1. Introduction

The research on ultrasmall semiconductor devices has strongly increased in the recent years [1–6], due to the need to realize low consumption devices with ultra-fast response. Moreover, reducing dimensions down to the tens of nanometers allows for quantum size effects to be exploited; at these length scales, energy bands or discrete states, which do not exist in larger structures, are created [7–9]. Quantum cascade lasers (QCLs), which currently exploit quantum size effects in 1D, represent one of the most reliable semiconductor sources of coherent radiation in the terahertz spectral range [10]. They can provide milliwatt optical power levels, but their operation is still confined to cryogenic temperature.

In this work we present an experimental demonstration of electrically driven nanopillars cavities for THz QCLs. This result is a promising step toward the realization of zero-dimensional structures for these active media. As demonstrated in presence of a high magnetic field [11], the 3-dimensional confinement can enable the suppression of non-radiative intersubband transitions or level relaxations, to reduce the laser threshold and increase the operating temperature [12–14].

In the recent years the highest operating temperature for THz QCLs is still slightly increasing with the years and the most recent is 199.5 K [15]. Its rate of increase sensibly slowed down since it reached 150 K, in other words since the thermal energy becomes comparable to the lasing energy. In fact as shown by Belkin et al. [16], a quantum cascade structure in the GaAs/AlGaAs material system would need to have nearly 100% injection efficiency in the upper lasing state and also an extremely efficient extraction from the lower lasing state in order to operate at room temperature. Therefore, room temperature operation of terahertz QCL will probably require the implementation of innovative ideas rather than active region optimization in GaAs structures [17].

Innovative materials systems for the cascading structures offer one promising path to higher temperatures. In particular, GaN/AlGaN is particularly appealing due to the large phonon energy (90 meV), which provides a large barrier for thermally activated optical phonon scatterings [18] and at the same time a fast depopulation scheme for the carriers in the lower lasing state. However, high quality GaN/AlGaN material growth is currently difficult to achieve and the best THz lasers are still obtained using the GaAs/AlGaAs material system. Another approach is based on the reduction of the LO phonon scattering thanks to the realization of 0-dimensional structures. Using InGaAs self-assembled quantum dots on GaAs, Zibik et al. [19] already demonstrated long intradot lifetimes using terahertz pump-probe spectroscopy. Lifetimes as long as 1.5 ns were measured in this case in contrast to the quantum-well based devices, where scattering times are of the order of ps. Even though these results are really encouraging and prove that carrier relaxation can be properly engineered in 0 dimensional structures, it remains unclear if self-assembled quantum dots could be successfully incorporated in complex multi-layer structures as in the case of THz QCLs [17].

Liverini et al. [20] have recently reported on the integration of InAs dot with AlInAs/InGaAs mid-infrared quantum cascade active regions. In this case InAs dots elongate to form quantum dashes. They demonstrate narrow electroluminescence at 150 meV up to room temperature; the emission is associated to a transition from the first excited state of the dashes, related to the lateral confinement provided by their width, to the ground state of the dashes.

One possibility to realize truly 0-dimensional structures is to define the dots lithographically after the growth of a standard quantum-cascade laser structure and to etch a dense array of pillars with dimension of the order of hundreds of nanometers in order to obtain electron confinement. In this case the quality of the epitaxial growth is preserved. Until this work, this proposed fabrication method had never been applied in the case of THz quantum cascade structures, due to the need to complex, high-end technology procedures. To achieve nm size device features using conventional processing requires e-beam lithography. However, due to the serial nature of defining the pattern with e-beam, writing high resolution structures over the large areas required for commercial applications is prohibitively expensive. In addition, demanding procedures of planarization and deposition of low-loss polymeric insulation layers would be required to enable the electrical biasing of the whole structure in parallel [13,21]

In this work we present the successful use of a controlled colloidal mask deposition to realize a large-area dense array of nanopillars of quantum cascade structures with an average radius <200 nm. In this case the potential in the pillars, due to the Fermi level pinning, leads to an estimated splitting of the energy of the order of 5 meV. For this size of the pillars the energy separation due to the lateral confinement is expected to be larger than the typical broadening of 2-3 meV of the THz transition in QCL. However a further reduction of the diameter of the pillars combined with an increase of the doping of the active region is needed to achieve the energy separation of ~30 meV, demonstrated in presence of magnetic field [11] for high temperature performances of THz laser. With the nanospheres deposition technique proposed here it is possible to have pillars with dimensions between 40 and 500 nm, distributed at controlled distances varying between 2 to 10 particle diameters.

The fabrication process proposed here is cost-effective and permits the realization of nanostructures avoiding for expensive electron beam lithographic steps and any polymeric layer deposition for planarization. Thanks to gold-gold thermocompression we achieve an efficient electrical injection in arrays of nanopillars over 1mm² surface. Moreover THz emission as function of the bias is measured for such structures, with clear emission at the expected frequency of the laser.

2. Experimental

Sample preparation

The active medium used in this work is a THz QCL based on a four well resonant phonon scheme in a GaAs/AlGaAs material system, operating around 3 THz (100µm) [22]. The active medium stack consists of a repetition of 28 periods for a total height of 2 μm. The designed device cavity consists of a wide array of nanopillars embedded in a double metal waveguide, where the active region is sandwiched between two metal layers. As presented in reference [23] the double metal waveguide can be engineered to have low losses even for reduced active region thickness, as in our case. Moreover, the reduced height of the active region is needed to have an aspect ratio for the pillars (height/diameter) not higher than 10, so as to not affect mechanical stability of the structure. The preparation of the samples, before the etching of the nanopillars, consists of the standard steps of a double metal process [24]. The original substrate is gold coated on top of the epi-growth and wafer bonded by gold-gold thermocompression to a metalized host substrate, polished down and selectively etched down to the active region. We then use a colloidal mask [25] for the etching, to have a rapid and parallel process compatible with large scale production.

The deposition of this mask exploits the self-assembly of latex nanoparticles at a liquid-liquid interface (SALI) prior to the deposition [27]. Thanks to this technique, we can control independently the size and the separation of nanoscale features in a non close-packed pattern driven by long-range electrostatic dipolar repulsion between the particles at water/n-hexane interface (Fig. 1(a).

 

Fig. 1 a) SEM image of the starting substrate after the SALI deposition of the latex nanospheres. b) Fourier Transform of the deposited pattern.

Download Full Size | PDF

In this paper we used 220 nm ± 10 nm diameter particles, obtaining an average nearest neighbor separation of 870 nm ± 10 nm corresponding to a surface coverage of 5.5%. The Fourier transform of the deposited pattern presented in Fig. 1(b) demonstrates high hexagonal order.

The nanospheres were used as hard mask for the etching of the nanopillars down to the bottom metal layer, throughout the whole thickness of the active medium; inductively-coupled plasma dry-etching was carried out using Cl₂:N₂ chemistry (Fig. 2 top panel). Latex nanosphere residues are removed after the etching with UV ozone cleaning. The electrical contact of the structure is realized by vertical evaporation of a Ti/Au layer without any need of additional lithography steps, in fact thanks to the small undercut in etching of the pillar profile, the short circuit of the active region is prevented (Fig. 2 bottom panel). After the contact deposition, to enable electrical injection, a gold-coated doped substrate is connected on the top of the pillars via gold-gold thermocompression. The bonding is performed for 15 minutes at a temperature of 300◦ C and a pressure of 2 MPa. As is shown in Fig. 3, the mechanical stability of the nanopillars is not affected by this procedure. The electrical contact is made through the top substrate, where an ohmic contact is deposited.

 

Fig. 2 SEM image of the sample after the etching of the nanopillars using the nanospheres as hard mask. On the top of the pillars some latex residuals are still visible. Bottom panel: closer view of the nanopillars.

Download Full Size | PDF

 

Fig. 3 SEM image of the sample after the gold-gold thermocompression on the top of the pillars. The mechanical stability of the nanopillars is preserved throughout this procedure

Download Full Size | PDF

Sample characterization

The light-voltage-current curves at 10 K of the nanopillars array are presented in Fig. 4. This sample is compared with a standard double metal ridge process of the same growth. The two voltage versus current curves overlap considering a surface coverage of the pillars of 7%. This last value is comparable to the 5.5% evaluated from the Fourier analysis of the distribution of nanospheres (Fig. 1(b)).

 

Fig. 4 Measured voltage-light versus current characteristic at 10 K, in pulsed mode, for the nanopillars array (red line) and a standard ridge waveguide (blue line). The current scale relates to the nanopillars device, while the current density is valid for the two samples. The dissipated electrical power vs current is shown as a dashed line. Colored circles correspond to the bias where the light emission has been spectrally studied in Fig. 5.

Download Full Size | PDF

Unfortunately, while in the standard waveguide a clear laser threshold is observed, for the nanopillars sample only electroluminescence emission is reported.

Spectral measurements of the light emitted from the nanopillars are shown in Fig. 5(a), for three different voltages with a spectral resolution of 0.6 meV. The emission spectra are quite complex but two main features are observed; one at 10 meV, corresponding to the expected laser frequency, more evident with the increasing bias, and the second at 40 meV, present also for lower voltage, corresponding to transitions between higher energy states.

 

Fig. 5 a) Top panel: calculated intersubband electroluminescence in the case of a lossy cavity. Bottom panel: measured spectra of the light emission from the nanopillars cavity for different voltages. b) Conduction band diagram of a period of the active region at an average field of 7.5kV/cm. The squared moduli of the relevant envelop wave functions are shown. Inset: Calculated potential and first five eigenstates in the conduction band for a nanopillar of bulk GaAs of 160 nm diameter. The splitting in energy is of ~5 meV

Download Full Size | PDF

The calculated intersubband electroluminescence curve assuming the presence of a lossy cavity and neglecting the effect of the nanopillar is reported in the top panel of Fig. 5(a). The labeling of the states is reported in Fig. 5(b), where we present the calculated squared moduli of the relevant envelop wave functions, neglecting the 3-dimensional confinement of the electrons. The main characteristics are in agreement with the experimental data apart from the region around 30 meV, where probably additional effects connected with the phonon absorption are involved [27]. In the inset of Fig. 5(b) we show the calculated energy levels in a single GaAs nanopillar using the freely available software “nextnano”, neglecting the embedded cascade structure; we calculated an energy difference of ~5 meV. We believe that a calculation of the eigenfunction including the cascade potential and the lateral nanopillars confinement is needed to gain a full understanding of the THz spectrum of such structures and of the electric transport. This calculation goes beyond the scope of this paper.

3. Conclusion

In conclusion in this work we have shown the successful realization of nanopillars array for THz quantum cascade active region. These structures could be electrically driven despite of their unconnected geometry. We presented electrically driven THz emission for this structure, tuning with the bias. This same technique using nanospheres smaller than 100 nm can be applied to THz QCL active region of thickness lower than 2 μm to have stronger effect of the lateral confinement.