafer was completed by growing lower (n-side) and upper (p-side) InGaAlP cladding layers and, finally, the cap GaAs layer. Simple metal stripe lasers were fabricated to evaluate the material performance. The metal stripe was 75 m wide, and lasers were cleaved to a number of cavity lengths. The as-cleaved lasers were tested under pulsed conditions, with a repetition frequency of 2 kHz, a duty cycle of 2% and at room temperature. Fig. 7 plots the VFF of the three structures at same the power level, and Fig. 9 shows the light-current (L–I) characteristics. It can be seen in Fig. 7 that the values of VFF are 27 , 24 , and 19 for structures A, B and C, respectively, which are very close to the designed values. It is also salient that the far-field profile for structure C becomes non-Gaussian when the beam divergence angle is large. As was analyzed in [9], the features are linked to the part of the corresponding near-field distribution in which the gradient is steep. In spite of this non-Gaussian feature, more than 79% of the optical energy is contained within an angle of 20 . To fully assess the power distribution with respect to the beam divergence, one can calculate the fraction of the total power within a given total coupling angle (see Fig. 8) by integrating the far-field profiles in Fig. 7. Obviously, for a total coupling angle below 35 , Structure C offers better coupling than the other two structures; however, the coupling for structure C becomes similar to or even slightly poorer than that of structure A and B when the beam divergence is larger, due to the non-Gaussian features in the far-field profile of structure C. Modeling indicates it is still possible to improve the far-field profile further by carefully tuning parameters associated with the waveguide core, the FRL, and the


Page 568



Fig. 9. Broad area laser L–I plots for structure A, B, and C.

Fig. 10. Coated 1-mm-long 3-m wide single-mode ridge laser spectrum (Structure C).

separation between the waveguide core and the FRL. Of course, a tradeoff has to be made between the beam divergence and factor and, hence, threshold current. From the L–I plots (see Fig. 9), one can see that structure C has the smallest threshold current, even though its beam divergence is also much smaller. This exactly follows the modeling prediction (see Fig. 6) that a structure with an FRL can have much smaller threshold variation over a range of beam divergence. Also, it is interesting to observe that while structures A and C have very similar values, the threshold current for structure C is significantly smaller. We are not able to account for this discrepancy, beyond attributing it to the GRINSCH waveguide design which would provide stronger carrier confinement, resulting in improvements in laser threshold and slope efficiency. Further analysis of the dependence of laser power on the cavity length reveals that structure C indeed has a higher internal quantum efficiency (67%), compared with 51% for structure A. The internal losses for the three structures are similar, albeit structure C also has lowest value of 2.4/cm compared with 3.3/cm for structure A. Three-micrometer-wide ridge waveguide lasers were also fabricated in structure C after confirmation of its superior performance in broad area lasers. The etch depth was controlled such that only the fundamental mode was supported by the ridge waveguide. After completion of the fabrication process, including ridge definition, contact window opening, substrate thinning, metallization, and contact annealing, the wafer was chipped into bars with the laser cavity being 1 mm long. Subsequently, a facet coating was performed with antireflection (9%) for the front facet and high reflection (96%) for the rear facet. The ridge lasers were again tested pulsed using the previously specified conditions for L–I, spectra, and beam divergence. The beam divergence is identical to that measured on broad area lasers, but the spectral peak at 632 nm is about 3 nm shorter than the broad


Page 569



Fig. 11. Coated 1 mm long 3 m wide single mode ridge laser L–I curve (Structure C).

area lasers (see Fig. 10). Fig. 11 shows the measured L–I characteristics with a threshold current of 30 mA at 0.9 W/A and a slope efficiency of 1.04 W/A.

5. Summary We have demonstrated that the vertical beam divergence at FWHM of a red laser can be significantly reduced by incorporating an FRL into a laser structure while having little impact on laser performance. We have further shown that, by using a graded refractive index separate confinement heterostructure, growth tolerances can be significantly improved compared with those of the more commonly used SCH waveguide, and this improved growth tolerance not only improves yield in manufacture but also allows an improved design margin. Experimental results confirmed our design concept, with tests on 1-mm-long single-mode lasers exhibiting a threshold current around 30 mA, slope efficiency 9 1.0 W/A, an emission wavelength of 632 nm, and beam divergence of only 19 FWHM.

References [1] M. Ikeda, Y. Mori, H. Sato, K. Kaneko, and N. Watanabe, BRoom temperature continuous wave operation of an AlGaInP double heterostructure laser grown by atmospheric pressure metal–organic chemical vapor deposition,[ Appl. Phys. Lett., vol. 47, no. 10, pp. 1027–1028, Nov. 1985. [2] K. Kobayashi, S. Kawata, A. Gomyo, I. Hino, and T. Suzuki, BRoom temperature CW operation of AlGaInP double heterostructure visible lasers,[ Electron. Lett., vol. 21, no. 20, pp. 931–932, Sep. 1985. [3] O. P. Kowalski, S. D. McDougall, B. C. Qiu, G. H. Masterton, M. L. Armstrong, S. Robertson, S. Caldecott, and J. H. Marsh, BUltra-fine pitch individually addressable visible laser arrays for high speed digital printing applications,[ in Proc. SPIEVNovel In-Plane Semiconductor Lasers VIII (Photonics West), 2009, vol. 7230, p. 723 01J. [4] J. H. Lee, Y. K. Mun, S. W. Do, Y. C. Ko, D. H. Kong, B. S. Choi, J. M. Kim, C. W. Hong, and D. Y. Jeon, BLaser TV for home theatre,[ in Proc. SPIE, 2002, vol. 4657, pp. 138–145. [5] D. Botez, BDesign considerations and analytical approximations for high continuous power, broad-waveguide diode lasers,[ Appl. Phys. Lett., vol. 74, no. 21, pp. 3102–3104, May 1999. [6] B. Qiu, S. McDougall, D. Yanson, and J. H. Marsh, BAnalysis of the thermal performance of InGaP/InGaAlP quantum wells for high-power red laser diodes,[ Opt. Quantum Electron., vol. 40, no. 14/15, pp. 1149–1154, Nov. 2008. [7] B. Qiu, S. McDougall, G. Bacchin, X. Liu, and J. H. Marsh, BDesign and fabrication of low beam divergence and high kink-free power lasers,[ IEEE J. Quantum Electron., vol. 41, no. 9, pp. 1124–1130, Sep. 2005. [8] J.-M. Verdiell, M. Ziari, and D. F. Welch, BLow-loss coupling of 980 nm GaAs laser to cleaved single mode fibre,[ Electron. Lett., vol. 32, no. 19, pp. 1817–1818, Sep. 1996. [9] B. C. Qiu, O. P. Kowalski, S. D. McDougall, B. Schmidt, and J. H. Marsh, BHigh-performance red lasers with low beam divergence,[ IEEE Photon. J., vol. 1, no. 3, pp. 172–177, Sep. 2009. [10] J. Hong, W. Huang, and T. Makino, BOn the transfer matrix method for distributed-feedback waveguide devices,[ J. Lightw. Technol., vol. 10, no. 12, pp. 1860–1868, Dec. 1992. [11] M. A. Afromowitz, BRefractive index of Ga1x Alx As,[ Solid State Commun., vol. 15, no. 1, pp. 59–63, Jul. 1974.


Page 570