1. Introduction

Terahertz quantum cascade laser (QCL) emission has been demonstrated over a wide range of frequencies, from 0.84–5.0THz [1]. This has been possible due to the intersubband nature of the optical transition and band structure engineering. The ability to precisely control the band structure through the entire active region (AR) has lead to multi-colour and broad-band emission in mid-infrared QCLs [2, 3].

Extending multi-frequency emission to terahertz QCLs could find applications in chemical identification in the security and pharmaceutical industries, amongst others, where it has been shown that many important compounds have distinct spectral features [4, 5, 6]. The spectral fingerprint of substances can be used for identification by employing differential absorption spectroscopy; the absorption at two frequencies is compared, with only one of the frequencies being absorbed by the substance. Though this could be achieved with two separate lasers, monolithic integration of the two colours into the same active region greatly simplifies the optical arrangement and driving electronics.

To date, there has been only one demonstration of electrically switchable two-colour emission in a THz QCL [7]; the method used was selective injection into a large well design. The states of the large well that received injection from the mini-band changed with electrical bias, injecting into a higher excited state increased the emission frequency. This design, however, required a magnetic field to achieve gain. The method we use here is the heterogeneous cascade, used for mid-infrared multi-colour and broad-band QCLs. This has the benefit of being able to optimise designs independently before being included in the same structure.

2. Fabrication

The heterogeneous cascade, formed by incorporating different active regions (ARs) into the same waveguide structure, consists of ARs based on the 2.9THz design by Barbieri et.al. [8]. We have shown previously [9] that it is possible to systematically manipulate the frequency of this design by simply altering the thickness of each well and barrier in the AR. The two ARs used here are one thinned by 5% from the Barbieri design, we shall call ‘A’ and one thickened by 5%, we shall call ‘B’, producing frequencies of 3.0 THz and 2.6 THz respectively. Wafers were grown on semi-insulating GaAs substrates by solid source molecular beam epitaxy [10]. After an etch-stop layer, 700nm of GaAs doped at 1×10 ¹⁸cm^-3 was deposited, then 45 periods of the lower AR. On this was grown a 250nm bulk GaAs spacer layer doped at 1×10 ¹⁵cm^-3, to allow for field adjustment, followed by 45 periods of the upper AR, finished with 80nm of GaAs doped at 5×10¹⁸cm^-3. Sample V428 has the 3.0 THz AR, ‘A’, on the substrate side of the structure and the 2.6 THz AR, ‘B’, assigned to the surface side; sample V429 has these positions reversed.

It was decided to grow two wafers with different arrangements of ARs and include a separator layer because mid-infrared QCLs fabricated in the GaAs/AlGaAs material system can exhibit domain formation [11], potentially distorting the electric field profile through the device. It is expected that if domain formation was an issue then the two wafers would have significantly different electrical and optical characteristics.

 

Fig. 1. Pulsed light-current-voltage characteristics for wafer V428 (left) and V429 (right) operating at a range of temperatures. The insets show the ordering of ARs when processed in the double metal configuration.

Download Full Size | PDF

3. Results

Portions from each wafer were Au-Au thermo-compression bonded to n⁺ GaAs substrates and processed into double metal waveguides 80µm wide and ~3mm in length. These ridges were indium soldered to copper blocks and mounted in a Janis ST-300 cryostat. The light-current-voltage (LIV) characteristics for pulsed operation at a 5% duty cycle are shown in figure 1 for V428 and V429. THz radiation was collected with an F/1 off-axis parabolic mirror and focused onto a Golay cell for detection. For each wafer two distinct peaks in the L-I curve are clearly visible, the first ceases completely before the second reaches threshold.

In pulsed mode, at 4K, the first colour of V428 reaches laser threshold at 107Acm^-2 and ceases lasing at 236Acm^-2, the second colour appears at 242Acm^-2 and comes to an end at 366Acm^-2. Under the same conditions, the first colour of V429 operates from 79 to 199Acm ^-2 and the second peak appears between 209 and 305Acm^-2.

Figure 2 shows emission spectra for the devices, obtained in CW operation. This clearly shows that for both wafers the first, larger peak in L-I corresponds to the lower frequency, AR B, the laser action then ceases at some intermediate current and a second, weaker, peak corresponds to the higher frequency of AR A. Devices from wafer V428 lase at 2.55THz and 2.93 THz, whereas those from V429 lase at 2.68THz and 3.12 THz. The difference in emission characteristics between wafers may be partly accounted for by a slight variation in growth rates, the thickness of each wafer, measured by high resolution X-ray diffraction, was found to have a 1% difference, corresponding to a 0.04THz shift.

The temperature performance of each frequency differs, figure 1 shows that the 2.55THz mode of V428 works to 91K, whereas the 2.93THz mode only works to just over 60K. Devices from wafer V429 show similar performance; the 2.68 THz mode working to 86K and the 3.12THz mode stops just before 60K. The CW temperature performances of AR A and B are 31K and 70 K, for V428, and 37K and 78K for V429, respectively.

There are some interesting common features in the L-I-Vs of Fig. 1. For each wafer, the sequence in which the frequencies reach threshold and the relative magnitude of each colour is the same: the position of ARs within the wafer has no effect on the sequence in which each AR reaches threshold. A second point to note is that, for both wafers, during laser action of the higher frequency there is a rapid increase in the voltage dropped across the device with no effect on the power or frequency of the emission. This happens at around 300Acm^-2 for V428 and between 200 and 250Acm^-2 for V429. Since the power and frequency of the laser are unaffected by a jump in bias, this suggests that the extra voltage is not dropped across the AR involved in laser action and that extra bias is dropped across AR B, that has finished lasing. This also implies that the two sections of the device have quite different electric fields across them, the change in electric field occurring in the spacer.

 

Fig. 2. (a) shows CW spectra from wafers V428 (left pane) and V429 (right pane). Each shows emission at the lower frequency ceasing then lasing recommencing at the higher frequency. Figure (b) and (c) show the corresponding IVs for V428 and V429, respectively, taken in CW operation.

Download Full Size | PDF

 

Fig. 3. LIVs for ARs A and B when grown and processed as single colour devices.

Download Full Size | PDF

These observations of the electrical and optical behaviour of the devices are consistent with the data from the two active regions when grown and processed as single colour lasers. Data from AR A and AR B when grown with 45 repeats and processed into single plasmon waveguides, using the same method of ref.[9], are shown in Fig. 3. It is clear that the range of currents at which the single colour devices lase is consistent with the dynamic range of each colour in the heterogeneous cascade devices described above. The maximum operating temperatures in pulsed mode of AR A and B are 33K and 53K for these devices. Performance is poor due to the combination of a reduced number of periods and the single plasmon waveguide, which provides much lower overlap factors. The difference in temperature performance observed here is commensurate with the difference between the two ARs when included in the heterogeneous cascade devices above. The relatively poor performance of the higher frequency design A is due to the fact that this design has not been optimised.

The form of the IVs in Fig. 3 support the conclusions drawn above on the electric field distribution through the devices in Fig. 1. One may think of the two sections, A and B, in the devices from V428 and V429 as effectively in-series; the current through each must be identical but the voltage dropped across each section may be different. This leads to features in voltage at particular currents in the IVs of Fig. 3 appearing in the IVs of Fig. 1, such as the jump in bias of V428 around 300Acm^-2 and of V429 between 200 and 250Acm^-2 which we identify this with the increase in bias of design B around 200Acm^-2 in Fig. 3.

 

Fig. 4. Upper pane: Calculated intersubband absorption for design A in a low field state, estimated to be 0.5kVcm^-1, and design B aligned at 1.9kV cm^-1. The lower pane has calculated intersubband absorption for design A aligned at 2.6kVcm^-1 and design B at 9kVcm^-1.

Download Full Size | PDF

4. Waveguide

Waveguiding of each frequency takes place in the same double metal waveguide, the calculated overlaps of the 3.0 THz optical mode with AR A and the 2.6 THz optical mode with AR B are both 49.0% for V428, the values are 43.8% and 52.8% respectively for V429. The differences in overlaps are due to the slight asymmetry of the mode, partly caused by the different thickness and doping level of the contact layers, and also the different thicknesses of the two ARs.

The modes were calculated from a 2-D model in COMSOL and only take into account the losses of the mode in the bulk doped contact layers, calculated for the 2.6THz and 3.0 THz modes in both wafers as 12.9cm^-1 and 15.6cm^-1 respectively. The extra absorption from the 250nm thick separator region was estimated to be just 0.2cm^-1. These values allow us to calculate the figure of merit for each frequency in each waveguide, typically defined as,

where Γ is the optical overlap, α _w is the waveguide loss and α _m is the mirror loss, assumed to be negligible in this case. Using this we obtain 0.038cm and 0.031cm for the 2.6 THz and 3.0 THz modes of V428 respectively and 0.041cm and 0.028cm for V429.

In addition to the waveguide loss calculated above, another potential source of loss is intersubband absorption of the miniband. When design A and design B are aligned the self-absorption is small, however it is possible that the non-lasing design, that is not in an aligned state may have a transition resonant with the emission frequency, causing some parasitic absorption. This has been calculated, the intersubband absorption curves are shown in Fig. 4. When the lower frequency mode is lasing (i.e. design B is aligned) we expect design A will have little field dropped across it; from the IVs of Fig. 1 we estimate this field across design A to be 0.5kVcm^-1 when design B is aligned. When the higher frequency mode is lasing (i.e. design A is aligned), design B is in a high field regime. From the IVs we estimate this high field across design B, after accounting for a Schottky barrier and contact resistance, to be 9kVcm^-1. The absorption curves are calculated at an electron temperature of 100K, which seems to be a reasonable estimate [12] under normal conditions. However for the high field case, this may not be a good approximation and the actual effective temperature may be significantly higher, reducing the intersubband absorption for this case. The results of Fig. 4 indicate that for the lower frequency case (upper pane) the intersubband absorption of design A is very similar to that of B. However, for the high frequency case (lower pane) the absorption of the spectator active region, B, is significantly higher than that of the aligned, operational design, A. This large increase in intersubband absorption could contribute to the poor performance of higher frequency AR and should be a consideration in the design of future heterogeneous QCLs.

5. Conclusion

We have demonstrated the first THz quantum cascade laser with a heterogeneous cascade. The structures presented show sequential lasing on frequencies around 2.6 and 3.0 THz. Using the method described in this paper and optimised active regions it should be possible to produce simultaneous emission from two terahertz modes.