Influence on plasmonic mixing efficiency

In Chapter 2 some fundamentals of plasma wave mixing at frequencies aboveωτ_p 1 were discussed. An efficiency factor f(ω) was introduced to account for plasmonic enhancement of distributed resistive mixing. In Section 5.1 we presented simulations of the enhancement factor in the case of gate-source mixing. Both these earlier discussions were based on precisely defined drain-coupling boundary conditions, which were employed to derive the analytic forms of f(ω) in Eqs. (2.17) and (2.18).

Moreover, it was assumed that all available radiation power was coupled to the intrinsic gated region of the FET and therefore available for the mixing mechanism.

Results in the previous section, on the other hand, showed that when ungated detector elements are included, the exact satisfaction of boundary conditions may not be valid anymore depending on the extrinsic device architecture. In such cases, analytic solutions for the enhancement factor f(ω) in general may not exist. We also showed that significant amounts of power can be lost to the ungated resistances. It is one of the most important advantages of the implementation of the TeraFET transport

5.2 Circuit simulations with ungated access regions

model in in a numerical circuit solver that simulations of real device situations can be performed.

The influences of extrinsic detector elements on the exact radiation coupling and power distribution is naturally included in simulation results from the circuit solver when such elements are added to the circuit model. In Section 4.2.3 it was shown that the intrinsic efficiency factor f(ω) could be retrieved from circuit simulations as the ratio of quasi-static and high frequency current response (cp. Eq. (4.22)).

Accordingly, a total efficiency factorf_tot(ω) can be extracted from simulations, which then not only reflects plasmonic mixing efficiency but also extrinsic influences on the overall detection process.

Figure 5.7 shows simulations of f_tot(ω) for the drain-coupling scheme displayed in Fig. 4.4 with the ungated parts activated (in contrast to Fig. 4.5) and for different applied gate bias voltages. As expected already from the discussion in the previous section, for all gate biases the mixing efficiency is always reduced compared to the intrinsic case – indicated as dashed lines in the figure - due to partial distribution of incident power to the ungated channel parts. Note that the efficiency decreases towards higher gate voltages above the threshold towards V_G = 0 V, which reflects the rising influence of the drain-side ungated part and decreasing impedance of the gated region according to the overall power distribution shown before in Fig. 5.3.

Again, the message here is that reduction of the ungated regions should be aimed for during device design in order to prevent negative influence on the plasmonic detection efficiency of TeraFETs.

TeraFET characterization

TeraFET have up until today been successfully implemented in various material systems and high sensitivity THz detection capability has been demonstrated. After a number of initial, pioneering works [106]–[108] the first thorough detection experi-ments were performed in the early to mid-2000’s using commercial GaAs/AlGaAs and GaN/AlGaN HEMTs [67], and Si MOSFETs [109], [110] for the non-resonant detection of THz radiation. In some studies, the observation of resonant plasmonic detection under non-zero drain-bias conditions was claimed [65], [66], [111]. The first devices employed for proof-of-principle experiments, however, were not intentionally fabricated for the purpose of THz detection, and in particular, the radiation cou-pling situation was mostly undefined. It was supposed in most of these works that bonding wires, contact pads for electrical contacting of the transistors, and other electrical leads or wires served as largely unspecified “antennas”. Nevertheless, the plasma wave detection principle proposed by Dyakonov and Shur [18] was successfully demonstrated and investigated in some detail in these early works.

In the succeeding years, design and fabrication of specialized TeraFET devices intended for THz detection, and particularly the integration of efficient antenna structures – resonant and broadband designs - has led to a tremendous improvement of TeraFET performances in terms of responsivity and sensitivity. Eventually, recently fabricated TeraFETs exhibit sensitivities well comparable to other state-of-the-art THz detector technologies, foremost Schottky diode-based receivers for the lower THz regions [20], [21], [43], [71]. The most sensitive (resonant antenna) devices have so far been realized in Si MOSFET technology, which is mainly due to the maturity and commercial availability of this semiconductor technology. A wide range of literature on the above experimental works exists and cannot be presented extensively in the scope of this thesis. Some review-type publications are, e.g., Refs. [20]–[22], [71].

In the course of this work, a number of TeraFETs with integrated broadband antennas were designed and fabricated in different materials with two main inten-sions: (a) to optimize the detection performance by careful design and modeling of the TeraFETs, and (b) to further investigate the underlying physical detection mechanisms in the devices. The selected material systems were AlGaN/GaN HEMTs and graphene FETs with integrated broadband antennas. During the characteriza-tion experiments, addicharacteriza-tional physical deteccharacteriza-tion mechanisms became evident, which cannot not be explained by the intrinsic plasma wave mixing principle [18], [19],

6.1 TeraFET figures of merit

[30], [34], [41], [42], [47], [48], [74], [90] alone. We argued before that these signals are thermoelectric signals due to local heating of charge carriers in the transistor channel.

This chapter presents results from THz detection experiments of the fabricated AlGaN/GaN and graphene TeraFETs in terms sensitivity of THz detection. Details of detector design and fabrication, the evaluation procedures of DC and THz mea-surements, and important critical issues therein are discussed. The focus of this chapter lies on the achievements in record detection sensitivity for both technologies, which could be reached by careful design based on the theoretical considerations of the previous chapters. The subsequent chapter then discusses in particular the obser-vation of the thermoelectric signal contributions, which are compared to predictions from the implemented device model.

6.1 TeraFET figures of merit

When characterizing and evaluating the performance of radiation detectors, two important figures of merit should be considered. On one hand, the responsivity describes the detectors radiation response signal relative to the amount of available radiation power - it is thus a quantity reflecting the efficiency of the underlying detection mechanism in the device. On the other hand the noise equivalent power (NEP) relates the detector’s responsivity to it’s inherent noise level, the NEP is thus

understood as a measure of the detector’s sensitivity.

We discussed before that the rectifying detection mechanisms in TeraFETs produces a measurable DC current I_det(ω) or voltage V(ω) response to an incident high frequency radiation signal (cp. Chapter 2). Both quantities are related to each other following Ohm’s law by the detector’s DC drain-source resistance

V_det =I_det·R_DC (6.1)

The magnitude of the measured responses depends on the available power of incident radiation, hence, in order to compare the performance of multiple detectors, the current responsivity and voltage responsivity RI and RV are defined as

RI = I_det P_in R_V = V_det

P_in = I_det P_in ·R_DC

(6.2)

The responsivity relates a measured THz response signal to the THz power available to the detector. Therefore, it is a measure for the efficiency of detection. Note that saturation effects may occur for very high radiation power levels and the dependence of responsivity on the incident power may no longer be linear in such cases. Also, it cannot be stressed enough here that a fundamental aspect in the concept of responsivity is the definition of the incidenct powerP_in and the exact meaning of the term “available power”. A detailed discussion on the topic follows in Section 6.1.2.