Optical Far-IR wave generation - state-of-the-art and advanced device structures

Optical Far-IR Wave Generation -

State-of-the-Art and Advanced Device Structures

V. Krozer

, B. Leone

, H. Roskos

, T. Löffler

, G. Loata

, G. Döhler

, F. Renner

, S. Eckardt

,

S. Malzer

, A. Schwanhäußer

, T. O. Klaassen

, A. Adam

, P. Lugli

, A. Di Carlo

, M. Manenti

,

G. Scamarcio

, M. S. Vitiello

, M. Feiginov

Technical University of Denmark, Kgs. Lyngby, Denmark

b)

ESA Directorate of Technical and Operational Support, ESTEC, Noordwijk, The Netherlands

c)

University of Frankfurt, Frankfurt, Germany

d)

University of Erlangen, Erlangen, Germany

e)

Delft University of Technology, Delft, The Netherlands

f)

Technische Universität München, München, Germany

g)

University of Rome "Tor Vergata", Rome, Italy

h)

University of Bari, Bari, Italy

i)

Technical University Darmstadt, Darmstadt, Germany

ABSTRACT

A recent study initiated by the European Space Agency aimed at identifying the most promising technologies to significantly improve on the generation of coherent electromagnetic radiation in the THz regime. The desired improvements include, amongst others, higher output powers and efficiencies at increasingly higher frequencies, wider tunability and miniaturization. The baseline technologies considered revolve around Photomixing and novel laser based technologies compared to all electronic techniques. Some of the most significant findings will be presented together with technological developments and experimental results selected for medium to short term development. These technologies include advanced p-i-n photomixer with superlattice structures and, THz quantum cascade lasers. Recent results achieved in these fields will be put into the potential perspective for the respective technology in the future.

1. INTRODUCTION

The optoelectronic local-oscillator concept is a new approach to provide THz frequency power for heterodyne receivers. Currently, no compact power sources exist for the generation of the LO signal in the frequency range between 1-10 THz, as indicated in Figure 1. However, this range of frequencies is interesting for numerous space and earth exploration experiments, for imaging and bioengineering applications. The introduction of optoelectronics in these applications was long delayed by the limited power level reached up to date. As this power level gradually increases and the power requirements for mixers decrease, the advantages of photonics, i.e., easy transport and distribution of the ultra-high-frequency signal, straightforward frequency stabilization, broad tunability range, and the lack of important limitations concerning the potential increase of the oscillation frequency, begin to bear fruit. A comparison of different optical techniques for THz power generation has been presented on a conceptual level in [2,18].

Optoelectronic local-oscillator systems are currently under development in the context of the joint European-American-Japanese ALMA project. Future space missions will also rely heavily on interferometric techniques over ranges beyond the reach of current electronic technologies. The need for interferometric interconnection over many kilometers, necessitate suitable optical LO power generation devices.

The purpose of the present study is to survey and assess the various novel techniques for photonic generation of THz power, to demonstrate the experimentally achievable capabilities of photo-mixer devices based on LT GaAs, p-i-n diodes, and of lasers for THz power generation, and to survey and assess the measurement, characterization and system requirements for optical far-infrared power generation.

Figure 1. Overview over the available power sources versus operating frequency. Power slopes with frequency are also indicated. Original data has been prepared by J. Hesler [1].

The state-of-the-art in the area of photomixing is currently set by two device concepts, namely the LT GaAs MSM structures and the p-i-n diode structures. Both device concepts have already demonstrated the working principle but are still far from the requirements in output power for receiver applications at THz frequencies. These techniques will be considered in section 2. Another possibility considered in this study is the direct generation of THz power with either p-Ge lasers or quantum-cascade laser structures, which are described in section 3. These techniques have been introduced only recently and their potential in performance, especially for operation at the higher end of the frequency range indicated above, is of great interest in the preparation of new applications. We also consider a large number of other known techniques for the generation of THz power, with special emphasis on their CW THz power capabilities. Based on the discussion of the various concepts for optical THz power generation, we have identified the most promising technologies and possible strategies for their development, which are summarized in detail at the end of the paper.

2. THZ POWER GENERATION WITH PHOTONIC DEVICES

Semiconductor-based photo-mixers are very promising THz sources in terms of most desirable (operation over a broad temperature range (100-300K), e.g.) or necessary (tuning range, LO line-width, e.g.) requirements. However, the output power at THz frequencies is still low. Photomixing means the superposition of two, in frequency slightly detuned laser-beams, which leads to a beat signal with the difference frequency. This beat frequency can be set in the THz-range (0.1-10THz). The modulated laser-beam is absorbed in the photoconductor and periodically generates electrons and holes, which are separated in an electric field. The resulting THz-modulated current is fed into an antenna, which emits CW THz radiation.

The first promising THz power generation with photonic devices was demonstrated in 1994, using an LT-GaAs photo-mixer. Recently, photomixing in PIN diodes, which has proven as a very attractive approach up to about 100 GHz, is being considered as an alternative. LT-GaAs-based and PIN diode based state-of-the-art photomixers, advanced photomixing concepts and recent activities for improving these devices are reviewed. The theoretical limitations of conventional PIN photomixers are discussed and a novel n-i-pn-i-p-superlattice photomixing concept is proposed.

2.1. THz power generation with LT GaAs devices

State of the art conventional photo-mixers are based on MSM-structures on LT-GaAs driving a log spiral (broadband) or a double dipole (resonant) antenna. The state-of-the-art results are set by two groups at MIT Lincoln Laboratories and Jet Propulsion Laboratories (California Institute of Technology), respectively [3,4,11]. The overall output power of LT-GaAs-based THz photomixers is limited by a number of factors. The key problems are:

100 1000 10000 100000 Output Frequency (GHz) 0.01 0.1 1 10 100 1000 10000 100000 O ut p u t Pow er ( m W ) D D D DD D D D D D D D D DD D DD_DD D D D D DD DD D D D D D T T T T T T T T T T TT T T T_T T T T G GGGG_GGG G G GG G G G G G GG GG G G G G G G_G R R R I II II II I I I I I I I AA A AA AA A A A A D D D D D L L L L L L L L L L L L L L = Laser D = Doubler T = Tripler G = Gunn Diode I = Impatt Diode R = RTD A = Amplifier Cooled L Pulse d

upper limit to power generation with multipliers

~f

~f

~f

~f

• Thermal failure due to high thermal resistance of the LT-GaAs and/or the substrate material limits both the current density through the devices and the optical power on the devices,

• Impedance mismatch between the mixer (~10 kΩ) and the antenna, and

• Inhomogeneous vertical electrical field distribution in lateral mixer structures.

These devices are pumped with ≈100mW optical input power (for devices with special heat sink with up to 250mW). The bias voltage applied to the typically 8x8 µm2 large MSM-structures (finger contact with 0.2µm width and 2µm spacing) is about 30V. As seen from Figure 2, the THz-power at 1THz is less than 2µW even for improved mixers [5,6]. Despite a lot of research activities during recent years the benchmark set by Brown et al. in 1994 [Bro95] is still a state of the art result. Various groups around the world are working to optimize the THz output power. Most approaches are based on an improved heat sinking to allow for a higher optical input power.

Figure 2. Simulations and measurements for four different twin-dipole photomixers are shown. Also a log-spiral antenna that rolls off at 12 dB/octave is depicted [5]. Recent results from [6] scaled to the correct laser power are indicated.

Conventional photo-mixers have been fabricated with many different electrode geometries. The active area of the MSM structures has been varied between 4×4 µm2 and 20×20 µm2 and many different values for the finger width and spacing have been investigated. As a compromise between photoconductive gain, optical spot size and capacitance best performance was achieved with 8×8 µm2 area, 0.2µm finger width and 1.8µm spacing [4,7]. The typical bias voltage applied to the finger contacts is about 30V. The output frequency can be tuned over several THz by adjusting the appropriate wavelengths difference of the two single-frequency lasers. The output signal is used to drive an antenna. At present usually log spiral (broadband) or double dipole (resonant) antennae are used. The most attractive features of photomixers are room temperature operation and their wide tuning range.

From a theoretical study we can derive that a further improvement of the THz output power of conventional photo-mixers based on LT-GaAs by a factor 10 or 50 (for an optical input power of 100mW and 230mW, respectively, should be possible, however, with large technological efforts. In addition there are strong arguments that fundamental key parameters, like the lifetime of photo-generated carriers in the presence of an electric field, the internal field screening and the voltage dependence of the photocurrent are not completely understood and hampering further developments.

The THz output power is given by PTHz ∝I2photo·RA

/2

impedance of the antenna. The photocurrent is given by Iphoto

≈

.

efficiency, Popt the optical input power, and τe / τtr the gain factor, where τe is the lifetime of the photo-generated carriers

and τtr the transit time for drifting between the contact fingers. The most efficient way to improve on the THz output

power is to increase the AC component of the photocurrent. This can be done either by improving the optical input power Popt or by optimizing the responsivity. The optical input is limited by the thermal damage threshold of the

1µW

Mikulics

2002

photomixer due to optical and ohmic heating [8]. This limitation can be overcome by either using traveling-wave concepts or improving the heat-sink, including the reduction of the thermal resistance and optimization of the field distribution in the devices. Most of approaches to increase the optical input power are presently based on improvements of the heat sink due to the difficulty to find efficient traveling-wave structures. A very promising approach is based on using hetero-structures with thin LT-layers. With such a design, using only every thin layers of LT-GaAs embedded in standard GaAs, it is possible to fulfill the partly contradicting requirements desired for photomixers: short lifetime of photo-generated carriers, high carrier mobility and high thermal conductivity. The thermal conductivity of standard GaAs is much better than for LT-material.

An alternative strategy to improve the overall thermal budget is to achieve a higher photo current for a given optical input power. Apart from optimizing the external losses (transmission and reflection) the responsivity of a photoconductor can be improved by the gain factor τe / τtr (see formula above). In principle the lifetime of the

photo-generated carriers τe can be influenced by the growth parameter and post-growth treatment like annealing. The

dependence of the THz output power on the lifetime τe is given by PTHz ∝I 2

photo·RA / (2 +2ω ·τe). However, the shortest

values of about 300fs obtained up to now are already a good compromise for THz-devices.

The photoconductive gain for carriers generated deep in the active region is lower than for those generated close to the surface. Optimized lateral field distribution can provide an enhancement of about 50% with buried contacts (on high resistivity substrates) instead of surface contacts [9]. Another strategy is to use a material with a larger absorption coefficient or choosing a suitable wavelength for the incident light. The overall absorption in this case can be increased by using a high reflectivity mirror beneath the photomixer [11] and/or exploiting cavity effect to increase the effective interaction length of the light.

Investigations of the voltage dependence of the photocurrent have shown a super and sub quadratic bias dependence, which has been explained with strong field screening due to charge accumulation. Strong field screening by photo-generated carriers and ionized defects in the LT-GaAs material is expected for electric fields higher than 5kV/cm. For the transport on a short time scale also ballistic transport has to be taken into account, with largest displacements ∆zmax ≈ 150 nm for typical times te≈250fs at fields of about 20 kV/cm [10].

2.2. THz power generation with PIN devices

Generation of THz radiation with p-i-n photomixers is a relatively new approach. Nevertheless, encouraging results have already been demonstrated by several research institutions in Europe, USA and Japan [12-15]. In general, lumped-element photomixers are RC-time limited and generated THz power rolls-off quite fast with frequency. There are two main directions of research in the field. One is to try to overcome limitations of the lumped-element photomixers by implementation of travelling-wave designs [12]. In this case, the optical wave should propagate in the same direction and with the same speed as the THz wave. In principle the approach can give possibility to overcome the limitations of the lumped-element photomixers, but it is mandatory to velocity match the 2-3 times slower optical wave to the THz wave in the p-i-n photomixers. The typical example of the spectrum of the THz wave is shown in the Figure 3. Simulations show that the advantages of TW photomixers are limited due to the large velocity mismatch. Optical mode is the spectrum that has to be achieved to match the velocities of the THz and optical modes. One can see that the THz mode is approximately 3 times slower than the optical one. The output THz power vs. the length of photomixer for a typical TW PIN photomixer is shown in the Figure 4. One can depict that maximum output power is delivered for lengths approximately equal to λ/5 of the THz wavelength. The maximum of the photocurrent density is fixed when

αoptical changes, since photocurrent is limited be the finite thermal dissipation. One can see that the output power has

maximum when the optical absorption length (1/αoptical) is approximately 1/5 of the THz wavelength.

Close-to-ideal performance of TW PIN photodiodes can be achieved for frequencies < 200 GHz. So far, there are no reports on 1-THz performance of PIN photomixers. The highest frequency results reported so far are an emitted output power of 1 µW and 100nW at 460 GHz and 625 GHz, respectively [12,13]. The best value reported for 100 GHz is 2 mW. For an optimized PIN photomixer at 20 mW absorbed laser power the theoretically predicted value for the THz output power is about 2.5 µW and at 50 mW absorbed laser power around 60 µW value, which are necessary for pumping of mixers.

Figure 3. The spectrum of the THz mode in the TW photomixer designed by Duisburg University [12]. γ is the complex propagation constant of the mode.

Figure 4. The output power vs. optical absorption constant (αoptical), β is the propagation constant of the THz wave.

Velocities of the THz and optical waves can be matched in the passive THz waveguides periodically loaded by the lumped-element photomixers, but the dimension of the active sections with the lumped-element photomixers and distance between them should be small as compared to the THz wavelength in the waveguide. This condition can be met at the frequencies of the order of 100GHz, but is difficult at THz frequencies. The wavelength decreases linearly from 80µm at 1THz to 25µm at 3THz, and it is difficult to accommodate devices, which still behave as lumped elements.

A very good approach to improve the performance of lumped-element photomixers has been proposed and demonstrated experimentally by NTT group: uni-travelling carrier (UTC) photodiode [15]. In the UTC photodiode the optical absorption and electron transport regions are separated and can be optimized independently. The major advantage of the approach is that the electron velocity can be maximized in the photodiode. The output power of 2mW at 100GHz has been demonstrated with the UTC photodiodes with the optical pump signal at 1.55µm wavelength and 100mW of optical power. An RC-time constant of 0.3ps and an electron transit time of 0.5ps can be estimated for this photomixer with an area of 2x5µm2, a thickness of the intrinsic region of 0.2µm and the electron drift velocity of 4x107cm/s. The time constants are in reasonable agreement with the 310GHz 3dB-bandwidth and 750GHz 10dB-bandwidth reported.

We have developed a comprehensive model for simulation of TW p-i-n photomixers. The model is based on equivalent-circuit (EC) shown in the Figure 5. The TW p-i-n photomixer can be considered as two coupled waveguides: one is formed by the metal coplanar line on the top of the structure (the waveguide is represented by Ccoplanar, Lcoplanar and

Rskin,met,side in the EC) and another one is formed by the central metal stripe of the coplanar waveguide and the n-doped

conducting layers underneath (the layers are separated by the isolating undoped region) -- Cstrip, Gi, Lstrip and Rskin,sem in

the EC. The two waveguides are coupled via the conducting n-doped region (resistance Rs in the EC) and skin-layer

resistance of the common metal stripe (Rskin,met,strip). The resistances Rskin,met,side, Rskin,met,strip, Rskin,sem and Rs are the

complex values. Rskin,met,side takes into account skin effect in the side metal stripes of the coplanar waveguide and Rskin,sem

and Rs allow for the skin effect in the n-doped semiconductor region. The current source (Ioptical) has been introduced in

the EC to simulate the current generated by the optical pump.

The photomixer designed by the University of Duisburg has been simulated using the EC above. The generated power vs. frequency is shown in the Figure 6. The optical input coupling efficiency was supposed to be approximately 7% in the calculations and the input optical power was 300mW. The THz-signal was coupled out to the antenna with impedance of 19 Ω and the length of the photodiode was 50µm. The calculated THz-power level is in good agreement with the experimental value of 1µW at 460GHz. Also, the roll-off of approximately 10dB between 25 and 160GHz is in good agreement with the experimentally measured data. So, our model is in very good agreement with experimental data. Optical mode Re(γ) P TH z (arb. un it s ) αoptical/β 5 1 λ αoptical ≈

Figure 5. The equivalent circuit of the TW p-i-n photodiode.

Figure 6. Simulation of the output THz power vs. frequency for TW p-i-n photomixer fabricated by University of Duisburg [12].

2.3. THz power generation with nipnip devices

Conventional photomixers like LT-GaAs-mixers [11] or pin-photomixers [16] are limited by the RC-roll-off of the THz-power due to the finite device-capacitance and the transit-time-roll-off of the carriers, which have to reach the contacts according to their drift-velocity before new carriers are generated. Thus, the THz-power is given by

2 2 2 2 1

)

(

1

1

)

(

1

1

)

(

I

R

P

ν

+

⋅

+

⋅

=

Iph is the generated ac-photocurrent, RA the antenna-resistance, νRC the RC-frequency of the photomixer and νtr the

transit-time-frequency of the carriers, depending on their drift-velocity and the distance they have to cover. LT-photomixers are limited by the capacitance-roll-off and to a great extent by the fact, that the achievable photocurrent is limited by the photoconductive gain g = τrec/τdr≈ 10-2, given by the recombination-lifetime τrec of the carriers in the

LT-material and the transit-time τdr of the carriers until they reach the contacts. Pin-photomixers are limited by the trade-off

that by minimizing the drift-length of the carriers (i-layer) and consequently the transit-time-roll-off, the capacitance increases and therefore the RC-roll-off. Minimizing the capacitance by increasing the i-layer-length leads to an increase of the transit-time.

The nipnip-THz-emitter (Figure 7) can overcome both limitations. It can be seen from Monte-Carlo-simulations of the field-dependent electron-transport in Figure 8, that electrons can cover a distance of approx. 200nm travelling quasballistically according to the velocity-overshoot before intervalley-scattering into the L-valley sets in. Therefore, the i-layer-lengths are chosen approx. 200nm, leading to transit-time frequency νtr≈ 1 THz [19,20]. Thus, the emitted

THz-power is not transit-time limited up to 1 THz. The RC-roll-off is minimized by growing a stack of these transport-optimized nano-pin-diodes, leading to a significant reduction of the capacitance. Carriers are photo-generated near the layer of the pin-diodes by changing the Al-content across the i-layer, leading to a minimum-bandgap near the p-contacts. Thus, only electrons contribute to the transport, the slower holes remain near the p-contact. One period of the emitter is depicted in Figure 7. Under illumination, the photo-generated carriers are separated in the electric field inside the i-layer. Under optimized working conditions, the resulting dipole-field screens the internal field to the value optimum for transport. External voltage doesn’t have to be applied necessarily. The electrons, accumulating in the triangular potential-well have to recombine with holes across the internal np-junctions to prevent a complete flattening of the field. For the use in the emitter, current densities of approx. 1-100 kA/cm2 have to be achieved under 1V forward bias inside these recombination-np-diodes. These high values were achieved by growing semi-metallic ErAs between the np-junction or by growing the np-junction with LT-material [21].

Gi Cstrip Ccoplanar Rskin,met,strip Rskin,sem Lstrip Rskin,met,side Lcoplanar Rs Ucopl, Icopl Ustrip, Istrip 0 Ioptical Ω =19 A R m L=50µ mW Popt=300 -30dBm -40dBm

Figure 7. Band diagram, Al-content and doping profile of one period of the n-i-pn-i-p-superlattice.

Figure 8. Monte Carlo results velocity and position vs. time.

a) b)

Figure 9. Results for the advanced PIN diodes: a) emitted THz-power vs. the mixing-frequency. The measured data is fitted with the RC-roll-off only; b) THz output power versus optical input power. Some water-absorption lines are also shown.

A representative measurement of the emitted THz-power vs. frequency and vs. the input optical power is depicted in Figure 9. The emitter consists of 7 periods, including ErAs-recombination-diodes. The sample was structured leading to photomixers with a mesa-area of 7µmx7µm and a capacitance of C = 7fF. A log-periodic antenna was attached, operating between 0.1 THz and 3 THz, as depicted in Figure 10. The measurement clearly shows, that only the

RC-roll-h =EνL G GaAs EG AlGaAs lz,h lz,e lz Φp Φn Φnp ∆h ∆h ∆e -z nD nA nD nA Al-content LT-GaAs or ErAs defect level + recombination layer 0,000 0,25 0,50 0,75 1,00 3x107 6x107 9x107 1x108

onset of intervalley scattering

onset of polar optical emission 5 kV/cm 10 kV/cm 20 kV/cm 50 kV/cm 100 kV/cm vz (c m/ s ) time (ps) 0 ,0 00 0 , 2 5 0 ,5 0 0 , 7 5 1 ,0 0 1 0 0 2 0 0 3 0 0 ∆t ∆z 1 0 0 k V /c m 5 0 k V /c m 2 0 k V / c m 1 0 k V / c m 5 k V / c m z (n m) tim e ( p s ) 100 1000 1 10 100 PTH z [ a .u .] frequency GHz 7 periods fit: RC-roll-off 0 10 20 30 40 50 60 70 0 25 50 75 100 125 150 175 200 225 250 275 300 Lo ckI n s ig n al 10 -6V

optical power photomixing signal (mW)

P_THz measured PTHz

max

with applied bias quadratical fit of P_THzmax

off limits the THz-power. The carriers are not transit-time-limited due to the velocity-overshoot. Therefore, the nipnip-THz-emitter is superior to a conventional pin-photomixer.

a) b)

Figure 10. a) Layout of the log-periodic 120-3000 GHz antenna with bias circuit designed for advanced PIN device and b) THz output power versus frequency for standard PIN devices and advanced PIN devices.

3. THZ POWER GENERTION WITH CRYOG ENIC AND LASER STRUCTURES

3.1. THz power generation with quantum cascade lasers

Quantum cascade lasers are unipolar lasers that exploit the radiative transitions between two subbands of a multi quantum well system [22] or two minibands of a superlattice [23]. The basic cell is composed of an injector and an active region. The injector provides electrons to the upper level of the active region. Under sufficiently high electric fields, a population inversion builds up between such upper level and the lower one, which in turn depletes itself via non-radiative transitions with the level(s) of the active region sitting at even lower energy. The basic cell is repeated for 30 to 40 periods in order to enhance the output power. Until recently quantum cascade lasers could be built only to operate in the range of 5-20 µm, with output power of the order of few mW, under CW conditions (at least of to 200 K) or under pulsed condition up to room temperature and higher [24–27].

Recently, the first QCL emitting in the THz region was reported working at 4.4 THz with an output power of 2 mW up to a temperature of 45 K [28]. Now, with rapid progress in the field, also lasers emitting in the 3.4 - 2.3 THz range have been reported, at temperatures up to 135K [29-33]. The two existing structures, namely those fabricated by Tredicucci (Pisa) and Faist (Neufchatel) have been simulated with Monte Carlo in order to test both the Schrödinger and Poisson equation solver and the MC program.

We have studied the influence of technological parameters such as barrier/well thickness and composition, injector design, level spacing, in order to determine the processes that control the achievement of population inversion between the states involved in lasing. We have also analyzed the effect of the external temperature on the QCL optical characteristics.

Then, the activity has been focused on the design of two 4-quantum well GaAs-based QCL’s, operating at 4.1 and 4.2 THz, based on different depopulation schemes. In the laser operating at 4.2THz (see Figure 11) we have investigated an active region design based on bound-to-continuum transitions. In this design electrons are injected into an isolated state created by a thin well adjacent to the injector barrier while electrons are extracted through a lower miniband. Owing to the diagonal nature of the laser transition both the injection efficiency and lifetime ratio are maximized. The

100 1000 1 10 100 1000 pinpx4 x 10 pinpx7 x 10 pin (LI µV) THz pow e r [a .u.] frequency (GHz)

THz-results ErAs series; 50mW;

f

f

radiative transition occurs between an upper level and a group of two levels in the lower miniband; because of level broadening, caused by interface roughness and impurities it’s difficult to identify these states individually.

In the laser operating at 4.1 THz, as indicated in Figure 12, we explore, in terms of electrical performances and potential features of laser emission, a different active region scheme based on the direct use of LO-phonons for depopulation of the lower state. This scheme offers two distinctive advantages. The collector state is separated from the lower state by at least ELO, so depopulation can be extremely fast and it does not depend much on temperature or on the

electron distribution. A second important point is the large separation that provides intrinsic protection against thermal backfilling of the lower radiative state. We have also optimized the long upper state lifetime; the active region is based on 4 (see Figure 12) quantum wells in which we have a vertical laser transition and a highly selective depopulation. In fact the lower laser level is anti-crossed with a state in the adjacent well where fast LO phonon scattering takes places.

A typical characteristic of this kind of design is the intra injector barrier (25 Å) that has a strong influence on the threshold current density. In fact, a large value of Jthdoes not necessarily reflect a weak gain or high optical loss, but it

rather reflects the presence of a strong parasitic current channel.

We have designed different active regions changing the barrier thickness. Analyzing our results, we found a progressive injector levels anti-crossing reduction and a reduced coupling between the lower injector state and the upper LO phonon transition level. Besides, we found a more selective injection into the upper laser level and a progressive electrical dipole decrease associated with the increase of the barrier thickness. The importance of these effects is in the possible reduction or suppression of the parasitic channel associated with the transitions between the two injector levels and the lower laser level and in the high-field domain prevention that gives us the possibility to obtain narrower emission line width and more modules contributing to optical gain. In the Monte Carlo simulations the population densities of states and scattering rates between laser and injection levels, gain spectra and phonon distribution functions have been investigated for different temperatures.

The optical waveguide has been also optimized, as depicted in Figure 13. We have designed an optical single plasmon waveguide based on the confinement provided by the interplay between a metallic reflection at the top metallization and the quasi metallic confinement provided by a thin, heavily doped buried contact. Because of the large dielectric constant of the doped buried contact, the overlap factor Γ between the field and the buried contact is very small, leading to much lower waveguide losses α. However for longer wavelengths a thicker buried contact layer is necessary and this lead to a larger value of the active region confinement at the cost of a larger loss α., with an overall decrease of Γ/α. We have carefully optimized all the details, still maintaining a large value of Γ/α. Great care has been taken in order to optimize the substrate residual doping with the aim to increase the optical mode confinement.

Figure 11. Bound-to-continuum chirped superlattice GaAs/Al0.15Ga0.85As (4.2 THz)

Figure 12. QC- structure based on LO scattering depopulation (4.1 THz)

Figure 13. THz mode structure in the single plasmon waveguide for the 4.2 THz (left) and the 4.1 THz (right) QC laser Two samples have been grown by MBE. The device fabrication has been optimized; in particular, great care has been taken in order to optimize metal deposition and annealing recipes in order to improve electrical performances and decrease the differential resistance (see Figure 15). Electrical measurements indeed show a strong decrease of the differential resistance and the typical trend I-V characteristics of an e-LO THz laser, as can be seen in Figure 14.

The device show a differential resistance of 8Ω and a voltage threshold at 12V; this value correspond to the condition of alignment between the injector level and the upper radiative level. In the region between 1.2A and 1.8A there is the typical kink that underlines the presence of a parasitic current channels and the onset of high field domains (NDR); this effect is due to fast LO phonon scattering allowed to lower–energy states witch are located far away.

0 10 20 30 0 0.5 1.0 1.5 2.0 2.5 I(Ampere) V(Vol t) 8Ω

Figure 14. Pulsed I-V characteristic of a 4.1 THz laser sample (200µm wide, 1.4 mm long) at T=16K

Figure 15. Photograph of a typical laser facet

3.2. THz power generation with p-Ge lasers

A second subject of investigation is the p-Ge laser under uniaxial stress [34]. With uni-axial pressure the light- and heavy hole sub-band split up (see Figure 16) and for P > 2.5 kbar // [100] and P > 4 bar // [111], the acceptor ground state connected with the heavy hole band enters the continuum of the light hole sub-band and becomes therefore a “resonant”- or “auto-ionizing” state. Under impact ionization conditions, the scattering of the holes at this resonant state results in a trapping of the holes near this state, leading to a maximum of the non-equilibrium hole distribution function near this energy [28]. The proposed population inversion mechanism is illustrated in fig. b [26,27].

A second subject of investigation is the p-Ge laser under uniaxial stress [25]. With uni-axial pressure the light- and heavy hole sub-band split up (see fig. a) and for P > 2.5 kbar // [100] and P > 4 kbar // [111], the acceptor ground state connected with the heavy hole band enters the continuum of the light hole sub-band and becomes therefore a “resonant”- or “auto-ionizing” state. Under impact ionization conditions, the scattering of the holes at this resonant state results in a trapping of the holes near this state, leading to a maximum of the non-equilibrium hole distribution function near this energy [37]. The proposed population inversion mechanism is illustrated in Figure 17 [35,36].

Figure 16. Pressure dependence of the splitting of the Ge light- and heavy hole valence bands and the acceptor ground states [35]

Figure 17. Energy level scheme showing the heavy hole 1s resonant state in the light hole continuum possible laser transitions [35]

The pressure dependence of the most intense transition, 1sr → 2p±, for P // [111] is depicted in Figure 18. Easy

frequency tunability from 5 – 10 THz is obtained by changing the applied pressure // [111], whereas for P // [100] a tunability from 2.5 THz at 3.9 kbar to 10 THz at 8 kbar is observed [35-38]. The line width of these transitions is about 3 cm –1, probably due to the rather large lifetime broadening of the resonant state. This would enable cavity fine-tuning without changing the applied pressure. Depending on crystal quality and especially on the quality of the “internal reflection” cavity the laser threshold varies. For P // [001] a laser threshold below 10 V/cm [39] has been observed. With such low thermal CW emission at 2.5 THz has been obtained with a reported output power of 1-2 µW.

Figure 18. Pressure dependence of the frequency of the main emission peak in p-Ge. [35].

In close cooperation with the Moscow group (Kagan and Altukhov) we are starting a program to further investigate

this type of laser, with special emphasis on the possibilities of elevated temperature operation (T>4 K), CW operation and the reproducibility of the pressure tuning.

3.3. THz power generation with cryogenic devices

In general it has to be pointed out that cryogenic devices have only limited applicability at frequencies above 1 THz. However, a number of interesting experiments need still to be done, such as e.g. direct illumination of HEB devices, etc. Devices for THz power generation with Josephson junctions based on low Tc and high Tc superconductor have been

• THz emitters based on low Tc superconductors (e.g. Nb) are highly developed but limited to frequencies

below 700 GHz.

• For THz emitters based on high Tc material a lot of work is in progress, however the best results reported so

far are not very promising (a THz power of less than 1 nW at a frequency of 1.1THz).

From the physical point of view, Josephson devices are ideal high frequency emitters. They are tunable simply by applying a voltage, the generation mechanism is intrinsic, no high-speed electrical circuit is needed, neither optical pumping. The only necessity is the use of cryogenic temperatures.

The problems arise due to two main issues: power generation and frequency of operation, both referring to either high TC or low TC junctions. As mentioned low TC, devices based on Niobium can be fabricated with high accuracy in

commercial production facilities. The power scales with the number of junctions and output powers as high as 400µW @ 410GHz have been reported. But this technology is not available for THz radiation. The frequency referring to the binding energy of a cooper-pair is 750 MHz. Above this value no radiation will be emitted based on Nb junctions.

An Emitter based on high TC technology, could theoretically deliver radiation up to THz and frequencies little above

1 THz have been reported, but the power is below one nW. To rise the number of devices in a circuit, they must be sufficient uniform to have them oscillate synchronously. To compare the situation to the low TC history, where it took

decades to evolve the necessary production technology, one may assume junction is out of side, but may come up in the next decade.

4. MEASUREMENT, CHARACTERISATION AND SYSTEM ASPECTS

The photomixer performance was characterized by the setup shown in Figure 19 in which the optical pump power is derived from two independent Ti:Sapphirelasers operating near 830 nm. Each laser, in a ring cavity configuration (travelling wave), operates at a single frequency which is continuously tunable over a wide range using a birefringent (Lyot) filter inserted in the cavity. The two beams are spatially combined in a 3 dB single-mode 2 × 2 fiber-coupler.. As shown in the Figure 19, both laser beams pass through polarization controllers to match the electric field for maximum interference. After combining, the THz-modulated output beam is transformed back into free space and focused on the photomixer active region using a 15 mm achromatic lens (estimated laser spot size in the focus ~ 7 µm). The generated THz radiation propagates mainly into the GaAs substrate, where it is coupled out into free space via a 10-mm-diameter high-resistivity silicon hyperhemispherical lens abutted to the backside of the substrate. The radiation coming out of the lens forms a diverging quasi-Gaussian beam that is readily focused into a liquid helium cooled InSb bolometer using two identical off-axis parabolic mirrors (parent focal length of 152.4 mm). The field-enhanced InSb bolometer is very broadband with a frequency response over the range 0.1 – 1.5 THz. To enable lock-in detection of the difference frequency an optical chopper (set at a chopping rate of 1 kHz) was inserted in the output optical beam.

Figure 19. Experimental setup for the characterization and testing of the photomixers.

For characterization of the frequency dependence of the THz output power, the THz frequency was changed by tuning one of the lasers (with its birefringent filter) in steps of 0.1 nm (corresponding to 25-GHz steps in the frequency

Ti:Sa ring laser I Ti:Sa ring laser II ν1 ν2 Polarization c ontroller 3 dB fiber-c oup ler

c hopp er lens

photom ixer

Si hyperhem ispheric al lens

Spectrometer/ Fabry-Perot etalon

4.2 K InSb bolometer

sweep). The difference frequency and the single-mode operation of the lasers were both monitored with a grating spectrometer and a scanning Fabry-Perot interferometer.

The considerations given above lead to the conclusion that the implementation of a compact system for photomixing is clearly feasible based on either diode lasers or on diode-laser-pumped solid-state lasers. Stabilization of the lasers will be needed, which is a common research field in the case of diode lasers. The advantages of diode lasers are compactness, low weight, and high conversion efficiency, whereas they are sensitive to optical feedback and need for optical amplification.

For research purposes solid-state lasers may be used with their good performance, but are bulky and require more optical elements

Summing up, we would suggest employing a stabilized diode-laser set-up, based on the arguments of lower weight, compactness, better conversion efficiency, and more mature status with respect to the requirements.

5. SUMMARY

We have investigated the possibilities of the optoelectronic local-oscillator concept as a new approach to provide ultra-high-frequency power for heterodyne receivers. Such a compact LO power source could be used to generate LO signal in the frequency range between 1-10 THz.

Based on an extensive literature survey and in-house results we have discussed and evaluated various photonic and optical techniques with respect to their applicability as compact THz optoelectronic LO sources. Among the different techniques we have identified the following concepts suitable for future consideration:

• Photonic mixing with advanced p-i-n GaAs photodiodes

• Direct LO signal generation with III-V and/or SiGe QCL laser structures and with stressed p-Ge lasers It has been found that advanced p-i-n GaAs photodiodes are the most prospective devices for future generation of THz LO power. The new original structure proposed here, also referred to as “n-i-pn-i-p” structure, consists of a periodic sequence of p-i-n nanodiodes with thin, but strongly doped n- and p-layers. We expect that these devices outdate optimized conventional p-i-n mixers by large factors. At a frequency of 1 THz and a laser power of 50 mW a THz power of 400 µW, corresponding to a THz conversion efficiency of 0.4 %, is predicted for these devices. The highest frequency results reported so far are an emitted output power of 1 µW and 100nW at 460 GHz and 625 GHz, respectively. In contrast to this stands the 2 µW for LT-GaAs based photomixers at 1 THz and it is predicted that such devices could meet LO requirements only while operating close theoretical limit.

PIN based diodes are currently considered by a number of other projects including the ALMA project operated by ESO. These projects use InP based diode structures, either travelling-wave (Univ. Duisburg) or lumped element (U2T Berlin and NTT Japan). The initial experiments confirm the concept at the required frequencies.

The generation of THz power with lasers, either QCL or stressed p-Ge, are both very interesting in mid-term perspective in the range of 2.5–10 THz. Both offer direct THz power generation capabilities, broadband tunability for some laser structures, and potential applications in other space related programs focusing on frequencies above 4 THz up to the far–IR. Both exhibit a number of advantages as well as disadvantages and could both lead to efficient sources at higher frequencies. One of the most important issues in both laser technologies is the impact of the cavity on laser operation. While the p-Ge laser requires a high quality cavity for efficient lasing, QCL structures could in principle operate with moderate cavity quality factors. We conclude that an in-depth understanding of the role of the cavity for lasing in both technologies, would lead to a big leap towards devices applicable in various environment. With a high quality cavity the experimental results obtained for the p-Ge laser demonstrate wide-band tunable operation in the range of 2.5 – 10 THz at pressures up to 10kbar. In this range pulsed and CW operation has been achieved with large power levels of Pout,pulsed>1mW, Pout,CW>1µW, respectively, and only 50mW power dissipation. Operation with closed-cycle

coolers at 10K seems therefore feasible. In the case of QCL results based on well-known fabrication technologies demonstrated an output power of 2 mW up to a temperature of 45 K at 4.4 THz, however without tuning capabilities.

Currently, a large number of groups in Europe are active in the field of design and fabrication of QCL devices. We believe that it is important to investigate potential properties of QCL technology in the future for frequencies below 3 THz. Further questions concerning tunability, linewidth and stability, and temperature range of operation should also be considered.

ACKNOWLEDGEMENTS

The authors would like to acknowledge partial financial support by the European Space Agency under the contracts ESA/ESTEC- AO/1-3661/00/NL/PB and ESA/ESTEC- TOS-MMR/2002.01/BL/BL. We also acknowledge partial financial support in the early stages of work by the ALMA consortium through the Max-Planck Institut für Radioastronomie in Bonn.

REFERENCES

1. J. Hesler, Univ. of Virginia, Virginia, USA, 1999, private communication.

2. J. B. Khurgin, “Comparison of different optical methods of THz generation”, SPIE Vol. 3624, pp. 128-139, Jan. 1999.

3. E. R. Brown, F. W. Smith, and K. A. McIntosh, “Coherent millimeter-wave generation by heterodyne conversion in low-temperature-grown GaAs photoconductors”, J. Appl. Phys., vol. 73, pp. 1480-1484, 1993.

4. K. A. McIntosh, E. R. Brown, K. B. Nichols, O. B. McMahon, W. F. Dinatale, and T. M. Lyszczarz, “Terahertz photomixing with diode lasers in low-temperature-grown GaAs”, Appl. Phys. Lett., vol. 67, pp. 3844-3846, 1995. 5. S.M. Duffy, S. Verghese, A. McIntosh,; A. Jackson, A.C. Gossard, and S. Matsuura. “Accurate modeling of dual

dipole and slot elements used with photomixers for coherent terahertz output power“, IEEE Trans. Microwave Theory Tech., vol. 49(10), pp. 1032-1038, 2001.

6. M. Mikulics, F. Sicbe, A. Fox, M. Marso, A. Forster, H. Stuer, F. Schafer, R. Gusten, and P. Kordos, “Generation of 460 GHz radiation by photomixing in low-temperature-grown MBE GaAs“, The 4th Intern. Conf. on Advanced Semiconductor Devices and Microsystems, pp. 129-132, 2002.

7. E. R. Brown, K. A. McIntosh, K. B. Nichols, and C. L. Dennis, “ Photomixing up to 3.8 THz in low-temperature-grown GaAs”, Appl. Phys. Lett, vol. 66(3) pp. 285-287, 1995.

8. S. Verghese, K. A. McIntosh, and E. R. Brown, “Optical and terahertz power limits in low-temperature-grown GaAs photomixers”, Appl Phys. Lett., vol. 71, pp. 2743-2745, 1997.

9. M. Mikulics, A. Fox, M. Marso, A. Förster, and P. Kordoš, “Material and device issues of low temperature grown GaAs for THz photomixer”, 3rd Symposium On Non-Stoichiometric III-V Compounds, edited by T. Marek, S. Malzer, and P. Kiesel, pp. 31-36, 2001.

10. M. Eckardt, A. Schwanhäuser, L. Robledo, A. Friedrich, P. Pohl, P. Kiesel, S. Malzer, G.H. Döhler, and A. C. Gossard, “LT-GaAs recombination layers used in photomixers based on quasi-ballistic transport”, 3rd Symposium On Non-Stoichiometric III-V Compounds, edited by T. Marek, S. Malzer, and P. Kiesel, pp. 37-42, 2001.

11. E.R. Brown, “THz generation by photomixing in ultrafast photoconductors”, International Journal of High Speed Electronics and Systems, Vol.13, No. 2, pp.497-545, 2003.

12. A. Stöhr, R. Heinzelmann, K. Hagedorn, R. Güsten, F. Schäfer, H. Stuer, F. Siebe, P. Van der Wal, V. Krozer, M. Feiginov, and D. Jäger, “Integrated 460 GHz photonic transmitter module“, Electronics Letters, Vol.37, No. 22, pp.1347-8, 2001.

13. A. Malcoci, A. Stöhr, R. Heinzelmann, K. Hagedorn, R. Güsten, F. Schäfer, H. Stuer, F. Siebe, P. Van der Wal, V. Krozer, M. Feiginov, and D. Jäger, “Photonic (sub)millimeter wave local oscillators”, MIKON-2002. 14th International Conference on Microwaves, Radar and Wireless Communications, Vol. 3, pp.722-734, 2002.

14. M. S. Islam, S. Murthy, T. Itoh, M. C. Wu, D. Novak, R. B. Waterhouse, D. L. Sivco, and A. Y. Cho, “Velocity-Matched Distributed Photodetectors and Balanced Photodetectors with p-i-n Photodiodes”, IEEE Trans. Microwave Theory Tech., Vol. 49 (2001), pp. 1914-1920).

15. Y. Hirota, T, Hirono, T. Ishibashi, and H. Ito, “Wave-Wave Photodetector for 1.55µm Wavelength Fabricated with Unitraveling-Carrier Photodiodes”, Appl. Phys. Lett., Vol. 78 (2001), pp. 3767-3769.

16. H. Ito, T. Furuta, T. Ito, Y. Muramoto, K. Tsuzuki, K. Yoshino, and T. Ishibashi, “W-band uni-travelling-carrier photodiode module for high-power photonic millimeter-wave generation”, Electronics Letters, Vol. 38, No. 22, pp. 1376-7, 2002.

17. H. Ito, T. Furuta, T. Ito, Y. Muramoto, K. Tsuzuki, K. Yoshino, and T. Ishibashi, “W-band uni-travelling-carrier photodiode module for high-power photonic millimetre-wave generation”, Electronics Letters, Vol. 38, No. 22, pp. 1376-7, 2002.

18. B. Leone, V. Krozer, M. Feiginov, H. Roskos, H. Quast, T. Löffler, G. Loata, G. Döhler, P. Kiesel, M. Eckardt, A. Schwanhäußer, T. O. Klaassen, P. Lugli, “Optical Far-IR wave Generation - An ESA review study”, Proceedings 14th International Symposium on Space Terahertz Technology, Tucson, USA., April 22-24, 2003.

19. G.H. Döhler, M. Eckardt, A. Schwanhäusser, F. Renner, L. Robledo, A. Friedrich, P. Pohl, S. Malzer, P. Kiesel, D. Driscoll, M. Hanson, A.C. Gossard, “Ballistic transport w nanostructures used for novel THZ-emitters”, 2002 Conference on Optoelectronic and Microelectronic Materials and Devices, pp. 275-280, 2002.

20. M. Eckardt, A. Schwanhäusser, F. Renner, L. Robledo, A. Friedrich, P. Pohl, P. Kiesel, S. Malzer, G.H. Döhler, D. Driscoll, M. Hanson, A.C. Gossard,” Novel concept for efficient THz-emitters based on quasi-ballistic transport in an asymmetric superlattice”, Physica E 17 (2003) 629 – 630.

21. P. Pohl, F. Renner, M. Eckardt, A. Schwanhäusser, A. Friedrich, O. Yuksekdag, S. Malzer, G.H. Döhler, P. Kiesel, D. Driscoll, M. Hanson, A.C. Gossard,” Enhanced recombination tunneling in GaAs pn junctions containing low-temperature-grown-GaAs and ErAs layers”, Applied Physics Letters, Vol. 83, No. 19, pp. 4035-4037, 2003. 22. J. Faist, F. Capasso, D. L. Sivco, C. Sirtori, A. L. Hutchinson, and A. Y. Cho, Science, 264, 553 (1994).

23. G. Scamarcio, F. Capasso, C. Sirtori, J. Faist, A.L. . Hutchinson, D. L. Sivco, and A. Y. Cho, Science, 276, 773 (1997).

24. F. Capasso, C. Gmachl, R. Paiella, A. Tredicucci, A. L. Hutchinson, D. L. Sivco, J. N. Baillargeon, A. Y. Cho, and H. C. Liu, IEEE J. Sel. Top. Quantum Electron. 6, 931 (2000).

25. D. Hofstaetter, M. Beck, T. Aellen, J. Faist, U. Oesterle, M. Illegems, E. Gini, and H. Melchior, Appl. Phys. Lett.

78, 1964 (2001).

26. J. Faist, F. Capasso, D. L. Sivco, S. N. G. Chu, and A. Y. Cho, Appl. Phys. Lett. 72, 680 (1998).

27. A. Tredicucci, C. Gmachl, F. Capasso, D. L. Sivco, A. L. Hutchinson, and A. Y. Cho, Appl. Phys. Lett. 74, 638 (1999).

28. R. Köhler, A. Tredicucci, F. Beltram, H.E. Beere, E.H. Linfield, A.G. Davies, D.A. Ritchie, R.C. Iotti, F. Rossi, Nature 417, 156 (2002).

29. B.S. Williams, S. Kumar, H. Callebaut, Q. Hu, Reno, J.L.,” 3.4 THz quantum cascade laser operating above liquid nitrogen temperature”, Electronics Letters, Vol. 39, No. 12, pp. 915-916, 2003.

30. G. Scalari. Private communication, 2004.

31. B.S. Williams, S. Kumar, H. Callebaut, Q. Hu, , Appl. Phys. Lett.83, 5142 (2003).

32. B.S. Williams, S. Kumar, H. Callebaut, Q. Hu, , J.L. Reno, Appl. Phys. Lett.83 , 2124 , (2003).

33. S. Barbieri, J. Alton, J. Fowler, H.E. Beere, E.H. Linfield, D.A. Ritchie, Proceedings of the Photonic West SPIE 5354 Conference, San Jose, U.S.A., 25-26 January 2004.

34. I.V. Altukhov, E.G. Chorkova, M.S. Kagan, K.A. Korelev, V.P. Sinis, F.A. Smirnov, Sov. Phys. JETP 74(1992)404.

35. Yu. P. Gousev, I.V., Altukhov, K.A. Korelev, V.P. Sinis, M.S. Kagan, E.E. Haller, M.A. Odnoblyudov, I.N. Yasseievich, K.-A. Chao, “Widely tunable continous-wave THz laser,” Appl. Phys. Lett. 75 (1999), pp. 757.

36. I.V. Altukhov, M.S. Kagan, K.A. Korelev, V.P. Sinis, E.G. Chorkova, “Resonant acceptor states and terahertz stimulated emissionof uniaxially strained germanium, ” JETP 88 (1999), pp. 51.

37. M.A. Odnoblyudov, I.N. Yasseievich, M.S. Kagan, Yu. M. Galperin, K.-A. Chao, “Population inversion induced by resonant states in semiconductors,” Phys. Rev. Lett. 83 (1999), pp. 644.

38. I.V. Altukhov, M.S. Kagan, Yu. P. Gousev, I.V. Altukhov, “Continuous stimulated THz emission due to intra-centre population inversion in uniaxially strained germanium,” Physica B272 (1999), pp. 458.

39. M.S. Kagan, I.N. Yasseievich, “Resonant acceptor states and stimulated THz emission in semiconductors and semiconductor structures,” Proceedings of the 11th Internat. Symp. on Ultrafast Phenomena in Semiconductors, Vilnius, 2001.