InP Offers A Promising Base For Very-short-wavelength QCLs
The quantum cascade laser (QCL) is no longer just a lab curiosity, caged in a cryostat. Room-temperature devices are now commercially available, and they are generating revenue for European firm Alpes Lasers, and for Daylight Solutions and Physical Sciences Inc in the US.
Many of these QCLs are being deployed in gas-sensing tools. These include instruments made by Cascade Technologies of Stirling, Scotland, that reveal traces of methane, ammonia and carbon monoxide at the parts-per-billion level. What s more, this versatile laser can have very different uses, such as high-speed free-space communication.
But QCLs would be even more useful if they could be extended to shorter wavelengths and cover the 2–4 µm spectral range, which is difficult to reach with III-V edge-emitters and VCSELs. This region is rich in molecular absorption lines, and spectroscopy within this domain can expose small organic molecule contaminants within the atmosphere. In addition, the same materials that feature in these QCLs could be used to make ultra-fast optical switches for telecommunication networks if intersubband transitions at 1.55 µm can be realized.
Why turn to II-VIs?
Stretching QCLs to these shorter wavelengths is far from easy, however. Emission is governed by the conduction-band offset between the pair of III-Vs employed in the device, and the commercial twosome of InGaAs and InAlAs, which are grown on InP, is limited to the 7–24 µm range.
This impasse has spurred the exploration of alternative materials, and our group at the City College of New York is going down this road. We ve teamed up with Claire Gmachl s group at Princeton University and recently observed the first quantum cascade electroluminescence from a non-III-V semiconductor – a II-VI wide-bandgap material system that is based on ZnCdMgSe.
We are not the first QCL researchers to enjoy success below 4 µm. Antimonide-based lasers have stretched to 2.75 µm thanks to conduction-band offsets of up to 2.1 eV. However, this material system cannot replicate the performance of longer-wavelength InGaAs/InAlAs QCLs, and it fails to deliver continuous-wave room-temperature emission. Inferiority stems from intervalley electron scattering, which reduces the effective size of the conduction-band offset.
Wide-bandgap II-VIs, and in particular the direct bandgap pair ZnCdSe and ZnCdMgSe, are promising materials for QCLs operating below 4 µm. The materials offer type-I band alignment, a broad bandgap range and a conduction-band offset that can hit 1.1 eV. The only trouble is that well developed, high-quality native bulk substrates do not exist.
So we have turned to III-V substrates as a platform for II-VI growth. This is not a new idea, and in the 1990s GaAs was the popular choice for ZnSe-based structures, which were viewed as a promising material system for blue-green laser diodes. Success relied on getting this material combination to work well together, and demanded careful control of lattice mismatch and the chemical imbalance that is inherent in a heterovalent interface between a II-VI and a III-V compound.
But GaAs isn t the only III-V that provides a suitable platform for II-VI materials. In the late 1990s and early years of this decade we demonstrated that InP can be an ideal basis for the ZnCdMgSe family of compounds. Progress wasn t easy, however, and we had to invest a great deal of time in developing a process for a device-quality II-VI/InP interface.
We produce our material by MBE and our approach appears to deliver higher-quality material than that produced by other methods. Growth takes place in a solid-source MBE Riber 2300P system with two growth chambers. One is dedicated to III-Vs, the other to II-VIs, and ultra-high vacuum modules enable sample transfer.
Cleaning the surface
We strip away our substrate s oxide surface layer prior to growth by heating the InP platform in the III-V chamber under arsenic overpressure. This step is monitored by reflection high-energy electron diffraction (RHEED), which can pick up the switch from an arsenic-stabilized surface to one terminated by indium.
Once the surface is ready, we rapidly drop the substrate temperature by 30 °C and deposit a lattice-matched 0.25 µm thick In0.53Ga0.47As buffer layer. Growth proceeds at 1 µm/h in an arsenic-rich atmosphere to create an atomically smooth III-V surface with group-V termination. This is ideal for initiation of the II-VI layer.
The wafer is then transferred under vacuum to the II-VI chamber, where its top surface is exposed to a zinc flux for a period of 40 s. This step, which is carried out at 200 °C, suppresses In2Se3 and Ga2Se3 formation at the III-V/II-VI interface. These compounds can degrade material quality by forming defects such as stacking faults.
The wafer s temperature is maintained at 200 °C, and a 9 nm thick Zn0.43Cd0.57Se buffer layer is added to promote two-dimensional nucleation. This provides the basis for the growth of epitaxial material with a typical defect density of just 104 cm–2, comparable to the best ZnSe-on-GaAs.
A selenium flux is then applied while the wafer is heated to 300 °C. Growth of the II-VI structure then resumes, with ZnCdSe and ZnCdMgSe deposited at growth rates of 0.30 and 0.85 µm/h, respectively. These layers are grown under selenium-stabilized conditions, which are maintained by ensuring that the selenium to group-II element flux ratio exceeds four. Accurate growth rates are calculated from RHEED intensity oscillation measurements, which can also confirm layer-by-layer growth.
We have refined our process conditions for the deposition of lattice-matched ZnCdSe and ZnCdMgSe layers and quantum wells through the growth of calibration samples (figure 1). Hall-effect measurements have determined the carrier concentrations in n-type layers that were doped with ZnCl2.
This work has enabled us to select lattice-matched Zn0.20Cd0.19Mg0.61Se barriers and Zn0.43Cd0.57Se wells for the quantum cascade emitter structure. They have bandgaps of 2.08 and 3.03 eV, respectively, at 300 K, and produce a conduction-band offset of 0.78 eV.
Having developed this deposition process, we then designed a quantum cascade emitter structure with 20–30 asymmetric coupled quantum-well active regions, which are sandwiched between multilayer injector regions (figure 2). Each injector contains 14 Zn0.43Cd0.57Se and Zn0.20Cd0.19Mg0.61Se layers that are several angströms thick, which means that the device has more than 500 layers. This takes 6–8 hours to grow. To aid the fabrication of such a complex device, we have also produced a simpler structure that just contains several repeats of the asymmetric coupled quantum-well layers separated by simple quaternary barrier layers (see box "Calibration structures for fine tuning").
These calibration structures have helped us to improve our growth process, and we employed these new conditions in a quantum cascade emitter structure. The devices were processed by photolithography, before contacts were added by electron-beam metal deposition. Electroluminescence at 4.8 µm was produced at 78 K – which isn t far away from the design target of 4.5 µm (figure 3) – and persisted up to room temperature.
With proof-of-principle demonstrations behind us, we are now focused on optimizing a structure that should ultimately deliver impressive device characteristics. We will be helped by the National Science Foundation-funded Engineering Research Center at Princeton, known as MIRTHE (Mid-Infrared Technologies for Health and the Environment). These collaborators are actively pursuing the optimization of the materials and device design, which should help us to meet our targets of sharper emission lines and lower turn-on voltages.
One of our primary goals is to design quantum cascade structures emitting below 4 µm. We believe that this is possible by increasing the barrier height in the QCL and then taking full advantage of the 1.12 eV conduction-band offset. Another option at our disposal is to turn to symmetrically strained heterostructures. Once we ve hit our short-wavelength target of less than 4 µm, we will develop an effective waveguide design and finally build a QCL.
View pdf of article