Numerical study on heterodyne terahertz detection in field effect transistor

Zhifeng Yan; Jingxuan Zhu; Yinglei Wang; Xinnan Lin; Jin He; Juncheng Cao

doi:10.1364/OE.18.007782

1. Introduction

In recent years, terahertz frequency radiation has received a lot of attention due to its potential for non-invasive imaging. Dyakonov and Shur had proposed a theory for using field effect transistors as detectors of terahertz radiation [1,2]. The operations of these detectors are based on plasma wave excited in GaAs, Si, and GaN FETs [3–6] under high modulation frequency. Plasma waves excited in un-gated two dimensional electron layers [5] and the influence of DC current on THz detection had been studied [7,8] in earlier works. FET to be used as an ultra fast detector of modulated terahertz radiation was predicted in 2007 [9]. Imaging of field-effect transistors by focused terahertz radiation was developed in 2009 [10].

In 2008, our group has developed a numerical tool for simulating plasma wave transmission in FETs based on the structure shown in Fig. 1 [11,12]. In the figure, the THz radiation induced AC voltage between gate and source is assumed to be $u_{a} \cos w t$ , and the frequency of the incoming THz signal is $w$ . The radiation induced voltage drop $Δ U$ between source and drain is considered to be the detector response.

Fig. 1 Schematic geometry of an N type MOSFET operating in detector mode

Download Full Size | PDF

A theory of nonlinear response of a field effect transistor subjected to two terahertz beams with close frequencies has been proposed in 2008 [13]. It predicts that the heterodyne efficiency can be significantly increased by the electric current flowing in the transistor channel. In addition, it has been demonstrated that such a heterodyne detector is capable to operate at very high intermediate frequencies up to 50~100GHz. However, existed simulation programs developed so far only shows the relationship between the amplitude of heterodyne signal and drain-to-source current in the strong inversion region rather than the entire region including the sub-threshold region. In the sub-threshold region, since the existence of gate leakage current in the transistor decreases the magnitude of detection response remarkably and the enhancement of magnitude emerges close to the sub-threshold voltage, developing the numerical tool covering the sub-threshold region plays a significant role in the further investigation on heterodyne terahertz detection. Thus, in this work, numerical study on heterodyne terahertz detection is made, and a numerical simulation tool for calculating the terahertz response under different incoming signal amplitudes and frequencies at all operation regions from sub-threshold to strong inversion is established. By using our developed numerical tool, not only can the validity of the theory of nonlinear response of the FET channel subjected to two terahertz beams with the close frequencies be demonstrated, but also the characteristics on heterodyne detection in sub-threshold region can be simulated.

2. Numerical model

The basic principle of the mechanism of heterodyne detection is described as follows. An incoming signal expressed with the amplitude $u_{a}$ and a frequency $w$ is mixed with a strong signal from a local oscillator with the amplitude $u_{b}$ and a frequency $w + Ω$ $(Ω < < w)$ . We assume that the voltage at the source is the sum of the fixed gate-to-channel swing and an oscillating part induced by the two beams, so that

V_{c h} (0, t) = U_{o s c} (t)

Where

U_{o s c} (t) = u_{a} \cos (w t) + u_{b} \cos ([w + Ω] t + φ)

.

φ

is a certain phase shift between two signals. The hydrodynamic equations governing the THz signal are used to describe the nonlinearities in a field effect MOS transistor [1,2].

\frac{\partial n_{s}}{\partial t} + \frac{\partial (n_{s} V)}{\partial x} = \frac{j_{g}}{e}

\frac{\partial V}{\partial t} + V \frac{\partial V}{\partial x} + \frac{V}{τ_{m}} + \frac{e}{m} \frac{\partial V_{c h}}{\partial x} = 0

In the equations, $n_{s}$ is the electron density, $V$ is the average velocity, $e$ is the elementary charge, $τ_{m}$ is the electron momentum relaxation, $m$ is the electron effective mass and $V_{c h}$ is the quasi Fermi potential. $j_{g}$ is the density of the gate leakage current which is a constant. The general equation for the electron density in the FET’s channel is given by [3]

n = n^{*} \ln [1 + \exp (\frac{e (U - V_{c h})}{η k_{B} T})], n^{*} = C η k_{B} T / e

where

U = U_{g s} - V_{t h}

,

U_{g s}

is gate bias,

V_{t h}

is the threshold voltage,

C

is the gate capacitance per unit area and

η

is the ideality factor. To simplify the formulation, the following normalization is made:

n = n_{s} / n_{s 0}, v = V / s, η = x / L, τ = t s / L, τ_{m}^{'} = τ_{m} s / L, v_{c h} = V_{c h} / v,

j = j_{g} L / (e n_{s 0} s)

where

s = \sqrt{e V_{t} / m}

,

V_{t}

is the thermal voltage and

L

is the channel length of the FET detector. After the normalization, the following dimensionless equations are obtained:

\begin{array}{l} \frac{\partial n}{\partial τ} + \frac{\partial (n v)}{\partial η} - j = 0 \\ \frac{\partial v}{\partial τ} + v \frac{\partial v}{\partial η} = \frac{\partial v_{c h}}{\partial η} - \frac{v}{τ_{m}^{'}} \end{array}

where

n_{s 0}

is the electron density in the FET’s channel when the detector is not subjected to radiation. Define two variables

q_{1} and q_{2}

as

q_{1} = v_{c h}, q_{2} = n v

The following expressions are obtained

F_{1} (q_{1}, q_{2}) = [\begin{matrix} n \\ \frac{q_{2}}{n} \end{matrix}], F_{2} (q_{1}, q_{2}) = [\begin{matrix} q_{2} \\ \frac{q_{2}^{2}}{2 n^{2}} - q_{1} \end{matrix}], G (q_{1}, q_{2}) = [\begin{matrix} - j \\ \frac{q_{2}}{n τ_{m}^{'}} \end{matrix}]

Using the forward-difference method in the time domain and central-difference method in the spatial domain, one gains the difference equation:

\frac{{(F_{1})}_{k}^{n + 1} - {(F_{1})}_{k}^{n}}{Δ τ} + \frac{{(F_{2})}_{k + 1}^{n + 1} - {(F_{2})}_{k - 1}^{n + 1}}{2 Δ η} + {(G)}_{k}^{n + 1} = 0

where

Δ τ

and

Δ η

are the time step and incremental spatial separation respectively. As the quasi-Fermi potential (which is fixed at

u_{a} \cos (w t) + u_{b} \cos ([w + Ω] t + φ)

) is only induced by radiation at the source,

{(q_{1})}_{1}^{n + 1}

is obtained. The left boundary condition is described by using the first part of the Eq. (5)

\frac{{(F_{1, 2})}_{1}^{n + 1} - {(F_{1, 2})}_{1}^{n}}{Δ τ} + \frac{{(F_{2, 2})}_{2}^{n + 1} - {(F_{2, 2})}_{1}^{n + 1}}{Δ η} + {(G_{2})}_{1}^{n + 1} = 0

By fixing the DC current of drain at $j_{d}$ , the left boundary ${(q_{1})}_{N + 1}^{n + 1}$ is obtained. Similarly, the right boundary condition is described as

\frac{{(F_{1, 1})}_{N + 1}^{n + 1} - {(F_{1, 1})}_{N + 1}^{n}}{Δ τ} + \frac{{(F_{2, 1})}_{N + 1}^{n + 1} - {(F_{2, 1})}_{N}^{n + 1}}{Δ η} = 0

The response of the radiation induced by

V_{d s}

can be gained from the numerical program developed to solve the difference Eqs. (7), (8) and (9). The implicit numerical of this numerical model is stable for a wide range of conditions.

Here we have chosen to use the backward Euler method to carry out iterative calculation. In analyzing the error of the numerical method, it is found that the oscillation tends toward stability and eventually forms the stable oscillation, and it could attain dozens of the order of magnitudes at a determinate time step. Thus the analysis on the astringency of algorithm is crucial. According to the LAX equivalence theory, the stability is the necessary and sufficient condition of astringency. By keeping the $Δ τ / Δ η = K$ constant, the compatibility of the difference and discrete form of two basic questions can be guaranteed. The stability of algorithm can be ensured with the advancing of time steps. Similarly, the methods for discretion of fundamental equation set-up basing on Eq. (4) are compatible as well. Following the same analysis methods, it is drawn that the solutions of the differential equations can always be close to their genuine solutions under an adequate precision.

When it comes to the question of adding boundary conditions, since there are only two differential equations, and two boundary conditions were needed. However, due to the limits of the difference algorithm itself, the current density should be given on the left margin, and the electron concentration given on the right sector. So the situation of adding boundary conditions will also affect the accuracy of the algorithm. Here we use a stable implicit algorithm, and this algorithm is used in the current boundary layer grid points, the next boundary layer grid points and the boundary adjacent grid points. The stability of such boundary conditions is much higher than that of the explicit algorithm as well as the leap-frog scheme.

3.Result and discussions

The instantaneous source-to-drain voltage $V_{d s} (t)$ contains all information of the plasma wave excitation/detection and mixing details. By omitting high-order harmonics, $V_{d s} (t)$ can be described by [13,14]:

\begin{array}{l} V_{d s} (t) = V_{D S} + δ V^{b} (Ω + w) \cos [2 π (Ω + w) t + ϕ^{b} (Ω + w)] \\ + δ V^{a} (w) \cos [2 π w t + ϕ^{a} (w)] + \overset{— —}{Δ V} (w, w + Ω) \\ + δ V^{m} (w, Ω + w) \cos [2 π Ω t + ϕ^{m} [w, (Ω + w)]] \end{array}

where

V_{D S}

is the DC part of the source-to-drain voltage which is independent of the frequencies with both

u_{a} = 0

and

u_{b} = 0

. The next two terms are the harmonic responses induced by the two incoming signals, respectively. The last two terms are related to nonlinearity of the system.

\overset{— —}{Δ V} (w, w + Ω)

is the additional DC voltage that appears due to a rectification effect of the electrical excitations and the last term describes the effect of mixing which provides oscillations at a different frequency

δ f = Ω

[14].

First, we simulated a field effect transistor with gate oxide thickness $d = 2 n m$ , substrate doping concentration $5 \times 10^{15} c m^{- 3}$ and gate length $L = 1 μ m$ . The temperature is fixed at 300K in all simulations. The threshold voltage is set at $V_{t h} = 0.7 V$ and by applying an average gate bias $U_{g s} = 1 V$ , the mean swing voltage is $U = 0.3 V$ . Parameters of the fixed gate-to-channel swing and the oscillating signal induced by two beams are chosen as following: $u_{a} = 0.004 V, u_{b} = 0.005 V, w = 1.382 \times 10^{11} r a d / s$ , $Ω = 1.382 \times 10^{10} r a d / s, φ = 0$ . Figure 2 shows (a) the drain voltage $V_{d s} (t)$ and (b) the amplitude-frequency characteristic of the photoresponse corresponding to the spectrum of the drain voltage near the frequencies $w$ and $w + Ω$ . The two components at frequency of $w$ and $w + Ω$ are called the harmonic responses induced by two incoming signals. Then a low-pass filter is used to separate the photoresponse from the other frequency components and the corresponding results are shown in Fig. 3(a) . According to the theory introduced in [13], the phase of the intermediate signal $ϕ^{m} [w, (w + Ω)]$ is equal to the phase difference $φ$ of the signal and the phase of the local oscillation with a small correction $Ω t_{0}$ , here $t_{0}$ is the time delay and $t_{0} ~ L^{2} / μ U$ . Both $U_{g s}$ and $V_{D S}$ are fixed and the THz detection spectrum is picked up by the FET spectrometer. The predicted time delay is not included in the red curve so as to show the time resulting from the heterodyne detection clearly. The amplitude of the photoresponse is shown in Fig. 3(b). The component at the frequency $Ω$ is the signal generated at the intermediate frequency (IF) [14].

Fig. 2 (a) Source-to-drain voltage $V_{d s} (t)$ (b) Amplitude-frequency characteristic of source-to-drain voltage

Download Full Size | PDF

Fig. 3 (a) Comparison between the results from simulation and existing theory (b) Amplitude of the photoresponse

Download Full Size | PDF

In the preceding example, we showed the typical detection spectra at a gate bias $U_{g s} = 1 V$ . Now, we change the gate bias from 0.8V to 1.15V. Figure 4(a) shows that the photoresponse is modulated by the gate bias. Figure 4(b) shows the dependence of time delay $t_{0}$ and the amplitude of the response $δ V^{m} (w, w + Ω)$ as a function of the gate bias. It is observed that both the time delay of the response and the dc component decrease as gate bias $U_{g s}$ increases. The black and green dotted lines are extracted from the simulation and the red and blue dotted lines are calculated from the theory. It is observed that the numerical simulations have excellent agreement with the analytical prediction.

Fig. 4 (a) Photoresponse as a function of the time for four different gate biases (b) Time delay and amplitude of photoresponse

Download Full Size | PDF

To estimate the influence of the electron mobility $μ$ on the response of the MOSFET detector, we will use the output characteristics of the transistors as shown in Fig. 5 (a) . In  Fig. 5(b), the numerical time delays at different relaxation times, are compared with the theoretical curves derived from the output characteristics of transistor. Comparison between the theoretical calculation and numerical simulation in Fig. 5(b) indicates that the time delay decreases while $μ$ increases, and the simulation results agree with the theory. Besides, the amplitude of the response from the numerical simulation agrees with the theoretical results in the case of small electron mobility. However, for high electron mobility, the numerical results are a little larger than the theoretical results. It is because the theory introduced in [13] is restricted to the condition $L \sqrt{2 w / μ U} > > 1$ . With the increase in electron mobility, such condition will not be satisfied. The numerical simulation reflects the true THz detection in FET-based terahertz detectors more consistently.

Fig. 5 (a) Photoresponse of different electron mobility (b) Time delay and amplitude of photoresponse

Download Full Size | PDF

All the above simulations are restricted in the strong inversion region. Our simulator can also be used in the sub-threshold region while $U < 0$ [3]. In the experiment of THz detection in Ref. [3], for $U_{g s}$ smaller than the threshold voltage (negative values of $U$ ), the experimentally observed response decreases, so that the response always reaches a maximum value close to the FET threshold. Due to the fact that no systematic numerical tool for the investigation on heterodyne terahertz detection has been established till now, developing the numerical study on heterodyne terahertz detection in the sub-threshold region is necessary. Electron concentration in the FET channel becomes very small in such situation. The leakage current plays an important role and should be taken into account [3]. In Fig. 6(a) , four different gate leakage current densities $j_{g}$ (0A/m², 1A/m², 10A/m² and 100A/m²) are chosen to study their influences on the response, here $η$ is fixed at 3. When the gate leakage current density is neglected, the amplitude of the response in sub-threshold is independent on the gate bias. According to the theory [13], the amplitude of the response is just inversely proportional to the temperature. The analytical theory cannot reflect the actual situation when there is a gate leakage current. And the previous theory does not include the situation that gate voltage is near the threshold voltage, while the numerical tool constructed by us not only considers the influence of the gate current to response in the sub-threshold areas, but also has a good simulation capability in the situation near the threshold voltage. In Fig. 6(a), it is clear that when the gate bias decreases, the amplitude of the response rapidly decreases in the sub-threshold region, and the peak value of the amplitude appears when $U_{g s} = V_{t h}$ . A maximum value near zero swing voltage can be obtained by a decreasing leakage current, and the maximum broadens and shifts toward negative swing biases. A highland is gained at low values of the gate current. Gate leakage current does not influence the amplitude of response in strong inversion region. Well above threshold, the gate leakage is small compared to the current in the channel and does not affect the response of the detector. It may be interesting to study how the photoresponse of the MOSFET detector is influenced by the temperature. Three different temperatures (300K, 200K and 100K) are chosen for investigating their influences on the response. In Fig. 6(b), the peak amplitude of the photoresponse increases with the decrease of temperature. With decreasing temperature, the peak value shifts to smaller gate bias. That is because the threshold voltage decreases when the temperature decreases.

Fig. 6 (a) Amplitude of the photoresponse as a function of gate bias for four different gate leakage current densities (b) Amplitude of the photoresponse as a function of gate bias in three different temperatures

Download Full Size | PDF

The numerical description, presented above, shows that parameters necessary to describe the heterodyne terahertz detection of field effect transistor are: (i) electron mobility (ii) the threshold voltage $V_{t h}$ (iii) the leakage current. Therefore, the development of numerical tools is instructive to the study of the device structure about changing parameters on the heterodyne detection.

4. Conclusion

In this paper, a numerical method on heterodyne detection has been investigated which is used for calculating the terahertz signal transport in field effect MOS transistors basing on the hydrodynamic equations. By implementing the heterodyne characteristics into the fundamental framework of a numerical tool established by our group in 2008, a comprehensive numerical method particularly for heterodyne detection is formed. By comparing the numerical results with the existing theory, the validity of the proposed numerical method has been proved, and the developed numerical tool has been demonstrated to work well in all operation regions of field a MOS transistor with continuous transit from the sub-threshold to the strong inversion and from the linear to the saturation regions. Thus the proposed numerical method has potential to be a useful tool for studying the nonresonant detection response by changing certain parameters of field effect MOS transistors.

Acknowledgments

This work is supported by the National Natural Science Funds of China (60976066), the Shenzhen Science & Technology Foundation (JSA200903160146A) and the Open Project of State Key Laboratory of Functional Materials for Informatics.

References and links

1. M. Dyakonov and M. Shur, “Shallow water analogy for a ballistic field effect transistor: New mechanism of plasma wave generation by dc current,” Phys. Rev. Lett. 71(15), 2465–2468 (1993). [CrossRef] [PubMed]

2. M. Dyakonov and M. Shur, “Detection, Mixing, and Frequency Multilication of Terahertz Radiation by Two-Dimensional Electronic Fluid,” IEEE Trans. Electron. Dev. 43(3), 380–387 (1996). [CrossRef]

3. W. Knap, V. Kachorovskii, Y. Deng, S. Rumyantsev, J. Q. Lu, R. Gaska, M. S. Shur, G. Simin, X. Hu, M. A. Khan, C. A. Saylor, and L. C. Brunel, “Nonresonant detection of terahertz radiation in field effect transistors,” J. Appl. Phys. 91(11), 9346 (2002). [CrossRef]

4. A. El Fatimy, S. Boubanga Tombet, F. Teppe, W. Knap, D. B. Veksler, S. Rumyantsev, M. S. Shur, N. Pala, R. Gaska, Q. Fareed, X. Hu, D. Seliuta, G. Valusis, C. Gaquiere, D. Theron, and A. Cappy, “Terahertz detection by GaN/AlGaN transistors,” Electron. Lett. 42(23), 1342 (2006). [CrossRef]

5. W. Knap, F. Teppe, Y. Meziani, N. Dyakonova, J. Lusakowski, F. Boeuf, T. Skotnicki, D. Maude, S. Rumyantsev, and M. S. Shur, “Plasma wave detection of sub-terahertz and terahertz radiation by silicon field-effect transistors,” Appl. Phys. Lett. 85(4), 675 (2004). [CrossRef]

6. R. Tauk, F. Teppe, S. Boubanga, D. Coquillat, W. Knap, Y. M. Meziani, C. Gallon, F. Boeuf, T. Skotnicki, C. Fenouillet-Beranger, D. K. Maude, S. Rumyantsev, and M. S. Shur, “Plasma wave detection of terahertz radiation by silicon Field Effects Transistors: Responsivity and Noise Equivalent Power,” Appl. Phys. Lett. 89(25), 253511 (2006). [CrossRef]

7. M. Dyakonov and M. S. Shur, “Current instability and plasma waves generation in ungated two-dimensional electron layers,” Appl. Phys. Lett. 87(11), 111501 (2005). [CrossRef]

8. D. Veksler, F. Teppe, A. Dmitriev, V. Kachorovskii, W. Knap, and M. Shur, “Detection of terahertz radiation in gated two-dimensional structures governed by dc current,” Phys. Rev. B 73(12), 125328 (2006). [CrossRef]

9. V. Yu. Kachorovskii and M. S. Shur, “Field Effect Transistor as ultrafast detector of modulated terahertz radiation,” Solid-State Electronics , (2007).

10. D. B. Veksler, A. V. Muravjov, V. Yu. Kachorovskii, T. A. Elkhatib, K. N. Salama, X.-C. Zhang, and M. S. Shur, “Imaging of field-effect transistors by focused terahertz radiation,” Solid-State Electron. 53(6), 571–573 (2009). [CrossRef]

11. X. Mou, Y. Chen, C. Ma, Y. Chen, and J. He, “A Numerical Method to Simulate THz-Wave Generation and Detection of Field-Effect Transistors,” IEEE ICSICT , 396–399 (2008).

12. Y. Wang, Y. Chen, F. He, X. Mou, and C. Ma, “Numerical Study on Detection Response of Field Effect MOS Transistor to Modulated Terahertz Radiation Signal,” IEEE ELECTRO-2009, India, (2009).

13. B. Gershgorin, V. Yu. Kachorovskii, Y. V. Lvov, and M. S. Shur, “Field effect transistor as heterodyne terahertz detector,” Electron. Lett. 44(17), 1036 (2008). [CrossRef]

14. H. Marinchio, L. Chusseau, J. Torres, P. Nouvel, L. Varani, G. Sabatatini, C. Palermo, P. Shiktorov, E. Starikov, and V. Gruzinskis, “Room-temperature terahertz mixer based on the simultaneous electronic and optical excitations of plasma waves in a field effect transistor,” Appl. Phys. Lett. 96, 013502 (2010). [CrossRef]

Numerical study on heterodyne terahertz detection in field effect transistor

Abstract

1. Introduction

2. Numerical model

3.Result and discussions

4. Conclusion

Acknowledgments

References and links

Cited By

Figures (6)

Equations (11)

Optics Express