Accessing the mid-infrared and beyond

Merging the quantum cascade laser and transistor promises new applications involving mid-infrared wavelengths through to terahertz frequencies

BY JOHN DALLESASSE AND KANUO CHEN FROM THE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN

It is far from easy to create a coherent source of emission between mid-infrared wavelengths and terahertz frequencies. But if success is possible, great rewards may follow − if these new devices could be  commercially viable, they could serve new applications and open the door to new markets.

Many of the opportunities for frequency-tuneable sources in the mid-infrared and beyond are associated with identification of chemical species. Numerous substances have distinct molecular resonances in this spectral range, so frequency-tuneable sources could be used for chemical process monitoring, automotive applications, biomedical imaging and histopathology, chemical and explosive detection, greenhouse gas monitoring, ground and waste water monitoring, homeland security, and a host of other applications that rely on detecting the presence and quantity of a given chemical.

A frequency-tuneable source can be used to detect chemical species by sweeping its frequency so that it excites a molecular vibration. Hitting a frequency that causes molecules to resonate leads to the absorption of light, and this can be picked up by the dip in the response of a detector. To detect with the best signal-to-noise ratios, high-intensity, spectrally narrow, tuneable source are required.

Such sources are also preferred for imaging, with power changes in a detector array enabling spatial localization of compounds in various samples, including biological specimens.  When imaging techniques with mid-infrared lasers identify cancerous cells, this approach can realise greater speed, accuracy, and resolution than conventional methods. 

Another application for optical sources emitting from the mid-infrared to terahertz frequencies is point-to-point communication links. To serve this, sources must be independently modulated in amplitude and frequency at high data rates.

For all the applications outlined above, the ideal source is compact and highly efficient, so that a battery can power it. Operation at typical ambient temperatures, without the need for cooling is also essential for widespread uptake in portable handheld systems, or in systems that are field deployed for environmental monitoring.

QCLs: Pros and cons

A great breakthrough that has paved the way to the realisation of practical sources in the mid-infrared and beyond came in 1994, with the development of the first quantum cascade laser (QCL).  This two-terminal unipolar n⁺-i-n device can generate photons with desired wavelengths if layer thicknesses are carefully engineered, and the applied voltage adjusted. Changing the voltage modifies the separation between the energy states, so it determines the emission wavelength and influences the rate at which these transitions occur, because the electron wavefunction in the cascade region is field dependent. 

Thanks to the wavelength of the QCL being only loosely constrained by material parameters "“ and governed by the engineering of the layer thicknesses for a specific bias voltage "“ this class of laser has demonstrated emission over a very broad wavelength range.

Tuning emission through adjustments in voltage is an attractive feature of the QCL, but it comes at a price: changes in voltage also influence the lifetime of the photons. This means that when there are adjustments to the electric field across the cascade region, this changes both the shape of the electron wavefunction and the state energy, and because the QCL is two-terminal device, changes in output power follow. So in short, adjusting the applied voltage changes both the emission wavelength and the output power. Clearly this is undesirable, because in many applications it is preferable to fix the output power while adjusting the wavelength.

It is possible to see how to improve the QCL by examining the operation of the bipolar junction transistor (BJT).  When formed as a n-p-n structure, the electrons in this class of device are injected from a forward biased n-p emitter-base junction into the base, before they diffuse across to the collector and are swept away by the field in the reverse-biased p-n base-collector junction. This chain of events occurs as the carriers take a random walk from the quasi-neutral base into the collector junction field region.  The "˜magic' of the BJT stems from the use of a small hole current into the base contact, which controls a large electron current sourced by the emitter and captured by the collector. 

Our team at the University of Illinois at Urbana-Champaign has combined the attributes of the BJT with those of the QCL to form an innovative device that provides independent control of field across, and current through, the quantum transition region. With this novel, hybrid design, which we refer to as the transistor-injected (TI) QCL, the electric field controls wavelength while the current controls the laser power.

The TI-QCL is similar in design to the HBT (see Figure 1). Key differences are the addition of layers on the top and bottom of the device to provide optical confinement, the inclusion of the cascade region, and some variations in layer thicknesses and doping levels.

Figure 1: The transistor-injected quantum cascade laser is a three-terminal device that is designed to obtain coherent radiation from mid-infrared wavelengths through terahertz frequencies. This device utilizes the transistor effect and minority carrier injection to enhance the performance of the quantum cascade laser.  Gold arrows show the electron flow through the structure, which is controlled by the grey hole flow into the p-type base.

It is possible to fabricate our novel laser on a range of substrates. It can be grown on an n-type conductive substrate, with the collector contact made to the back of the wafer (see Figure 2, which shows a simplified epitaxial layer structure); and it can be formed on an insulating substrate, with all contacts made to the top surface. What's more, thanks to the similarity of the TI-QCL and the HBT, in terms of design and processing steps, it should be possible to produce this device in a commercial GaAs IC foundry.

Figure 2: An example of a typical epitaxial structure for the TI-QCL. The collector contains low-refractive-index layers for optical-mode confinement.  Within the base-collector junction is the quantum cascade gain region.  Above the quantum transition region is a p-type, lightly doped base.  Graded doping that increases near the emitter-base junction minimises free carrier absorption.  The emitter itself is formed from InGaP to ensure a high emitter injection efficiency, and within the overall emitter structure are low refractive index layers to provide optical mode confinement.  The entire structure is capped with a heavily doped n-type contact layer.

The introduction of a third terminal offers plenty of freedom, in terms of device operation (see Figure 3 for a band structure diagram). Frequency modulation at fixed output power is possible by dithering the base-collector bias voltage while using a fixed emitter-base bias − adjustments to the voltage alter the slope of the bands, and thus the transition energy. 

Figure 3: The band structure of the TI-QCL showing the emitter (E), base (B), and collector (C) contacts.  In this n-p-n device, electrons are injected into the p-type base by the forward biased emitter-base junction and then diffuse across the base, before they are swept away by the field region in the reverse-biased base-collector junction.  Within the field region of the base-collector junction, electrons travel through a staircase potential in discrete steps, emitting sub bandgap photons at a wavelength established by the design of the quantum wells and barriers as well as the applied electric field "ε" set by the base-collector bias.

Varying frequency while maintaining a constant output power is extremely useful for identifying chemical species. By sweeping frequency back and forth across a molecular absorption line at a fixed output power, enhanced signal-to-noise performance is possible, enabling superior detection of low concentrations.

Communications applications may also be targeted with our TI-QCL, because frequency modulation has the potential to be very fast, as it is based upon the electric field change in the structure. Amplitude modulation is also possible, due to modulation of the emitter-base bias. In this case, modulation rates are limited by transit times across the base and through the quantum transition region, so are slower than those resulting from frequency modulation. In addition, frequency and amplitude can be modulated independently, by adopting a common base configuration that provides separate control of the emitter-base and base-collector junctions.

A further benefit of our TI-QCL is its low internal loss, which stems from the n-p-n structure and the fundamentally different method of current injection. Losses due to free carrier absorption are minimal, thanks to the absence of doping from the cascade region and much lighter doping levels near the cascade region. Further improvements to internal loss result from the creation of a depletion region around the QCL structure in the regions of high optical field intensity where absorption is greatest.

Due to these refinements, there is the promise of a lower laser threshold current, an increased differential quantum efficiency and a hike in wall plug efficiencies. An increase in the latter is highly valued, because this is critical for portability and energy-sensitive applications.

Modelling the device

Device modelling is not trivial. To design a TI-QCL that works as intended requires modelling of the electron wavefunctions and energy levels under various electric fields, as well as modelling of optical modes and simplified modelling of the electrical transport characteristics.

Modelling of the electron wavefunctions and energy levels offers insights into laser properties, such as the gain characteristic. Such calculations can reveal the voltage swing needed to produce a desired change in output frequency. Meanwhile, waveguide modelling allows engineering of the optical mode and an estimation of internal loss based upon free-carrier absorption. It is important to carry out this modelling for both the engineering of the epitaxial layer stack, and for determining the widths of the emitter and base mesa, which control lateral mode confinement. The last aspect of modelling, that of the electrical transport characteristics, is performed to enable a prediction of the current-voltage characteristic of the device. 

So far we have only published results from device modelling, due to an initial focus on design and concept validation. To prove that our device is capable of fulfilling its promise, is must be capable of generating stimulated emission, which requires a population inversion "“ the presence of more carriers in the excited state than the ground state.

Calculations of electron wavefunctions have enabled us to determine state transition rates, and from this state population densities. The good news is that the population densities for the two most important states for lasing action converge to values that will allow a population inversion (see Figure 4).

Figure 4: Modelling results reveal the electron sheet density in the two most important states for lasing action "“ the upper and lower lasing states.  The simulation shows convergence toward an electron population that is higher in the upper lasing state than the lower lasing state "“ a necessary condition for laser operation.

We have also undertaken waveguide modelling for our laser design, determining the expected mode profile (see Figure 5). Further data showing device operation is being prepared and will be soon be published.

Figure 5: Two-dimensional optical waveguide modelling of the TI-QCL showing mode confinement provided by the upper and lower cladding layers, as well as the emitter ridge.  The confinement factor and the layer doping levels are used to provide an estimate of the optical losses due to free carrier absorption.

The great potential of our device, and the encouraging preliminary results from our modelling efforts, are providing motivation for us to continue to pursue the development of a TI-QCL. Supported by funding from National Science Foundation, we will focus on device design refinement, fabrication, and characterization, with the primary goal of achieving a performance suitable for portable systems. This system-enabling work will focus on design iterations to improve device performance, demonstrate operating wavelengths that extend through terahertz frequencies, show extended tuning range, and expand the ability to modulate at high speeds. 

Beyond this initial device-level work, we will explore, in collaboration with partners, the use of our novel QCL in specific applications such as chemical sensing and biomedical imaging.  We look forward to a future where these devices are broadly deployed to help maintain air and drinking water quality, ensure process efficiency in chemical and semiconductor manufacturing, and help provide early detection of illnesses such as cancer.