Open Access
Add to BibSonomyAbstract
The dependence of laser performance on the active region position in broad-waveguide laser diodes is presented in this paper. Performance of structures with different position of active region is compared in simulation and actual devices. Lasers with active region displaced towards the p-cladding layer outperformed the lasers with active region undisplaced or displaced towards the n-cladding layer both in simulation and experimentally. Maximum output power increased by 25% for devices with active region displaced towards the p-cladding layer.
© 2014 Optical Society of America
1. Introduction
Usually the active region of a laser diode is placed in the middle of the waveguide, which is an acceptable practice for producing simple devices. However, to produce state-of-the-art lasers precise design of all laser parameters including the position of active region in the waveguide is required. This might be less obvious for thin waveguide lasers (transverse waveguide thickness of the order of few hundred nanometers) but in the design of broad-waveguide lasers (transverse waveguide thickness exceeding one micrometer), the position of the active region can dramatically affect the performance of the device. In this paper the influence of active region position on the performance of broad-waveguide lasers will be discussed.
Changing the position of active region in the waveguide changes a number of laser parameters. The most obvious effect is the change of the optical mode confinement factor. Displacing the active region to either side from the middle of the waveguide will reduce the optical confinement factor. This leads to a reduced modal gain with its the most obvious manifestation of increased threshold current (assuming it still satisfies the laser threshold condition). On the other hand the reduced optical confinement factor leads to increase of catastrophic optical damage energy density [1], the reduction of free carrier absorption in the active region [2,3] and laser gain-saturation [4]. There are also numerous reports of significantly increased output power from laser structures with reduced optical confinement [1–7].
Another positive effect of displaced active region is the possibility of transverse mode selection in very thick waveguide layers capable of supporting multiple transverse modes [8].
However there is one more parameter influenced by the active region position, which is usually neglected in the process of designing these broad-waveguide structures – active region carrier injection, especially hole injection. Despite two orders of magnitude difference in mobility values of electrons and holes in semiconductors and numerous reports demonstrating the fallibility of ambipolar injection model [9], (which uses single averaged mobility for both electrons and holes) it still has been used for laser design, resulting in the active region being placed in the middle of the waveguiding region.
There are papers highlighting the importance of hole injection, by reporting nonuniform and preferential localization of injected carriers in the quantum well closest to the p-cladding layer in multi quantum well structures [10], due to the lower mobility of holes, which preferentially occupy the closest QW and then electrostatically attract more electrons into the QW.
When considering the active region position in broad-waveguide high power single mode lasers some studies reported displacing the active regions from the center of the waveguide to reduce the optical confinement factor or for mode selection rather than improvement of the carrier injection [7,8]. Even though the reported structures have their active regions displaced towards the p-cladding layer instead of n-cladding layer the authors do this without referring to any quantitative study.
In this paper we investigate the influence of the active region position on laser performance, but unlike previous reports we isolate the influence of optical confinement factor and carrier injection. To achieve this, three laser structures were grown:
- • one symmetrical structure with the active region positioned in the middle of the waveguide layer;
- • two structures with the active regions displaced, one towards the p-cladding and another towards the n-cladding layer.
Both structures with the displaced active regions yield the same optical confinement factor (based on 1D analysis) since the active region displacements are symmetrical relatively to the middle of the waveguide layer. This isolates the effect of carrier injection in the two structures with displaced active regions.
2. Numerical modelling
To determine the optimum position of the active region a set of well-known formulas for minority carrier injection was used:
where p(x), n(x) – are the injected carrier density, pN0, nP0 – equilibrium minority carrier density, q – electron charge, Vf – applied bias, kB - Boltzmann constant, T – temperature, x – the coordinate, Lh, Le – carrier diffusion length.For simulation purposes we considered a 1 µm thick InGaAsP nominally undoped waveguide layer as lightly p-doped for electron injection calculation and lightly n-doped for hole injection calculation. Using well-known mobility values, diffusion length values and the value of applied forward bias corresponding to the typical InP laser threshold current, the concentration distributions of injected electrons and holes in the waveguide layer were calculated and the results are shown in Fig. 1. The abscissa represents the position in the waveguide layer, where holes are injected from the right (x > 3000 nm) and electrons from the left (x < 2000 nm). Needless to say that an equal number of electrons and holes are required for optimum laser performance and as shown in Fig. 1, in our case the curves intersect at around 2800 nm, which is 300 nm away from the center of the waveguide. Hence a value of 300 nm displacement was chosen for this study.
Fig. 1 Calculated concentration of injected electrons and holes into 1000 nm InGaAsP waveguide layer.
For numerical modelling we used a commercially available software package LaserMOD from RSOFT [12]. It is an integrated software package for designing and numerical modelling of semiconductor lasers. LaserMOD includes a number of mode solvers (including Beam Propagation Method), a free-carrier 8x8 band k·p gain calculation method, rate based carrier capture between bound and continuum states, and drift-diffusion equations fully coupled with thermionic emission and photon rate equations.
The structures are discussed in depth in “structure and fabrication” section. With the exception of the number of quantum wells (one quantum well was used in the simulations) the real structures were replicating the simulated structures as closely as possible.
The confinement factors for the structures were determined to be 0.69% for the undisplaced active region (from now onwards will be referred to as 000), 0.50% for the structure displaced towards the p-cladding (referred to as + 300) and 0.54% for the structure displaced towards the n-cladding (referred to as −300). The small difference in confinement factors of the ( + 300) and (−300) samples are due to ridge waveguide formation on top of thestructure which pushes the mode slightly towards the substrate, resulting in marginally better overlap with active region of the (−300) structure. The difference is very small and not expected to affect the results significantly.
Considering only the confinement factor, the performance of ( + 300) and (−300) structures should be almost identical, with their threshold current higher than that of structure (000), but less roll-over and as a result higher maximum output power [4]. However, as seen in Fig. 2, which depicts the results of numerical modelling in the form of light-current (L-I) curves, the performance of (−300) structure is significantly degraded compared to other samples, while the ( + 300) structure showed the best performance. The inset in Fig. 2 is the magnified image of the threshold region of the L-I curves, which illustrates the expected increase of threshold current with the decrease of the confinement factor. It is evident that other parameters rather than confinement factor have to be considered to explain the obtained results.
Fig. 2 Simulated light-current curves of the studded structures. The inset shows the enlarged threshold region.
The work of semiconductor devices is based on bandgap engineering, therefore the bandstructure has to be studied to explain the differences in laser performance. In Fig. 3, the numerically modelled conduction band diagrams of the structures under identical high injection bias are shown. The structure with the active region displaced towards the p-cladding ( + 300) appears to be the least deformed, while the structures (000) and (−300) are much more deformed, 28 meV and 32 meV lower than that of ( + 300) respectively.
The potential barrier for electrons consists of conduction band discontinuity between the waveguiding layer and the p-cladding layer, and the built-in potential. The different amount of deformation can be explained with the guide of Fig. 4 which shows the hole quasi-Fermi levels for the studied structures. The built-in potential is reduced by the applied voltage, which is equal to the hole quasi-Fermi level separation [11]. Efficient hole injection in structure ( + 300) provides enough holes in the active region for electrons to recombine with, while in the structures (000) and (−300) the number of holes in the quantum well issignificantly lower. This leads to lower number of electrons recombining in the quantum well and therefore, excess electrons spill over and accumulate at the waveguide – p-cladding interface. This excessive number of electrons deforms the band structure as seen in Fig. 3, by increasing the hole quasi-Fermi level separation, leading to the reduction of the potential barrier at the p-cladding layer and therefore increases electron thermionic leakage current.
Reduction of the potential barrier explains better performance of the ( + 300) structure. However the (−300) structure does not have significantly lower barrier height compared to the (000) structure, due to almost complete reduction of the built-in potential. To explain inferior performance of the (−300) structure as compared to the (000) structure we have to compare their electron quasi-Fermi levels. As shown in Fig. 5 electron quasi-Fermi level of the (−300) structure is 14 meV higher compared to the (000) structure in the vicinity of the waveguide – p-cladding interface, resulting in higher number of electrons capable of overcoming the barrier. As a result of both of these effects the electron thermionic leakage current across the waveguide – p-cladding interface is the lowest in ( + 300) structure, increases in the (000) structure and is the highest in the (−300) structure as can be seen in the inset of Fig. 6. It is the electron thermionic leakage that leads to an increased L-I roll-over in the (000) structure and even more so in the (−300) structure, as shown in Fig. 2.
Fig. 6 Simulated electron flux of the studied structures. The inset shows the blowup of the p-cladding region of the structure.
3. Structure and fabrication
All samples were MOCVD grown on (001) InP n+ substrates. An n+ (Si: 2x1018 cm−3) InP cladding was first grown, followed by 1 µm of undoped In0.86Ga0.14As0.30P0.70 waveguiding region, which also contained the active region. The active region consisted of five In0.60Ga0.40As compressively strained quantum wells separated by In0.79Ga0.21As0.45P0.55 barriers. In the first structure the active region is displaced by 300 nm from the middle towards the p-cladding ( + 300), in the second structure the active region is positioned in the middle of the waveguide layer (000) and in the third structure the active region is displaced by 300 nm from the middle towards the n-cladding (−300). After the waveguiding layer a 2 µm InP p-cladding (Zn: 2x1018 cm−3) layer was grown. The structures were terminated with a thin 50 nm In0.53Ga0.47As (Zn: 1x1019 cm−3) p+ contact layer.
Standard contact optical lithography was used to define the lasers’ 4 µm wide ridge waveguides. Ridges were formed by RIE etching with CH4-H2 chemistry using a previously established polymer free single step etching recipe [13], suitable for “soft mask” etching to the depth of 1.95 µm, followed by PECVD deposition of 200 nm of SiNx at 100 °C. After lift-off to expose the top of the ridge and solvent cleaning to remove photoresist residue, the top contact (Ti/Pt/Au) was deposited using an e-beam evaporator. After that the substrates were polished down to thickness of about 200 µm and cleaned with solvents, followed by the bottom contact deposition (Ge/Ni/Au) by e-beam evaporator. To form good ohmic contacts the samples were annealed in RTA for 1 min at 450 °C under Ar ambient. Subsequently the samples were cleaved into separate devices 1, 1.5, 2 and 3 millimeters long and tested, with the facets “as cleaved” (without any facet coating). Testing was carried out at a constant temperature of 10 °C. To avoid device heating the lasers were tested under pulsed condition (1% duty cycle and pulse duration of 100 ns). The light output was collected from a single facet and measured with an integrating powermeter.
4. Results and discussion
The averaged values (multiple devices for every device length of each structure were tested) of the maximum output powers for different device lengths of the studied structures are shown in Fig. 7. As expected the ( + 300) structure outperformed the other structures, with a clear trend of increasing margin with increasing device length, up to 25% for 3 mm long devices.
Fig. 7 Averaged measured maximum output power from different device length and different structures. (Lines are a guide to the eye).
Comparing the performance of the ( + 300) and (−300) structures clearly shows that the improvement in laser performance is due to improved hole injection in the ( + 300) structure and not merely due to the reduced optical confinement which is almost identical in both structures. This unquestionably proves the importance of carrier injection in broad-waveguide lasers, especially hole injection and the advantage of positioning the active region close to the p-cladding region for improved hole injection.
However, the experimental results shown in Fig. 7 agree with the modeled results only for 1 mm and 1.5 mm long devices, where the averaged maximum output power values of (000) sample falling between those of (−300) and ( + 300) samples. At 2 mm device length the averaged maximum output power values of samples (000) and (−300) overlap and at 3 mm device length the averaged maximum output power values of (000) sample become lower than that of (−300) sample. This can be explained by the gain’s sublinear dependence on the injection current. There is an optimum device length (where modal gain to injection current density ratio is maximum) for a certain modal gain provided by the active region [14]. Since the confinement factor for (000) structure is the highest its optimum device length is significantly shorter than for (−300) and ( + 300) structures [15]. Therefore for the 3 mm long devices the higher maximum output power of (−300) structure over the (000) structure is the result of lower confinement factor (higher gain to injection current density ratio) of the former. The same argument is applicable to explaining the increasing gap between the performance of ( + 300) and (−300) structures with increasing device length, since the confinement factor of the ( + 300) structure is slightly lower than that of the (−300).
5. Conclusion
We have presented a comparative study of three laser structures with the active region located in different positions in the waveguide. The structures were 1 µm broad-waveguide laser structures, with the reference structure (000) undisplaced (active region in the middle of the waveguide) while the other two structures have the active region displaced by 300 nm from the center towards n-cladding (−300) or p-cladding ( + 300) layer. The two displaced structures have almost identical confinement factors and by comparing their performance we could isolate the effect of carrier injection from the effect of reduced confinement factor.
The ( + 300) structure showed less roll-over and produced significantly higher averaged maximum output power, up to 25% for 3 mm long devices, compared to other structures. This illustrates the importance of carrier injection and active region position in broad-waveguide laser diodes and clearly proves that the improvement in laser performance is due to improved hole injection and not merely due to reduced optical confinement.
Acknowledgments
The Australian Research Council is acknowledged for its financial support. This work has been made possible through the access to the ACT Node of the Australian National Fabrication Facility
References and links
1. G. Iordache, M. Buda, G. A. Acket, T. G. van de Roer, L. M. F. Kaufmann, F. Karouta, C. Jagadish, and H. H. Tan, “High power CW output from low confinement asymmetric structure diode laser,” Electron. Lett. 35(2), 148–149 (1999). [CrossRef]
2. Y. Nagashima, S. Onuki, Y. Shimose, A. Yamada, and T. Kikugawa, “1480-nm pump laser with asymmetric quaternary cladding structure achieving high output power of >1.2 W with low power consumption,” in Proceedings of IEEE Semiconductor Laser Conference (Institute of Electrical and Electronics Engineers, 19th International, 2004), pp. 47–48. [CrossRef]
3. A. Guermache, V. Voiriot, N. Bouche, F. Lelarge, D. Locatelli, R. M. Capella, and J. Jacquet, “1W fibre coupled power InGaAsP/InP 14xx pump laser for Raman amplification,” Electron. Lett. 40(24), 1535–1536 (2004). [CrossRef]
4. M. Lysevych, H. H. Tan, F. Karouta, L. Fu, and C. Jagadish, “Merged beam laser design for reduction of gain-saturation and two-photon absorption in high power single mode semiconductor lasers,” Opt. Express 21(7), 8276–8285 (2013). [CrossRef] [PubMed]
5. L. J. Mawst, A. Bhattacharya, J. Lopez, D. Botez, D. Z. Garbuzov, L. DeMarco, J. C. Connolly, M. Jansen, F. Fang, and R. F. Nabiev, “8 W continuous wave front-facet power from broad-waveguide Al-free 980 nm diode lasers,” Appl. Phys. Lett. 69(11), 1532–1534 (1996). [CrossRef]
6. D. Garbuzov, L. Xu, S. R. Forest, R. Menna, R. Martinelli, and J. C. Connolly, “1.5 µm wavelength, SCH-MQW InGaAsP/InP broadened-waveguide laser diodes with low internal loss and high output power,” Electron. Lett. 32(18), 1717–1719 (1996). [CrossRef]
7. J. J. Plant, P. W. Juodawlkis, R. K. Huang, J. P. Donnelly, L. J. Missaggia, and K. G. Ray, “1.5-μm InGaAsP-InP slab-coupled optical waveguide lasers,” Photon. Technol. Lett. 17(4), 735–737 (2005). [CrossRef]
8. N. A. Pikhtin, S. O. Slipchenko, Z. N. Sokolova, A. L. Stankevich, D. A. Vinokurov, I. S. Tarasov, and Zh. I. Alferov, “16W continuous-wave output power from 100 lm-aperture laser with quantum well asymmetric heterostructure,” Electron. Lett. 40(22), 1413–1414 (2004). [CrossRef]
9. R. Nagarajan, “Carrier transport effects in quantum well lasers: an overview,” Opt. Quantum Electron. 26(7), S647–S666 (1994). [CrossRef]
10. H. Yamazaki, A. Tomita, M. Yamaguchi, and Y. Sasaki, “Evidence of nonuniform carrier distribution in multiple quantum well lasers,” Appl. Phys. Lett. 71(6), 767–769 (1997). [CrossRef]
11. R. F. Kazarinov and M. R. Pinto, “Carrier transport in laser heterostructures,” J. Quantum Electron. 30(1), 49–53 (1994). [CrossRef]
12. “LaserMOD user guide,” RSOFT Design Group, Inc.
13. M. Lysevych, H. H. Tan, F. Karouta, and C. Jagadish, “Single-step RIE fabrication process of low loss InP waveguide using CH4 / H2 chemistry,” J. Electrochem. Soc. 158(3), H281–H284 (2011). [CrossRef]
14. P. W. A. McIlroy, A. Kurobe, and Y. Uematsu, “Analysis and application of theoretical gain curves to the design of multi-quantum-well lasers,” J. Quantum Electron. 21(12), 1958–1963 (1985). [CrossRef]
15. F. Karouta, E. Smalbrugge, W. C. van der Vleuten, S. Gaillard, and G. A. Acket, “Fabrication of short GaAs wet-etched mirror lasers and their complex spectral behavior,” J. Quantum Electron. 34(8), 1474–1479 (1998). [CrossRef]
