1. Introduction
In recent years,there has been a significant increase in interest in high power semiconductor laser diodes emitting in the spectral range of 1500 nm for their important role in free space optical communication,laser illumination and military applications. Only a few commercial companies have released their single mode high power product with output power above 400 mW[1, 2, 3]. However,few design details of these high power laser devices were mentioned in published academic literature,which makes systematic high power laser design challenging.
When designing high power laser diodes,one important step is to carefully choose a proper waveguide structure,which is capable of achieving minimal internal loss (αi),and maximal internal quantum efficiency (ηi). Studies have indicated that αi can be reduced by reducing free carrier absorption through increasing the thickness of the waveguide confinement layer and decreasing the number of QWs[4, 5]. However,when the waveguide thickness reaches above one micron and αi minimized,ηi can still be degraded by various physical mechanisms[6],including carrier leakage across multiple quantum wells (MQW)[7, 8] and non-radiative recombination process in the expanded waveguide layer[9]. In this article,we further analyze physical mechanisms affecting ηi with commercial semiconductor laser simulation software,afterward we fabricated and characterized actual laser structures in order to verify our simulation predictions and provide systematic design optimizations for high power InP/InGaAsP laser diode operation.
2. Numerical modelling
Figure 1 shows the schematic structure of 1.55 μm InGaAsP/InP board area laser used in our study. In order to determine the optimal active region design for high power operation,we adopt a commercial electron-magnetic (EM) semiconductor laser simulation software to study the carrier transport effects,in particular in the undoped active waveguide region.
The active region of our basic structure[10] consists of six 5-nm-thick compressively strained In0.63Ga0.37As0.9P0.1 multiple quantum wells (MQW) and seven 10-nm-thick tensile strained In0.738Ga0.262As0.568P0.362 barriers. The MQW is sandwiched between a pair of undoped 650-nm-thick lattice matched InGaAsP separate confinement layers (SCLs). Three designs with different bandgap wavelengths of the separate confinement layer (SCL) (Q1.0,Q1.1,and Q1.2) were used in our simulation. Above and below the active region,heavily doped p and n type (Si: 2 × 1018 cm−3) InP layers are used to complete the entire layer structure. Close to the active region,for the p type InP layer,a gradient doping (see Figure 1) was used,in order to minimize internal loss from free carrier absorption due to overlap between the optical mode and heavily doped p-cladding semiconductor material. Figure 2 shows the band energy diagram of the laser structure with Q1.2 InGaAsP confinement layers under flat band conditions.
Figure 3 shows the electron and hole current density profiles in the vertical direction of our epitaxial structure,for different SCL compositions (Q values) under 13333 A/cm2 current density injection level. The electron current density in the p-type InP layer in Figure 3(a) can be interpreted as the electron leakage current through the MQW active region,which can be quite a significant effect,while Figure 3(b) shows that the corresponding hole leakage current into the n-type InP layer is negligible and virtually zero. As expected from current continuity requirement,the sum of the electron and hole current given in Figures 3(a) and 3(b) respectively remains constant at all positions inside the epitaxial structure,shown by green dashed lines in Figures 3(a) and 3(b). The electron and hole leakage current components in the InP-cladding layer represent electrons and holes bypassing the recombination process at the MQW and leaking to the other side (in our case here dominated by electron leakage current for the high mobility of electron together with a rather small effective mass),which rises proportionally with increasing biasing current and is not clamped at the lasing threshold,and therefore represents a degradation in the internal quantum efficiency ηi. One can see that the electron leakage current can be effectively suppressed by increasing the bandgap wavelength of the SCL (Q value). In order to minimize the leakage current effect and improve ηi,we chose Q1.2 for the SCL layers in our design.
3. Device structure and experiment results
We fabricated four different designs of InGaAsP/InP QW broad area lasers with a 50-μm contact strip width. Figure 4 shows the band diagram of different waveguide structures (Type A,Type B,Type C,Type D) for our experiment. The design variations studied including increasing the SCL thickness,two step SCL structures,and asymmetric structure with a thinner upper SCL. Their performances in improving αi,ηi,and the overall slope efficiency were investigated systematically by comparison of experimental and simulation results.

Table 1 shows the thickness and composition values of the waveguide steps,and the number of QWs used in our design. For the asymmetric structure Type D,the thickness of the upper and lower SCLs are 100 nm and 400 nm respectively,in additional to the pair of 50 nm thick inner InGaAsP layers forming a double step SCH structure (Eg1 = 1.033 eV,which corresponds to 1.2Q). All the epitaxial structures were grown on (001) InP n+ substrates by metal-organic chemical vapor deposition (MOCVD) under identical epitaxial growth and processing conditions. The epitaxial growth starts by growing the 1.5 μm n+ (Si: 2 × 1018 cm−3) InP cladding layer,followed by the different waveguide structures given in Figure 4. A 1.5 μm InP p-cladding (Zn: 1 × 1018 cm−3) layer and a 200 nm InGaAs (Zn: 1 × 1019 cm−3) p+ contact layer were grown on top of the active region,completing the epitaxial structure growth.
The grown epitaxial structures were fabricated into standard broad area lasers with 50 μm strip width using the standard semiconductor laser fabrication process. Laser bars were cleaved with different cavity lengths and characterized under pulsed conditions,from which αi and ηi values were extracted with the results given in Table 1. Figure 5 shows a typical plot of the inverse differential quantum efficiency versus the cavity length for structure D. The data in Table 1 shows that αi rapidly decreases as the total waveguide thickness increases,due to the decrease in free carrier absorption resulting from reduced overlap between the optical modal profile and the highly doped cladding layers,in particular in the Zn doped p-type InP layer. For calculating the internal losses,we use: αi=Γnkn+Γpkp,where kn and kp are electron and hole absorption coefficients at λ = 1.5 μm. According to published results[11],these values are 20 cm−1 and 9 cm−1 for n- and p-doping level of 1 × 1018 cm−3,respectively. The lowest value of αi (4.5 cm−1) is obtained for structure C (W1 = 0.5 μm,W2 = 1.4 μm and QW number is 3),however the slope efficiency of structure C is not the highest among the four designs due to its poor ηi. The data in Table 1 show a reasonable agreement between calculated and experimental values of internal loss αi,and confirm that αi can be reduced by increasing the waveguide thickness and decreasing the number of QWs.
The difference values of ηi for the different design structures (Table 1) can be systematically explained as the following,based on simulation analysis. For the symmetric waveguide designs Types A,B,C,even though as discussed above the internal loss αi is reduced,the value of ηi also decreases with increasing the total cladding thickness. While the two step waveguide design of Type C further improves carrier confinement inside the MQWs and suppresses carrier leakage[12],ηi is significantly degraded by non-radiative recombination of electron and hole population in the 650 nm thick upper SCL layer,which is not clamped above the lasing threshold and monotonically rises with increasing the current level. The asymmetric two step waveguide Type D structure with the narrower upper SCL (150 nm) supports a reduced hole population and correspondingly suppresses the non-radiative recombination contribution in the cladding layers,resulting in the highest ηi as well as the highest slope efficiency among the four designs investigated.
Figure 6 shows the continuous wave (CW) optical power versus current for a 1 mm uncoated long broad area laser diode based on the type D design at a heat-sink temperature of 10 ℃. The threshold current is 276 mA (the current density = 552 A/cm2) and the spectral wavelength is 1538 nm (Figure 6,inset) besides the threshold. The single side output slope efficiency is 0.156 mW/mA and the measured output power can reach 160 mW at 1.3 A drive current. With proper HR/AR coating,the output power and slope efficiency are both expected to double in value.
4. Conclusion
We have presented a comparative study of four high power semiconductor diode laser epitaxial designs in the InGaAsP/InP material system. Internal loss αi can be reduced to 4.5 cm−1 by increasing the SCL thickness and decreasing the number of QWs,suppressing lower free carrier absorption effects. Through theoretical analysis and experimental studies,we show that ηi can be increased using an asymmetric two step SCL waveguide design with a thin upper SCL,through suppressing both the electron leakage current into the p-InP cladding layer,and the carrier population present inside the SCLs. The CW single side output power of 160 mW has been obtained for uncoated 1 mm long,50 μm wide broad area lasers with an asymmetric waveguide design. The theoretical and experimental results are quite consistent with each other,allowing us to extract effective design rules by understanding the fundamental device physics for these devices under high current injection conditions.