Technical Insight
Closing in on the monolithic blue VCSEL
Blue VCSELs could produce continuous-wave emission following improvements to the indium tin oxide current-spreading layer
Today’s most powerful GaN-based VCSELs are formed using a series of complex processes to bond a pair of dielectric mirrors to an active region. That’s because the more elegant, monolithic approach currently delivers inferior output powers – but this performance gap could be slashed by optimizing the indium tin oxide layer, according to researchers at EPFL, Switzerland.
This is the next step for the team that is pioneering the development of monolithic blue VCSELs, which could be used for high-resolution printing and bio-sampling. Their latest breakthrough is the creation of VCSELs on "defect-free" epitaxial distributed Bragg reflectors (DBRs) made of lattice-matched InAlN and GaN.
“Here, defect-free means that no further dislocations are introduced during the growth, therefore the dislocation density seen by the quantum wells of the active region is that of the free-standing GaN substrate used for the growth”, says lead author Gatien Cosendey. Hepoints out that the device is less susceptible to over-heating when it is built with epitaxial DBRs, rather than those made from dieletrics.
Another important aspect of this team’s work is the development of a current confinement scheme based on plasma treatment of the surface of p-GaN. Going down this route avoids the requirement for a thin dielectric aperture.
“This prevents diffraction losses at the edges of this dielectric aperture, but increases the growth time – it’s about 20 hours for a 40-pair DBR,” explains Cosendey.
Growth times are much shorter for VCSELs made with a pair of dielectric mirrors, but this approach requires laser lift-off, wafer bonding, high-quality polishing and cavity thickness control.
Although laser lift-off and wafer bonding are widely used in the LED industry, it doesn’t follow that they are simple, high-yield steps for VCSEL fabrication. “Rough surfaces are required in LEDs to improve light extraction efficiency, while very smooth VCSEL cavity surfaces are needed to reduce diffraction losses and thus decrease lasing threshold,” explains Cosendey.
According to him, one of the advantages that VCSELs traditionally have over their edge-emitting lasers is mono-mode emission. This is not the case in the VCSELs made by Nichia and Panasonic that sport a pair of dielectric DBRs – these devices feature long cavities that support several modes.
“On the other hand, UCSB [which also uses a pair of dielectric DBRs] has developed a photoelectrochemical etch process that allows precise control of the cavity length while ensuring good surface morphology,” says Cosendey.
The VCSELs fabricated by the EPFL team feature: A 41.5 pair Al0.8In0.2N/GaN DBR; a pin diode structure, which contains an active region with five In0.10Ga0.90N quantum wells and an Al0.2Ga0.8N electron-blocking layer; and a dielectric DBR, which is added by electron-beam sputtering and is made from seven pairs of TiO2 and SiO2.
After the bottom mirror and cavity are grown by MOCVD, reactive ion etching forms a mesa structure and plasma treatment defines the aperture’s dimensions. The VCSEL is completed by the addition of an indium tin oxide (ITO) current-spreading layer and a dielectric DBR.
Driven at a duty cycle of 1 percent – 200 ns pulses at a 50 kHz repetition rate –VCSELs with current apertures of 4 µm, 8 µm and 12 µm exhibit threshold current densities of 130-150 kA cm-2. According to the team, this lack of variation in threshold current density with aperture diameter is proof of efficiency of both the current spreading layer and the current confinement.
Cavity-mode broadening is observed below threshold, due to heating effects that would prevent lasing under continuous-wave (CW) injection. However, calculations reveal that halving the thickness of the ITO layer, placing it at a node of the electric field and increasing the reflectivity of the top DBR could slash the current density from 140 kA cm-2 to 10 kA cm-2. “This is a value much closer to the state-of-the-art thresholds reported so far,” writes the EPFL team in its paper.
In the short term, these researchers plan to slightly modify the design of their VCSEL by introducing a shorter p-GaN layer so that the ITO is positioned in a node of the optical field. “Moreover, we will reduce ITO layer thickness by half,” explains Cosendey.“We think that by the combined effects of those two factors, we should reduce the threshold current density by a factor ten, which would eventually enable CW lasing.”
EPFL’s monolithic VCSEL combines a lattice-matched, 41.5 pair Al0.8In0.2N/GaN districted Bragg reflector with a multi-quantum well active region, an electron-blocking layer, an indium tin oxide current-spreading layer and a seven period dielectric top mirror.
This is the next step for the team that is pioneering the development of monolithic blue VCSELs, which could be used for high-resolution printing and bio-sampling. Their latest breakthrough is the creation of VCSELs on "defect-free" epitaxial distributed Bragg reflectors (DBRs) made of lattice-matched InAlN and GaN.
“Here, defect-free means that no further dislocations are introduced during the growth, therefore the dislocation density seen by the quantum wells of the active region is that of the free-standing GaN substrate used for the growth”, says lead author Gatien Cosendey. Hepoints out that the device is less susceptible to over-heating when it is built with epitaxial DBRs, rather than those made from dieletrics.
Another important aspect of this team’s work is the development of a current confinement scheme based on plasma treatment of the surface of p-GaN. Going down this route avoids the requirement for a thin dielectric aperture.
“This prevents diffraction losses at the edges of this dielectric aperture, but increases the growth time – it’s about 20 hours for a 40-pair DBR,” explains Cosendey.
Growth times are much shorter for VCSELs made with a pair of dielectric mirrors, but this approach requires laser lift-off, wafer bonding, high-quality polishing and cavity thickness control.
Although laser lift-off and wafer bonding are widely used in the LED industry, it doesn’t follow that they are simple, high-yield steps for VCSEL fabrication. “Rough surfaces are required in LEDs to improve light extraction efficiency, while very smooth VCSEL cavity surfaces are needed to reduce diffraction losses and thus decrease lasing threshold,” explains Cosendey.
According to him, one of the advantages that VCSELs traditionally have over their edge-emitting lasers is mono-mode emission. This is not the case in the VCSELs made by Nichia and Panasonic that sport a pair of dielectric DBRs – these devices feature long cavities that support several modes.
“On the other hand, UCSB [which also uses a pair of dielectric DBRs] has developed a photoelectrochemical etch process that allows precise control of the cavity length while ensuring good surface morphology,” says Cosendey.
The VCSELs fabricated by the EPFL team feature: A 41.5 pair Al0.8In0.2N/GaN DBR; a pin diode structure, which contains an active region with five In0.10Ga0.90N quantum wells and an Al0.2Ga0.8N electron-blocking layer; and a dielectric DBR, which is added by electron-beam sputtering and is made from seven pairs of TiO2 and SiO2.
After the bottom mirror and cavity are grown by MOCVD, reactive ion etching forms a mesa structure and plasma treatment defines the aperture’s dimensions. The VCSEL is completed by the addition of an indium tin oxide (ITO) current-spreading layer and a dielectric DBR.
Driven at a duty cycle of 1 percent – 200 ns pulses at a 50 kHz repetition rate –VCSELs with current apertures of 4 µm, 8 µm and 12 µm exhibit threshold current densities of 130-150 kA cm-2. According to the team, this lack of variation in threshold current density with aperture diameter is proof of efficiency of both the current spreading layer and the current confinement.
Cavity-mode broadening is observed below threshold, due to heating effects that would prevent lasing under continuous-wave (CW) injection. However, calculations reveal that halving the thickness of the ITO layer, placing it at a node of the electric field and increasing the reflectivity of the top DBR could slash the current density from 140 kA cm-2 to 10 kA cm-2. “This is a value much closer to the state-of-the-art thresholds reported so far,” writes the EPFL team in its paper.
In the short term, these researchers plan to slightly modify the design of their VCSEL by introducing a shorter p-GaN layer so that the ITO is positioned in a node of the optical field. “Moreover, we will reduce ITO layer thickness by half,” explains Cosendey.“We think that by the combined effects of those two factors, we should reduce the threshold current density by a factor ten, which would eventually enable CW lasing.”
EPFL’s monolithic VCSEL combines a lattice-matched, 41.5 pair Al0.8In0.2N/GaN districted Bragg reflector with a multi-quantum well active region, an electron-blocking layer, an indium tin oxide current-spreading layer and a seven period dielectric top mirror.
