Technical Insight
Quantum dot heterostructures progress to commercialization
A unique combination of self-organized quantum dot structures, an AlAs-GaAs material system and a defect-reduction technique promises to deliver dramatic advantages for GaAs-based devices in both optoelectronic and microelectronic applications, writes Nikolai Ledentsov.
The recent explosion of interest in quantum dot (QD) heterostructures for both microelectronic and optoelectronic applications is driven by the need to overcome some of the principal limitations of current technologies. For example, the increased complexity and number of channels required in telecom and datacom systems, together with increasing data bit rate per channel at equivalent or lower energy consumption, space and cost, demands energy-saving uncooled solutions with fewer components.
The benefits of QDs may be even greater than those of the two previous revolutions in this area: namely the replacement of p-n-homojunctions with devices based on double heterostructures in the early 1970s; and the introduction of quantum wells (QWs) in place of thicker layers at the beginning of the 1980s.
QD heterostructures feature size quantization in all three dimensions. Structurally, they represent tiny 3D insertions of a narrow bandgap material, coherently embedded in a wide-bandgap single-crystalline matrix. Because of their perfect crystal environment, the electronic and optical properties of heterostructure QDs resemble those of single atoms, rather than those of the bulk solid state, even though each QD may actually be composed of up to a few million single atoms. This distinguishes them from semiconductor QDs in glass matrices and QDs that are obtained by etching, where surface or interface states play a dramatic role.
Self-organized growthIn most cases, a combination of two or more self-organized growth techniques (see box) is used to fabricate QDs. For example, the most common approach to making GaAs-based QDs that emit light at relatively long wavelengths is based on a combination of the spontaneous formation of strained InAs nano-islands at a certain critical thickness of the deposit, followed by overgrowth of these islands with an InAs-based alloy. Phase separation of the alloy is activated by the non-uniformly strained InAs islands, the final size of which can be varied arbitrarily. Adding aluminum is known to increase the effect of alloy phase separation significantly, thereby creating large InAs islands that can emit light at beyond 1.3 µm.
Defect-free QDsOne way to reach these long wavelengths is related to the 3D nature of nano-insertions. Unlike the strained layer case, the energy accumulated in the strained nanodomain is finite, making the formation of extended defects, such as misfit dislocations, energetically unfavorable. Local defects, such as dislocated clusters, defect dipoles or dislocation loops, can be formed, but these can then be removed selectively using a sequence of in situ strain-sensitive deposition and annealing steps.
Defect-free QDs can be vertically stacked or merged. The lateral shape can be varied, and both approaches enable control over the emission spectrum and polarization of the light produced by radiative recombination. TM polarized or unpolarized edge emission can be obtained in vertically coupled InAs-GaAs QD structures.
The major advantage of QD lasers in datacom applications is that they allow access to the 1.31 µm emission wavelength, which is particularly important for ultrahigh bit rate data transfer at distances of 1-10 km. Low-cost vertical-cavity surface-emitting lasers (VCSELs) suitable for on-chip integration and applications in arrays are particularly desirable. Commercial VCSELs exist only on GaAs substrates and operate at 0.85 µm. (Editor s note: at the OFC conference in February, two companies claimed to have developed manufacturable VCSELs operating at 1.3 µm; see Guarded optimism reigns at fiber-optic industry s key event.) Recent development of 1.3 µm QD GaAs VCSELs currently operating at up to 2 mW single mode at room temperature may have a dramatic impact on the datacom industry.
QD laser advantagesAnother advantage of QD devices is their very high temperature stability. Optimized doping of the waveguide region enables completely temperature-insensitive operation, while maintaining a low threshold current and high differential efficiency.
High-temperature performance can be achieved in structures that have a low threshold current density and low optical nonlinearity. This enables ultra-narrow linewidth applications, which are important to reach high-frequency operation at a given link length. InGaAlAs QW lasers on InP substrates typically use 7-10 QWs to ensure high-temperature operation and, as a result, they suffer from high optical nonlinearity, wavelength chirp and beam filamentation.
Undoped QD lasers have recently been shown to operate with a perfect open-eye diagram at 5 Gbit/s. For p-doped QD lasers, the direct modulation speed is expected to be an order of magnitude higher. In results to be published at the Conference on Lasers and Electro-Optics (CLEO) in May, we have demonstrated passively mode-locked QD lasers operating at frequencies beyond 40 GHz, with low pulse width (2 ps), low jitter and high peak power of around 100 mW.
At the moment, high-performance QD GaAs lasers only operate at up to 1.35 µm. To extend this range, one can use metamorphic InGaAs buffer layers on top of GaAs-AlGaAs layers. Wafer-fused 1.5 µm VCSELs with InP lattice-matched active regions and GaAs-matched GaAs-AlAs distributed Bragg reflectors (DBRs) have been shown to be suitable for reliable continuous-wave operation, although they do suffer from low yield and high production costs.
Reducing dislocation defectsThe problem with metamorphic growth arises from threading dislocations, which propagate into the active region of devices and make practical applications, operating under high concentrations of injected non-equilibrium carriers, almost impossible.
The challenge is to block these threading dislocations from reaching the active region. As mentioned, QDs made of materials with similar lattice parameters tend to correlate vertically. For the same reason, there is a repulsive interaction between the relaxed region in a strained layer and a cap material with a lattice parameter close to that of the substrate.
In our defect-reduction technique (DRT), threading dislocations are blocked into trenches by faster growth of sidewalls. Regions near defects are removed by thermal etching. With thin layers, misfit dislocations can be healed, while local defects can be eliminated. Repulsive interaction between the temperature-stable cap layer and the relaxed region in the vicinity of the threading dislocation leaves the dislocation uncapped and enables selective evaporation of the dislocated region. For an appropriate depth of the etching pit, the growth front during the overgrowth stage merges before the defect penetrates into the cap epilayer. Stacking of DRT cycles further improves the effectiveness of DRT.
Using DRT, it is possible to make 1.5 µm QD lasers based on metamorphic InGaAs layers on GaAs substrates. The lasers operate at more than 80 ºC and emit over 7 W with a high slope efficiency (see figure). Unlike the 1.3 µm QD devices, which are robust at high temperatures and power levels, the reliability of 1.5 µm devices is yet to be demonstrated.
However, the development of GaAs VCSELs operating in the wavelength range beyond 1.42 µm will create enormous opportunities in metropolitan-area networks and coarse wavelength-division multiplexing.
Microelectronics applicationsDRT can be applied to both microelectronic and optoelectronic device structures. Metamorphic high electron mobility transistors (MHEMTs) and, in particular, metamorphic heterojunction bipolar transistors (HBTs) should benefit greatly from the reduced density of threading dislocations.
For example, by using vertically coupled type II QDs, where only electrons are confined, one can channel electrons through the heavily p-doped base while avoiding radiative and non-radiative leakage.
The main challenge now in QD growth by self-organized nanoepitaxy is the need for extensive structural and optical characterization of the media, which is required before optimum technological regimes, as well as deposition and annealing sequences, can be established. However, the experience gained in one material system is transferable to others, and progress is accelerating.
With commercial products now available (for example QD epiwafers for lasers operating over a broad spectral range, and metamorphic epiwafers for MHEMTs based on DRT), the real revolution has only just begun. To increase the overall level of competitiveness of the compound semiconductor industry and to dramatically extend its applications, the community should take a closer look at self-organized nanoepitaxy.
The benefits of QDs may be even greater than those of the two previous revolutions in this area: namely the replacement of p-n-homojunctions with devices based on double heterostructures in the early 1970s; and the introduction of quantum wells (QWs) in place of thicker layers at the beginning of the 1980s.
QD heterostructures feature size quantization in all three dimensions. Structurally, they represent tiny 3D insertions of a narrow bandgap material, coherently embedded in a wide-bandgap single-crystalline matrix. Because of their perfect crystal environment, the electronic and optical properties of heterostructure QDs resemble those of single atoms, rather than those of the bulk solid state, even though each QD may actually be composed of up to a few million single atoms. This distinguishes them from semiconductor QDs in glass matrices and QDs that are obtained by etching, where surface or interface states play a dramatic role.
Self-organized growthIn most cases, a combination of two or more self-organized growth techniques (see box) is used to fabricate QDs. For example, the most common approach to making GaAs-based QDs that emit light at relatively long wavelengths is based on a combination of the spontaneous formation of strained InAs nano-islands at a certain critical thickness of the deposit, followed by overgrowth of these islands with an InAs-based alloy. Phase separation of the alloy is activated by the non-uniformly strained InAs islands, the final size of which can be varied arbitrarily. Adding aluminum is known to increase the effect of alloy phase separation significantly, thereby creating large InAs islands that can emit light at beyond 1.3 µm.
Defect-free QDsOne way to reach these long wavelengths is related to the 3D nature of nano-insertions. Unlike the strained layer case, the energy accumulated in the strained nanodomain is finite, making the formation of extended defects, such as misfit dislocations, energetically unfavorable. Local defects, such as dislocated clusters, defect dipoles or dislocation loops, can be formed, but these can then be removed selectively using a sequence of in situ strain-sensitive deposition and annealing steps.
Defect-free QDs can be vertically stacked or merged. The lateral shape can be varied, and both approaches enable control over the emission spectrum and polarization of the light produced by radiative recombination. TM polarized or unpolarized edge emission can be obtained in vertically coupled InAs-GaAs QD structures.
The major advantage of QD lasers in datacom applications is that they allow access to the 1.31 µm emission wavelength, which is particularly important for ultrahigh bit rate data transfer at distances of 1-10 km. Low-cost vertical-cavity surface-emitting lasers (VCSELs) suitable for on-chip integration and applications in arrays are particularly desirable. Commercial VCSELs exist only on GaAs substrates and operate at 0.85 µm. (Editor s note: at the OFC conference in February, two companies claimed to have developed manufacturable VCSELs operating at 1.3 µm; see Guarded optimism reigns at fiber-optic industry s key event.) Recent development of 1.3 µm QD GaAs VCSELs currently operating at up to 2 mW single mode at room temperature may have a dramatic impact on the datacom industry.
QD laser advantagesAnother advantage of QD devices is their very high temperature stability. Optimized doping of the waveguide region enables completely temperature-insensitive operation, while maintaining a low threshold current and high differential efficiency.
High-temperature performance can be achieved in structures that have a low threshold current density and low optical nonlinearity. This enables ultra-narrow linewidth applications, which are important to reach high-frequency operation at a given link length. InGaAlAs QW lasers on InP substrates typically use 7-10 QWs to ensure high-temperature operation and, as a result, they suffer from high optical nonlinearity, wavelength chirp and beam filamentation.
Undoped QD lasers have recently been shown to operate with a perfect open-eye diagram at 5 Gbit/s. For p-doped QD lasers, the direct modulation speed is expected to be an order of magnitude higher. In results to be published at the Conference on Lasers and Electro-Optics (CLEO) in May, we have demonstrated passively mode-locked QD lasers operating at frequencies beyond 40 GHz, with low pulse width (2 ps), low jitter and high peak power of around 100 mW.
At the moment, high-performance QD GaAs lasers only operate at up to 1.35 µm. To extend this range, one can use metamorphic InGaAs buffer layers on top of GaAs-AlGaAs layers. Wafer-fused 1.5 µm VCSELs with InP lattice-matched active regions and GaAs-matched GaAs-AlAs distributed Bragg reflectors (DBRs) have been shown to be suitable for reliable continuous-wave operation, although they do suffer from low yield and high production costs.
Reducing dislocation defectsThe problem with metamorphic growth arises from threading dislocations, which propagate into the active region of devices and make practical applications, operating under high concentrations of injected non-equilibrium carriers, almost impossible.
The challenge is to block these threading dislocations from reaching the active region. As mentioned, QDs made of materials with similar lattice parameters tend to correlate vertically. For the same reason, there is a repulsive interaction between the relaxed region in a strained layer and a cap material with a lattice parameter close to that of the substrate.
In our defect-reduction technique (DRT), threading dislocations are blocked into trenches by faster growth of sidewalls. Regions near defects are removed by thermal etching. With thin layers, misfit dislocations can be healed, while local defects can be eliminated. Repulsive interaction between the temperature-stable cap layer and the relaxed region in the vicinity of the threading dislocation leaves the dislocation uncapped and enables selective evaporation of the dislocated region. For an appropriate depth of the etching pit, the growth front during the overgrowth stage merges before the defect penetrates into the cap epilayer. Stacking of DRT cycles further improves the effectiveness of DRT.
Using DRT, it is possible to make 1.5 µm QD lasers based on metamorphic InGaAs layers on GaAs substrates. The lasers operate at more than 80 ºC and emit over 7 W with a high slope efficiency (see figure). Unlike the 1.3 µm QD devices, which are robust at high temperatures and power levels, the reliability of 1.5 µm devices is yet to be demonstrated.
However, the development of GaAs VCSELs operating in the wavelength range beyond 1.42 µm will create enormous opportunities in metropolitan-area networks and coarse wavelength-division multiplexing.
Microelectronics applicationsDRT can be applied to both microelectronic and optoelectronic device structures. Metamorphic high electron mobility transistors (MHEMTs) and, in particular, metamorphic heterojunction bipolar transistors (HBTs) should benefit greatly from the reduced density of threading dislocations.
For example, by using vertically coupled type II QDs, where only electrons are confined, one can channel electrons through the heavily p-doped base while avoiding radiative and non-radiative leakage.
The main challenge now in QD growth by self-organized nanoepitaxy is the need for extensive structural and optical characterization of the media, which is required before optimum technological regimes, as well as deposition and annealing sequences, can be established. However, the experience gained in one material system is transferable to others, and progress is accelerating.
With commercial products now available (for example QD epiwafers for lasers operating over a broad spectral range, and metamorphic epiwafers for MHEMTs based on DRT), the real revolution has only just begun. To increase the overall level of competitiveness of the compound semiconductor industry and to dramatically extend its applications, the community should take a closer look at self-organized nanoepitaxy.
