LITERATURE SURVEY Last few decades reveal a considerable amount of effort by researchers in modeling and characterization of optically controlled MOSFET/MISFET as a photodetector. The major works reported in this area are reviewed extensively and presented in this chapter. Sections given below presents a summary of the major works reported on modeling and characterizing of optically controlled devices under the following headings: MOSFET modeling Requirements of a photodetector (i.e. gate material, device structure and optical widow) Optical Modeling and DC characterization of Silicon based Photodetectors AC analysis of Silicon based photodetectors Noise analysis InGaAs MISFET MISFET photodetector 0.1 MOSFET Modeling In the last decade, the IC business has grown explosively. Large numbers of circuit designers work without having frequent interaction with the process and foundries. They have access to fast prototyping as compared to manufacturing engineers to design VLSI chips. Circuit designers rely on the simulation of their design before building a prototype. To simulate a circuit, simulators make use of element models, which provide a mathematical description of the element behavior in the circuit. Computer Aided Design (CAD) tools are essential elements for circuit design. The productivity of circuit designers is intimately associated with the efficiency of the available arsenal of CAD tools. One of the most useful IC design tools is the circuit simulator, which allows users to enhance the understanding of a circuit and some fine details of its operation. The judicious use of electrical simulators allows quick evaluation of the circuit performance without the burden of a costly integrated prototype. Nowadays there are different CAD tools available in the market for circuit device and simulation such as TCAD, CADENCE, MATLAB, LabView, Microwind and many more. One must bear in mind, however, that the accuracy of the results provided by the simulator depends on the quality of the circuit element model. Like wise, MOSFET models, which serve as the critical communication vehicle between circuit designers and silicon foundries [11], play a crucial role in chip design productivity. In this way, modeling and simulation of the integrated circuit (IC) is an essential part prior to its fabrication. Number of models are developed to understand operations of MOSFET like SPICE, BSIM, MOS Models 9/11, EKV, ACM, PSP etc [14]. Table 2.1 illustrates pros and limitations of these models. These models are used by circuit simulators which help them to develop circuits efficiently. Thus device models act as a bridge between the IC designers and foundries. These models are mainly classified as numerical models and compact models. Numerical models are based on solving the partial differential equations describing the detailed physics of the device and compact models describe the device in a simplified manner [15,16]. The compact 1
MOSFET models provide most of the designers with essential information concerning electrical properties of the components associated with the manufacturing process of the chip. EKV model is the most popular compact model that has been used by many researchers to model the optoelectronic behavior of the device[17]. Many researchers have modified the basic structure of MOSFET to improve conductivity of the device which couls affect the cost of manufacturing. To increase the conductivity of the device, many parameters can be considered such as, device length, width, oxide thickness, doping density, etc [8]. Decrease in device length increases channel conductivity due to the shortening of channel length which leads short channel effects in the device. The other approach is broadening of the device width which leads to increase of the device size, hence not recommended. Oxide thickness reduction causes increase in insulator capacitance. Finally increasing doping profile is not at all recommendable or feasible in all cases. Table 1: Overview of Different Modeling Techniques[14-17]. Model Modeling Technique Generation Pros/Limitations SPICE Physical First suitable for long channel device BSIM Numerical equation Second Complex implementation, requires several parameters without clear physical meaning MOS Model 9 Physical Third Provides continuous and smooth behavior of device characterization EKV Interpolation function Third Valid in all operating regions 2
The traditional MOSFET device that is used extensively in communication systems is available in multifingered structure. It allows the increase of effective width and tends to increase channel conductivity. MOSFETs designed for radio frequency applications normally have large W/L ratios. It has a large transconductance therefore has large f t so that its noise figure is minimal. Increase in MOSFET width causes a decrease in f max. To counteract this effect, the MOSFET is laid out as a parallel connection of narrower MOSFETs, resulting in the fingered structure[18]. Further, it is required to the investigate sensitivity of these devices towards light. In recent decades researchers have shown keen interest in the development of an accurate and simple device model of the device at RF. Optical control to this device provides an additional port to control the device characteristics. 0.2 Requirements of a Photodetector Optical response of any photodetector depends on the optical window which is defined with the help of parameters such as frequency at which the detector operates, wavelength of incident radiations, intensity of incident radiations (in terms of flux density or optical power). Further, the optical absorption depends on the light penetration in the device and the absorption coefficient of the material. This penetration depth is more in silicon devices as compared to others. Absorption coefficient and wavelength of any material can be estimated from figure 1. Figure 1: The Absorption Coefficient of Common Detector Materials Vs Photon Energy (Bottom Axis) and Wavelength (top axis)[18 ]. Sensitivity of the MOSFET/MISFET to light has opened up the possibility of their use for a variety of optoelectronic applications. This sensitivity depends on the gate material and the semiconductor material of the device used for photodetection. Table 2.2 illustrates the types of gate, gate material used and nature of gate used for photodetectors. It is seen that the polysilicon gate is most suitable for photodetection as it is transparent in nature which can allow light to penetrate into the device. 3
Table 2: Gate Material Used for Photodetectors. Device Reference Type of gate Material Nature of Gate Si/SiGe Optoelectronic Switch [19] Metallic Gold Semitransparent SOI MOSFET [20] Nonmetallic Polysilicon Transparent MOSFET [21] Nonmetallic Polysilicon Transparent CMOS Device [22] Nonmetallic Polysilicon Transparent Figure 2 depicts the energy band diagram of MOSFET under illumination [6,12]. The photo carriers are generated through two different processes. One is an electron-hole pair excitation from the valance band to conduction band in the depletion layer of the space charge region. Another method is electron excitation from the interface states to the conduction band at the Si-SiO 2 interface. Both the processes are the function of surface potential[23,24]. Photo-carriers excited by light then drift to the drain in the lateral field while decreasing their density at a rate determined by an excess carrier lifetime. Figure 2: Energy Band Diagram of MOSFET under Illumination[13] 4
0.3 Optical Modeling and DC Characterization of Silicon Based Photodetectors Tools and techniques adopted by researchers for modeling optical effect, in the last few decades, are reviewed in this section. Katya[25] used MATLAB tool modeling electro-optical characteristics of photodiode. George K[20] carried out a theoretical investigation of light dependence of SOI MOSFET with non uniform doping which is purely simulation based. Metal Gate of the FET is replaced by the absorption region. The drain to source current and the transconductance increases significantly with optical flux density. 2D Poisson equation is solved for modeling surface potential and threshold voltage. Increase in photon flux density lower downs the surface potential barrier. There is reduction in threshold voltage with increase in intensity of illumination which increases drain current and transconductance of the device. The pinchoff voltage beyond which the current is saturated increases with photon flux density. The saturation current can be controlled by varying the flux density. H C Kim[26], has provided experimental set up to study photonic characterization of Current - Voltage characteristics in MOSFETS. Photovoltaic effect changes threshold levels and photoconductive effect increases the channel conductivity. Modeling is based on interface states in MOS System and uniform distribution of trap levels is assumed, practically which may differ. 1-D poisson equation is used for modeling. The device used for simulation was long channel and operating range is in KHz only. Yet there is need for new non destructive methods for efficient and accurate extraction of interface trap density. The author reported that that threshold voltage decreases with increase in optical power of incident radiations which increases the drain current. Abhinav Kranti[4], presented thin film fully depleted Surrounding/Cylindrical Gate (SGT) MOSFET model based on threshold voltage modeling. Tho riginal structure of the device is modified for modeling. Increase in the incident photon flux density, effectively lowers the effective bulk charges and the channel barrier thereby decreasing the threshold voltage and increasing the drain current. It was observed that this structure gives betters conductivity than that of conventional MOSFET. To reduce short channel effect in the SGT, it requires scaling down the device width which eventually decreases the drain current of the device. For confirmation of the validity of this approach, he compared the results in dark condition with the published data. M S Khim[21] presented an experimental model, in which illumination of the device causes reduction in the potential barrier due to PIBL. Threshold voltage reduces logarithmically with increase in photon flux density and saturation of the drain current can be controlled optically. M Schlosser[27] reported impact ionization MOSFET as an optical detector; he used the concept of cascaded APD isused for modelling. Increase in drain current is due to the impact ionization and photon induced charge carrier generation. There is scope for improvement in current structure to reduce unintentional doping level of i-zone (i-insulator). This model provides simple and accurate expressions for characterization of the device which are helpful to understand the physics of optical interaction in the device. S. D. Kirkish et. al.[28] tested CMOS fabricated MOSFET experimentally to observe the optical effect in which the P-well was floating on a n-type substrate. Light source used for experimentation was LED coupled to a fiber pigtail and collimating lens for wavelength of 820nm with a spot size of 0.4mm diameter. For large values of drain voltage, impact ionization occurs in the pinch-off region. Photodetection is observed in the well rather than in the channel. Holes are swept into the p-well and electrons are pulled into channel by the 2D electric field thereby increasing the drain current. With this approach the kink in drain current is overcome, but there is threat of performance degradation due to impact ionization. This device can be used as a digital 5
photodetector. Kazuhiko Shimomura and Tomonari Yamagata[29] have demonstrated a novel structure in which InGaAs/InP heterointerface acts as a light absorption area (p-i-n photodiode) which is directly integrated onto the gate of MOSFET. There are two electrodes in the absorption region, which are used to separate the photogenerated carriers and to pull them out from the external bias. He has used light source of long wave length laser light coupled with single mode optical fiber. In traditional FET devices, when illuminated, photo excited carriers are generated in the channel, the diffusion of these carriers limits the speed of the device. With his novel approach the drain current gets modulated not by the carriers but by the electric field which increases the speed and reduces noise. 0.4 AC Analysis of Silicon Based Photodetectors Extensive use of optical transreception demands more explanations about microwave interaction in the device as many authors have not reported frequency dependent characteristics of the silicon photodetectors. Hence photo effect on the device is influenced with different optical powers and the modulating frequency is required to be analyzed thoroughly to understand AC analysis. A small signal model for the intrinsic device is required with the inclusions of physical and optical effects, such as normal and reverse short-channel effects, channel-length modulation, photo-induced barrier lowering (PIBL), velocity saturation, mobility degeneration due to vertical electric field, impact ionization, band-to-band tunneling, small-signal parameters such as transconductance gm, and the intrinsic transcapacitance[30]. In real circuit operation, the device operates under time-varying terminal voltages. The dynamic operation is influenced by the device s capacitive effects. Thus a capacitive model describing the intrinsic components of the device capacitance is an essential part of a MOSFET model for circuit simulation. Many models have been suggested for capacitive modeling [27,31,32]. EKV MOSFET model is scalable and is built on fundamental physical properties [33,34]. [17] has used EKV model for high frequency application where as [35] developed mathematical model of the coordinate-sensitive photodetector which is composed of a system equations with partial derivatives describing the physical process of the photodetector structure. [21, 23, 36] found that surface potential decreases under illumination which results in reduction of depletion width, which is followed by a corresponding increase in capacitance. S. J. Song [36] presented a model based on interface states at Si/SiO 2 hetrojunction for pmos. He reported that MOS capacitance is less sensitive to light whereas there is significant effect of illumination on the drain current of MOSFET. At RF, transistors are usually characterized by its Y- and S-parameters. Very few researchers have provided insight on these parameters. A small signal model can be used to determine the Y parameters. MOSFET can be viewed as a two port network with two controlling ports, Gate and Drain with common grounded Source terminal. An S parameter is the proper tool to characterize the two-port network description of RF devices. It is required to analyze these parameters at RF to understand circuit behavior at RF. In [37], author has described these parameters for MESFET and reported that S parameters have significant effect of illumination. 0.5 Noise Analysis of Si Based Photodetectors Noise is an important issue at RF. Information transmitted by transmitter need not be the same at the receiver, in between noise is the factor that corrupts the original signal. Noise sources and 6
modeling of noise is well explained in [18]. The intrinsic and extrinsic noise sources in a MOS- FET operating at RF are predominantly thermal in origin. The extrinsic noise arises from the parasitic resistances found at the MOSFET terminal connections. Sources of intrinsic noise are as follows, Drain channel noise which is the noise generated by the carries in the channel region and appears as a noise current; Gate resistance noise is the thermal noise due to the resistance of the distributed gate material Induced gate noise is a gate noise current that is capacitively coupled onto the gate from the noise generated by the carriers in the channel and the channel resistance. Allam and I. M. Filanovsky [38], explore channel noise and characterization in deep submicron MOSFET at dark condition. In addition they reported a small signal model of MOSFET at RF frequency. Optical effect on noise and high frequency noise model of optically gated MOSFET is not reported by any of the authors. Researchers have referred to the possibility of applications of the devices they have modeled. This needs to be investigated how the devices will perform when they are used for real time applications. As per review, the Researchers have excluded this corner for further studies after completing initial modeling and simulation of device characteristics. Noise and S parameters can be used to prove modeling of the device application. To produce a new component, the researchers believe in on the 0.25µm process. Leading manufacturers of cell-phone and digital cameras fit CMOS chips in their products. It reduces demand on the battery as well as the size of the product[39]. 0.6 InGaAs MISFET The scaling of bulk Si MOSFET beyond the 45 nm technology node is extremely difficult due to a low Si mobility and quantum-mechanical tunneling through the sub nanometer thin gate oxide which results in an intolerably high leakage. Therefore, it is widely accepted that the bulk MOSFET architecture has to be changed and the industry is now adopting non-orthodox materials, new technologies and alternative device architectures. Naturally, the research into alternative transistor architectures recreated the possibility of building MOSFET based on III-V materials to profit from higher mobility in the channel. The development of suitable high-κ gate dielectrics for GaAs with an unpinned oxide/semiconductor interface is recognized as the most important step towards a real III-V transistor. Some of the past objections to III-V technologies are: III-V materials do not have a good native oxide; III-V devices have to be grown by epitaxial technology; III-V substrates are not in large diameters to offer competitive economies of scale; III-V transistors cannot be integrated within the current Si based technology. Today, industry is already adopting non-orthodox materials and technologies, such as: high-κ dielectrics for GaAs/AlGaAs with unpinned oxide/semiconductor interface by freescale semiconductor; 7
epitaxial SiGe in CMOS; alternative device architectures. 0.7 MISFET Photodetector Karol Kalna and Asen Asenov[40] have studied the potential performance of an n-type InGaAs MOSFET with an 80 nm physical gate length and a high-κ gate dielectric using ensemble Monte Carlo simulator MC/MOS. The Transport model is extended by using Fermi-Dirac statistics for the screening in ionized impurity scattering and Fermi exclusion principle (degeneracy) at each scattering process [41,42]. Further, they have incorporated a roughness scattering at the oxide/semiconductor interface and adopted an effective quantum potential to incorporate the effect of quantum confinement. The simulated performance of the InGaAs channel MOSFETs is then compared with the results for equivalent n-type conventional and strained Si channel MOSFETs which have been obtained from calibrated Monte Carlo simulations in order to fairly compare the intrinsic device performance. They found a drain current increase of about 50-60% and a maximum transconductance of 2080 ms/mm. Q.T. Do[43], fabricated the device by using compound III/V semiconductor material. The gate length and dielectric scaling behavior is experimentally studied by means of dc output- and transfer-characteristics and is modeled using the long-channel MOSFET equations. He found that using the long-channel MOSFET equations, it is possible to represent the InAs NWFET dc characteristics as a function of geometrical parameters in the investigated range. Excellent performance of output current and transconductance of the device is obtained. One way to improve the performance of an Si transistor is to change the material from Si into another semiconductor in which the charge carriers have higher mobility. One such material is the ternary III/V semiconductor In x Ga 1 x As(x = 0,.., 1). Electrons in InGaAs have a maximum mobility of 10000 cm 2 /V s while the electron mobility in Si is at max 1400 cm 2 /V s. The higher is the mobility of charge carriers the faster is the response of the device and higher is the current level[44]. Naofumi et. al [40] presented Uni-traveling-carrier photodiodes (UTC-PD s) with ultrafast response and high-saturation output are reported. The UTC-PD structures studied in this work were grown on semi-insulating InP substrates by MOCVD whereas In 0.53 Ga 0.47 As(x = 0.53) material is used in the absorption layer. It was observed, these are the fastest photoresponses ever reported for 1.55 µm wavelength ultrahigh-speed photodiodes. Y. M. Kim et. al, in [45], reported that a device using InGaAs or GaAsSb epitaxial base layers and InGaAs or InP epitaxial collector layers lattice-matched to InP currently exhibit significantly higher current-gain and power-gain cut off frequencies than GaAs-based device. The requirement to develop a simple structure of InGaAs MISFET based photodetectors still exists. 0.8 Conclusion To use the device as a photodetector, study is carried out for various requirements of detector like device dimensions, light source, gate material, operating frequency, absorption coefficient, semiconductor material etc. The study shows that there are different device models present for MOSFET under dark and illumination condition. To develop a MOSFET based photodetector which is cost effective, existing MOS devices can be used. Existing models incorporate light 8
effect modeling of drain current and transconductance only. Very little insight is provided on modeling of capacitances under illumination. As these detectors will to be used at very high frequencies, it is required to study frequency dependent characteristics of the device. None of the authors have focused on frequency dependence of these parameters. Further, HF noise is an important issue which is not under discussion as far as MOSFET based photodetectors are concerned. Y parameters and S parameters are needed to discuss as being important at RF. The study shows that the alloys of III-V elements are the materials of choice for detectors because they satisfy all the critical requirements of photodetectors. As the device is to be used as a photodetector, responsivity and quantum efficiency are main characteristics of a photodetector whose modeling technique is not discussed by researchers. Further, modeling of the device for specific applications is required to be analyzed to verify its utility. These shortcoming of the existing models are the source of motivation to develop a new model. For modeling of a MOSFET photodetector, multifingered structure is most suitable for optimization of the output current without modifying the existing device structure. In addition, Silicon and InGaAs are promising materials that can be used for optical detection. The EKV model is best suited for modelling of drain current and capacitances in different operating region, including all short channel effects. Further, the MATLAB tool can be used to test and tune the simulation results of device characteristics. The problem and scope of work have been defined in section 1.6. Since fabrication facilities are not there and no support from foundries, only simulation is possible and the results can be validated by comparison with the reported results. 9