More than ten years ago, Dyakonov and Shur proposed using submicron field effect transistors (FETs) as sources and detectors of terahertz electromagnetic radiation [1, 2]. Since that time both detection and generation in the sub-THz and THz frequency range have been experimentally demonstrated using compound semiconductor and silicon transistors [3–10]. For detection, the regimes of the resonant and non-resonant operation are distinguished. When wt<< 1,the FET operates in the non-resonant regime, where o is the radian frequency and t the plasmon decay time; the transistor response to the electromagnetic radiation is a decreasing function of the gate voltage. When wt>> 1,the transistor has a resonant response at resonant plasma frequencies Wn=Wo(1+2N), where N= 0, 1, 2, . . . , Wo= (pi=2Lg).SQR(e^2.ns/Cm) is the fundamental radian plasma frequency, Lg is the gate length, ns is the channel concentration, C is the gate capacitance per unit area, m is the effective mass, and e is the electron charge. The decay time is equal to or shorter than the scattering rate. It might be affected by the ballistic transport [11], the viscosity of the electron fluid owing to electron-electron collisions [1], and by a possible effect of oblique plasma modes [12].
Since the channel concentration is a function of the gate voltage, the FET in the resonant mode of operation is a tunable detector of THz radiation. In spite of the experimental demonstration of resonant sub- THz detection by GaAs-based transistors (see [7] and references therein) and THz detection by InGaAs-based transistors [13], this mode of operation remains a challenge to achieve. Since GaN/AlGaN heterostructure FETs (HFETs) have extremely high electron concentration in the channel (exceeding 1013cm 2), these transistors with gate length Lg <>
One of the most important parameters of the detectors is noise equivalent power (NEP), which can be found as N/RV, where N is the noise of the transistor in V/Hz^0.5 and RV is the responsivity in V/W. Fig. 2 shows the NEP for transistors with Lg=250 nm at 300 K for frequencies of radiation f=0.2 and 0.7 THz. The responsivity, which is the ratio of the output in volts to the radiant input in watts, was estimated in two different ways. For f=0.7 THz, the responsivity was estimated as RV=DV/(PtSa/St), where Pt is the total power of the source, St is radiation beam spot area, and Sa is the transistor area, which includes contact pads. In the experiment with 0.2 THz, the radiation was focused to the diameter approximately of the same size as the transistors, including the contact pads. Therefore, the total power of the source was taken for the NEP estimate. Since detection was studied at zero bias, the noise was taken equal to the thermal noise N=SQR(4kTRfet), where RFET is the gate voltage dependent drain to source resistance, which can be extracted from the transfer current voltage characteristic of the field effect transistor. Fig. 2 shows the NEP at 300 K against gate voltage. The inset in Fig. 2 shows the responsivity for the 0.2 THz experiment as a reference. As seen, in spite of a relatively low responsivity, the minimum NEP is of the order of 5 10 9 W/Hz^0.5. The minimum NEP corresponds approximately to the voltage, where the responsivity is the largest. This value of NEP is slightly higher than for such commercial detectors as Golay cells, pyroelectric detectors and Schottky diodes, having, meanwhile, the potential advantage of operation at very high sampling frequency of several tens of gigahertz.
Conclusions: We have investigated different GaN/AlGaN devices and demonstrated an efficient detection of electromagnetic radiation at THz frequencies. While varying the temperature in the range from 4 K to 300 K and the excitation frequency within 0.2–2.5 THz, we have shown the resonant detection due to excitation of the plasmon modes. These detectors demonstrate reasonably low noise equivalent power suitable for practical applications.
Fuente:
Gerald Soto, CRF 2010-1.
No hay comentarios:
Publicar un comentario