Introduction
Terahertz Wave and Applications
Due to the excellent characteristics of the THz wave such as good permeability (penetrability) and innocuousness to the human body due to its relatively long wavelength and low energy, the research on radiation and detection of the THz wave has been intensively carried out in the past year due to their potential applications in the various fields, including communication, spectroscopy and THz imaging technology [1-5]. THz imaging such as biomedical, security, impurity inspection is the most promising application of THz detection in the near future.
Plasmonic Terahertz Wave Detector
- Plasma Wave in FETs
- Operation Principle of Terahertz Detector
- Issues of Research on Terahertz Detector
Since Dyakonov and Shur proposed the behavior of plasma waves in the THz detector channel based on FETs [11, 12], THz detectors containing two-dimensional electron gas (2DEG) FET devices were theoretically reported [13-15]. and experimentally [16]. The plasma wave has a speed of s = (eU/m)1/2 and the propagation distance of the plasma wave is sτ. The plasma wave on the source side can reach the drain side and be reflected, increasing the amplitude of the plasma wave.
The plasma wave propagation distance l, which is an important parameter of a non-resonant THz detector, is calculated as s(τ /ω)1/2. The plasma wave can reach the drain side of the FET and be reflected. When ωτ < 1, the plasma wave cannot exist due to excessive damping and the FET operates in the non-resonant regime. a) Asymmetry of the 2DEG plasma distribution in the FET channel region for THz detection v.
There are three parts to a FET-based THz detector: the antenna, the FET, and the amplifier.
Motivation
6), the best RV can be obtained by making the output ∆U level of the FET as high as possible under the same input power from the THz source. 7), the best NEP can be achieved by keeping the total noise of the detector as low as possible under the same RV. The key factor for the highest RV and the lowest NEP is the photoresponse ∆U of the FET stage, which means that the ∆U of the FET (not the RV of the detector) should be up to the breakthrough level along with reducing the total noise of the detector.
Thesis Overview
This phenomenon leads to the asymmetry of the electron density distribution in the channel region of the FET. DC characteristics (Vth and SSW) of the FET-based THz detector according to x. The plasma wave behavior as the 2DEG of the THz detector can be represented as the electron density in the channel of FET.
The inset shows the mean electron density point (∂(De)/∂x= -1) for the extraction of lQP and NQP. Simulation results of photoresponse Δu as a function of gate voltage as variation tox and 4 nm) with the 2DEG quasi-plasma in the channel. In the same way, the asymmetry of the non-resonant THz detector based on the new NQS MOSFET design (adding Cgd) has been implemented.
The simulation results of the new NQS model in SPICE have the same tendency as the measurements [20] and TCAD simulation results [33].
Numerical Device Modeling of Terahertz Wave Detector Based on Si FET
Modeling of Terahertz Wave Detector with Quasi-Plasma 2DEG
As shown in Figure 2-1, the n-type Si MOSFET structure was designed by TCAD simulation with several structural parameters such as gate length Lg = 300 nm, source-drain junction depth Xj = 100 nm, channel doping concentration Nch = 1 × 1018 cm-3, source and drain doping concentration Nd = 1 × 1020 cm-3, poly-Si gate doping concentration Npoly = Nd and lightly doped drain concentration (LDD) NLDD = 1 × 1019 cm-3. When the THz wave irradiates FET-based detector, a photoresponse is induced in the. This DC voltage difference between the source and the drain is caused by the electron charge asymmetry as 2DEG in the channel.
The propagation length of the plasma wave, called the characteristic length, is considered an important parameter for the electron charge asymmetry as 2DEG in the channel region. The key point to extract the photoresponse is the definition and accurate analysis of the characteristic length as 2DEG length and 2DEG density in the channel. In the non-resonant THz sensing regime (ωτ < 1), an overdamped plasma wave cannot exist on the drain side of the channel.
Therefore, we introduce a quasi-plasma electron box as 2DEG [31], which is an over-damped plasma wave in the channel, for the asymmetric boundary condition and electron density distribution in the channel.
Asymmetric Boundary Condition of Terahertz Detector
Since the construction of FETs in the frequency and time domains is expensive, the numerical solution of the continuity equation and the hydrodynamic Euler equation with Poisson's equation is necessary to accurately analyze the behavior of the 2DEG plasma wave. 2DEG density modulation contours (when tox= 2.5 nm) along with the channel at each time scale show the change in 2DEG density modulation with increasing Cgd. Contour plots of electron density modulation along with channel position at each time scale over 2 cycles.
For the symmetrical boundary condition (in Figure 2-3(a)), the plasma wave as 2DEG can propagate equally at both the source and drain sides. 2-3(b), (c) and (d), respectively. the channel electron 2DEG propagates shorter than the symmetric state near the drain side. Simulation results of the peak-to-peak amplitude of the voltage difference between gate and drain according to increasing gate-to-drain external capacitance (Cgd) at tox= 1.1 nm.
The top-to-peak amplitude of the voltage difference between gate and drain (Ua) becomes almost zero in accordance with the increase of Cgd.
TCAD Simulation Results
- DC Characteristics
- Electron Mobility in Channel Region
- Normalization of Electron Density (l QP and N QP )
Therefore, the electron mobility should be extracted in the non-resonant state of the THz detection to analyze the electron density in the channel. Using normalization of the electron density, the values of lQP and NQP are obtained according to tox. The values of lQP and NQP from normalization of the electron density along the channel position according to the different tox.
The simulation results of ∆u and NEP in the subthreshold region provide the theory of the non-resonant plasmonic THz detection. For example, as shown in Figure 2-3, the simulation results of the electron density along the channel region can be extracted by using 2D device simulation TCAD. As shown in Figure 3-8(b), the simulation results of the new SPICE NQS MOSFET model are comparable to the TCAD simulation results.
-based verification of RF simulation results. a) Experimental results of the photoresponse when the input voltage is applied.
New Non-Quasi-Static Compact Model for MOSFET-Based Terahertz Detector
Quasi-Static and Non-Quasi-Static Analysis
In general, there are four modes of physical analysis according to time variation: static, quasi-static, non-quasi-static and dynamic analysis. The signal changes slowly with respect to the device transit time so that the channel load is always in the steady state region. When the rise time of the input signal is longer than 20 times the device transit time, the QS assumption is satisfied.
In this condition, the transition time has a different equation depending on the length of the channel in the case of a long channel and a short channel. In the QS approximation, the transit time of the carrier along the FET channel region is not taken into account, so the QS approach can introduce a fatal error in the simulation results for fast switching. In this situation, the analysis method is related to the oscillation period and the passage time of the plasma wave [7].
QS and NQS assumptions were distinguished when the oscillation period is equal to the plasma wave transit time.
DC Characteristics Matching Between TCAD and SPICE
Using these parameters, I can get the basic SPICE parameters and the input parameters to extract the SPICE DC parameters (in Table IV). In SPICE, lightly doped drain (LDD) doping concentration parameters do not exist and the temperature unit is degrees (not Kelvin) unlike TCAD. For efficient parameter extraction, important SPICE matching parameters should be set to the appropriate value (in Table V).
VTH0 is threshold voltage at Vbs = 0, U0 is the electron, LINT and WINT are structure fitting parameters, and RDSW and NFACTOR for subthreshold swing are critical parameters. Since all the other parameters are less important, they can be set to the default value. DC characteristics of SPICE simulation results, both linear and saturation region, are similar to TCAD simulation results.
New Compact MOSFET Model for Non-Quasi-Static Analysis
- Conventional Non-Quasi-Static MOSFET Model
- Non-Quasi-Static Elmore Model
- Transient Simulation of New Non-Quasi-Static MOSFET Model
However, this circuit model requires a lot of computation time and has formulation complexity [43-48], so the new compact NQS model is needed for the accurate analysis of channel charging delay. The Elmore model is one of the world's most widely used NQS MOSFET models in SPICE, as shown in Figure 3-5(d). However, the NQS Elmore model is not valid for transient simulation and fast switching mechanism of MOSFET.
Therefore, a new compact NQS MOSFET model should be needed to describe the transient fast switching and high-frequency operation in the THz region. To describe the switching delay with a very short rise time, I developed the new NQS compact. To verify SPICE's new NQS MOSFET model for transient simulation, the model was compared with TCAD 2D device simulation as a reference for the full numerical NQS simulation.
These results show that our model is more accurate and reliable for the rapid switching mechanism than the existing NQS Elmore model. a) Circuit configuration of the new NQS compact MOSFET model with stacked circuit elements.
Demonstration of New NQS Compact Model for MOSFET-based THz Detector
The applied asymmetric boundary condition in simulation results of the new SPICE NQS MOSFET model by adding the different gate-to-drain capacitance Cgd. Finally, I demonstrate the validity of the new model by achieving the successful photoresponse simulation as a function of the gate voltage at 0.2 THz. The peak of the photoresponse exists in the sub-threshold region, consistent with the simplified theory of the non-resonant THz detector [16].
Finally, I demonstrate the validity of the new NQS model as a plasmonic THz detector by extracting the photoresponse, which is the performance of the THz detector. By integrating the external gate capacitance, the parasitic capacitance will be eliminated and the intrinsic delay time of the THz detector can be estimated. In this measurement, the photoresponse overshoot delay is reduced by adding external gate capacitance.
As a result, the delay of the THz detector for real-time THz imaging will be demonstrated by the delay estimation. Masetti, “Issues in precision modeling of the MOS transistor for analog applications,” IEEE Trans. Gildenblat, “Quasi-static and non-quasistatic compact MOSFET models based on symmetric linearization of the bulk and inversion charges,” IEEE Trans.
Summary and Conclusions
Future Works
