1. Introduction

The terahertz (THz) spectral region is very attractive for a variety of applications such as real time imaging [1], heterodyne detection [2], sensing or spectroscopy [3]. These applications exploit the fact that the energy of rotational and vibrational transitions of molecules lies in this part of the electromagnetic spectrum. This leads to characteristic absorption features and allows a simple distinction between different molecules.

These applications require narrow band sources with reasonably high output powers. The preferred devices are THz quantum cascade lasers (QCLs) [4]. Due to their designable gain, it is possible to span a frequency range between 1.2 and 4.9 THz [5, 1] with maximum operating temperatures up to 178 K [6]. As the gain of THz-QCLs is inhomogeneously broadened and has a typical full-width half-maximum (FWHM) of 130 GHz [7], they normally lase in multi-mode. Single-mode operation at a defined frequency is typically achieved by one-dimensional (1D) distributed feedback (DFB) structures [8, 9] or microdisk resonators [10, 11].

Another concept for the frequency control of THz-QCLs is based on two-dimensional (2D) photonic crystals (PhC). The advantage of PhC is their designable dispersion, it allows for a manipulation of the flow of light on a sub-wavelength scale. It is perfectly suited for THz-QCLs due to their unipolar nature, the large wavelength and the in-plane emission. The unipolar nature of QCLs makes them insensitive to surface leakage currents, the large wavelength of THz-QCLs sets high tolerances for the precision and roughness, and the in-plane emission of light allows for the use of a standard planar processing technology. The basic concept uses 2D PhC-mirrors for the frequency selection, experimental results for two different waveguide configurations have been presented in Ref. [12] and [13]. These lasers have a central active region which is unstructured and provides the necessary gain for the lasing operation. The PhC surrounds the core and is used as a frequency selective mirror. A change in the period of the PhC shifts the spectral position of the bandgap and allows thereby for a simple control of the emission frequency. The PhC is used here only as a purely passive component, which means it has no gain and does not support laser modes on its own. This approach combines a reasonable frequency selection with a simple processing.

A major goal is the realization of active PhCs which are based on high-index, sub-wavelength and free-standing pillars providing the optical gain. These monolithic structures have a lot of advantages compared to the standard resonator geometries which will lead to versatile devices for future applications. A further miniaturization of the devices is achievable as the gain region and the frequency selective part need not to be spatially separated. The frequency control is improved as sharply defined bands instead of relatively broad PhC bandgaps are used. Additionally, a strong gain enhancement for active 2D-PhCs has been predicted theoretically [14]. However, the fabrication of these free-standing structures remains a challenging task. In a first experiment [15], Zhang and co-workers have realized active PhCs based on isolated pillars which have been embedded in a benzocyclobutene host. This approach allows for a simpler processing compared to free-standing pillars but reduces the refractive index contrast of the PhC and thereby the modal confinement. These devices have therefore not been showing single-mode emission in general. However, the temperature performance and the threshold current have been improved in comparison to standard Fabry-Perot (FP) resonators.

Therefore we have designed and fabricated an active 2D-PhC laser which is entirely made of semiconductor nanostructures providing gain. The structure consists of free-standing THz-QCL micropillars surrounded by air. The PhC is fabricated by reactive ion etching of a THz quantum cascade active layer structure. Instead of embedding the pillars in an insulator host, we use a thermo compression bonding technique to form the waveguide and have free-standing pillars at the same time. This leads to an very large index contrast between the active pillars and the surrounding medium. Therefore, the optical mode has a strong overlap with the pillars. This allows us to fabricate devices with low filling factors which reduces the total power consumption. The two major differences of this approach compared to previous work [15] are the open geometry and the excellent frequency control. The open laser resonators would allow their use for various applications such as gas sensors. The excellent frequency control allows for designing stable single-mode emitters in a wide range. A detailed description is given in the following section.

2. Design and fabrication of the photonic crystals

The active region of the THz-QCL used is based on a GaAs/Al_0.15Ga_0.85As heterostructure grown by molecular beam epitaxy. The lower laser level is depopulated by longitudinal optical phonon scattering [16]. The growth sequence is 9.2/3.0/15.5/4.1/6.6/2.7/8.0/5.5 nm where the barriers are marked with bold letters and the underlined well is homogeneously doped to 8 × 10¹⁵ cm^-3. It is designed to lase around 2.7 THz. The active region has a thickness of 15 μm and is similar to the one published in Ref. [17].

The active rods are arranged in a hexagonal lattice, every single rod consists of the GaAs/Al_0.15Ga_0.85As heterostructure. The resulting 2D-PhC has complete bandgaps for TM-modes [18]. The calculated bandstructure for an infinite PhC, using the MIT Photonic Bands-package [19], is shown in the inset of Fig. 1(a). The parameters for the calculation are: a dielectric function ε_r of 12.25 for the rods and a ratio r/a of 0.3 where r is the radius of the rods and a the period of the crystal. The first bandgap for TM-modes spans from 0.22 to 0.31 in terms of the normalized frequency f · a/c where f is the frequency and c the speed of light. Especially interesting for laser resonators are the flat-band regions at the band-edges. These regions correspond to extremely low group velocities and are predicted to give a strong gain-enhancement due to this anomaly [20]. The large contrast in the refractive index between the active pillars and the surrounding air confines the mode inside the pillars strongly. Our simulations show that 95 % of the energy are in the pillars.

 

Fig. 1. A schematic of the real device and the mode evolution inside the resonator. (a) The devices are built up by a defect-free 2D-PhC which is embedded in a double-metal waveguide, the active pillars provide the required gain. The inset is showing a calculated bandstructure of the ideal PhC. The pillar-based PhCs are typically showing bandgaps for TM-modes, the first one spans from 0.22 to 0.31 [fa/c] in our structure. (b) FDTD-results of the mode evolution as a function of time. The gaussian broad band source is placed in the central pillar, it has a center frequency of 0.18 [fa/c] and a spectral width of 0.1 [fa/c]. The gain maximum of the pillars lies at 0.24 [fa/c] and has a FWHM of 0.03 [fa/c], the loss peaks at 0.24 [fa/c] with a FWHM of 0.09 [fa/c]. Nevertheless, the surviving laser mode lies at 0.215 [fa/c] which corresponds to the K-point.

Download Full Size | PDF

To verify if a real device with a finite size can be described using an infinitely large PhC, we have performed finite-difference time-domain (FDTD) calculations on a device consisting of 37 pillars arranged in a hexagonal lattice, using the MIT Electromagnetic Equation Propagation-package [21]. There is no additional resonator incorporated. The ratio r/a is kept constant in respect to the previous simulation, but we have added an imaginary part to ε_r to be able to include the gain and loss of the real device. It is predicted theoretically that 2D-PhCs allow for lasing without the need of an additional cavity [22]. The time evolution of our 2D-PhC is studied after the system is kicked with a broad-band gaussian source, results for different times are shown in Fig. 1(b). The FDTD-calculations are showing the expected behavior, only the mode corresponding to the K-point in the infinite PhC survives, almost independently of the central frequency of the gaussian pulse and without the need for an external resonator.

All devices used in the experiments are embedded in a double-metal waveguide which has a modal confinement of almost 100 % in the vertical direction [23]. The waveguide surpresses any out-of-plane scattering, this reduces the overall losses and prevents the bandgap narrowing which is typical for dielectric slab waveguides [18]. The lasing modes are defined by the PhC itself, as devices with the same period are showing the same stable single-mode emission independently of the device sizes. We have varied the period of the PhC from 22.2 to 35.5 μm, this results in pillar diameters of 12 to 21 μm. It has to be pointed out here that the pillars are smaller than the emission wavelength and that the devices are not lasing on a central defect. The pillars have to provide the lateral confinement and the optical gain at the same time.

The processing of the actual devices consists of sequential metal deposition, etching and thermo compression bonding steps. As for a typical double-metal processing, the first steps are a gold-gold thermo compression bonding and a substrate removal by polishing and selective wet-chemical etching [24]. Afterwards the PhC can be etched through the entire active region by reactive ion etching. This guarantees steep and smooth sidewalls for the pillars. To form the waveguide a second thermo compression bonding step is required, a n⁺ wafer with a 400 nm thick gold layer is bonded onto the pillars. The n⁺ substrate is thinned down by polishing and etching to 10 μm. This n⁺ layer is only used to support the thin gold layer to allow for the wire bonding. An ohmic top contact is deposited onto the n⁺ layer and is used as a self-aligned etchmask for the remaining n⁺ layer and the 400 nm gold layer underneath it. Thereby, only the pillars below the top contact are contacted when the waveguide is formed. A schematic of the resonator is shown in Fig. 1(a).

 

Fig. 2. Spectra of a FP resonator and of 2D-PhC lasers with a period of 26.6 μm. (a) The reference FP spectrum is showing multi-mode emission around 2.6 to 2.7 THz due to the inhomogeneously broadened gain region and has a field dependence. (b) The PhC-based device with a period of 26.6 μm shows single-mode emission at 2.56 THz. The single-mode emission is independent of the driving conditions or the device size, it is defined by the PhC itself.

Download Full Size | PDF

3. Experimental results and discussion

To characterize the active region we have first fabricated standard FP resonators with the dimensions 1000 by 120 μm. A multi-mode emission between 2.6 and 2.7 THz is observed, shown in Fig. 2(a). The very strong optical confinement of the active PhCs allows us to realize cavities with a side length of only 120 μm, this corresponds to the free space emission wavelength or only 5 periods of the PhC. Together with the broad gain region of THz-QCLs we are able to achieve a wide tuning range. The devices with a period of 26.6 μm are lasing at 2.56 THz, shown in Fig. 2(b), which is close to the gain maximum of the material. Changing the period to 31.1 μm shifts the emission to 2.25 THz, shown in Fig. 3(a). A possible tuning of 400 GHz seems to be the typical value for THz-QCLs with pillar based PhCs [15]. The frequency range is significantly wider than the typical gain width of 130 GHz [7]. This is a clear evidence for the predicted gain enhancement in this type of structures and shows their potential. The total output power is comparable to other THz microcavities [10, 11]. However, it is reduced compared to typical Fabry-Perot resonators which have much larger dimensions [25].

The lasing frequency of the 2D-PhC fits nicely to the theoretical predictions, shown in Fig. 3(b). The PhC lasers are showing a stable single-mode emission under all driving conditions and independent of the device size. As already expected from our FDTD-calculations, the devices are lasing in the flat-band region at the K-point. It can be seen clearly that the lasing mode is fixed to the PhC, independently of the driving conditions or the period. The small deviations between the simulation and the experiments are caused by processing imperfections and the uncertainty of the refractive index of the active region.

 

Fig. 3. Spectra of the 2D-PhC with a period of 31.1 μm and normalized spectra of the 2D-PhCs. (a) The PhC-based device with a period of 31.1 μm shows a stable single-mode emission at 2.25 THz independently of the applied field. (b) The devices are lasing at the K-point in the PhC bandstructure. The gray shaded area shows the first bandgap. The deviations are caused by processing imperfections and uncertainties in the refractive index of the active region.

Download Full Size | PDF

Devices with a period of 35.5 μm do not show any lasing operation. For this period the gain region of the QCL overlaps with higher bands of the PhC. This reduces the overlap of the optical mode with the pillars. Devices with a smaller period of 22.2 μm are also not lasing. For these lasers the lowest band of the PhC would correspond to an emission frequency of 3.2 THz. This frequency requires an alignment field which lies beyond the negative differential resistance region of the QCL and therefore outside the stable operating conditions.

The excellent confinement of the mode in the vertical and lateral directions allows us to realize devices with a filling factor of only 33 %, therefore $\genfrac{}{}{0.1ex}{}{2}{3}$ of the waveguide is filled with air. This makes the devices ideal for gas sensing. The emission frequency is independent of the driving current, it is only defined by the PhC. The pillars are free-standing in air, therefore an injection of a gas into the chamber changes the refractive index of the surrounding medium and thereby shifts the optical mode of the laser. The change in refractive index can be significant if the laser line is close to a gas resonance. Even a change of the refractive index of 1 % leads to a shift of the laser mode by 0.3 GHz.

4. Conclusion

In conclusion, we have designed and processed active PhC lasers based on a THz-QCL active region. The devices are built up only by the PhC, there is no central defect incorporated. Lasing is obtained at the K-point in the lowest band. The strong refractive index contrast between the rods and the air leads to a 95 % confinement of the mode-energy inside the active region. A double-metal waveguide is used for the vertical confinement, therefore no mode leakage through the surface or into the substrate is possible. This concept allows us to achieve stable-single mode operation under all driving conditions and for all device sizes. The devices can be tuned more than 400 GHz by varying the period of the PhCs. The strong optical confinemt, in lateral and vertical direction, allows us to use a PhC with a filling factor of only $\genfrac{}{}{0.1ex}{}{1}{3}$.

This work was partly supported by the Austrian Scientific Fund FWF (SFB-ADLIS, DK CoQuS), the Austrian Nano Initiative project (PLATON), the EC (POISE) and the Society for Microelectronics (GMe, Austria).