High-Frequency Solid-State Electronic Devices - IEEE Xplore

1 downloads 0 Views 633KB Size Report
High-Frequency Solid-State Electronic Devices. Robert J. Trew, Fellow, IEEE. Abstract—Starting with exploratory work in the 1930s and development work in the ...



High-Frequency Solid-State Electronic Devices Robert J. Trew, Fellow, IEEE

Abstract—Starting with exploratory work in the 1930s and development work in the 1940s a variety of two-terminal and three-terminal solid-state device structures have been proposed, fabricated, and developed. This work parallels the development effort on vacuum electronic devices, and the two technologies share many applications. The solid-state and vacuum electronic devices work in tandem to enable numerous commercial and military systems. Solid-state device development is closely linked to semiconductor materials growth and processing technology, and advances such as the introduction of heterojunction growth technolog, permit complex multiple layer device structures to be fabricated and optimized for maximized device performance. This work has been very successful and a variety of high-performance diodes and transistors are now available for use from UHF into the millimeter-wave spectrum, approaching terahertz frequencies. The development, operating principles, and state-of-the-art of various diode and transistor structures are reviewed. Index Terms—Microwave solid-state amplifiers, microwave solid-state devices.



HE DEVELOPMENT of solid-state and vacuum electronic devices share much history and parallel development. Electron emission from materials was known in the 1800s, leading to the “Edison Effect” reported in 1883, which followed the invention of the electric light bulb. Similarly, work on semiconductor devices was reported by Braun in 1874 [1]. The first practical vacuum diode (called the “Fleming Valve”) was patented by J. A. Fleming in 1905 [2], and the first vacuum triode was patented by L. DeForest in 1907 [3]. The basic structure for the vacuum triode was simple, consisting of a cathode and anode separated by a distance and enclosed in a vacuum envelope. The triode is formed by placing a metal grid structure between the anode and cathode. In operation, a heater near the cathode causes electrons to be thermionically emitted from the cathode forming an electron cloud in the vacuum region around the cathode. The anode is positively biased, which results in an electron current flowing between the cathode and anode. The grid is negatively biased, which permits control and modulation of the cathode-to-anode current. The performance of the triode is a function of the device geometry, and particularly, the distance between the cathode and anode. In the early days the inability to closely space the cathode and anode limited the frequency response of the device. Manuscript received August 30, 2004; revised October 28, 2004. This work was supported in part by the U.S. Office of Naval Research under Grant N00014-03-1-0803 and in part by the U.S. Army Research Office under Grant DAAD19-03-1-0148. This paper is based upon an invited semiconductor device overview presentation at the 2004 IVEC conference held in Monterey, CA, April 27–29. The review of this paper was arranged by Editor W. L. Menninger. The author is with the Electronic and Computer Engineering Department, North Carolina State University, Raleigh, NC 27695-7243 USA. Digital Object Identifier 10.1109/TED.2005.845862

The first solid-state transistor was an attempt to reproduce the vacuum triode structure and implement it using the semiconductor to replace the vacuum. The solid-state device had essentially the same structure as the vacuum triode, but was implemented in a planar geometry. The first solid-state triodes were field-effect transistors (FETs) consisting of a layer of doped semiconductor supporting deposited metal electrodes. The role of the cathode is replaced with a source electrode and the anode is replaced with the drain electrode. When the source is grounded and a positive voltage applied to the drain, a current flows between the two electrodes. The role of the grid is performed by a gate electrode located between the source and drain. The semiconductor serves as a medium for current flow and as a mechanical structure to support the electrodes. Dissipated thermal energy is generally extracted through the semiconductor. An advantage of the solid-state transistor is that it can be operated at low bias voltage and low temperature, thereby significantly improving device reliability. However, the need to dissipate thermal energy through the semiconductor and supporting substrate can present difficult design challenges, particularly for devices designed for high power operation. Following the introduction of the early vacuum and semiconductor devices a wide variety of device types have been developed and introduced. These devices have been the enabling technology for numerous electronic systems. The process continues and new devices, both vacuum and solid-state, are continually being introduced and developed. In this paper, the various types of solid-state devices are reviewed. II. MATERIALS PARAMETERS The dc and RF performance capability of electronic devices is fundamentally dependent upon the electronic, thermal, and mechanical properties of the materials from which the devices are fabricated. Of particular importance are the charge transport characteristics as a function of electric field for the material. Each semiconductor has a different velocity-field characteristic, and semiconductors of interest will have high carrier velocity capability. From this perspective, a vacuum is simply another material, and serves as the limiting material to establish RF power performance. Once electrons are emitted from a metallic cathode they will establish an electron beam current as they flow through the vacuum toward a positively biased anode. The electrons encounter no scattering events in the vacuum and their velocity is the highest that can be obtained from any material, so from this perspective the vacuum is a perfect material. Semiconductor materials quality has continually improved and semiconductor devices have advanced in performance as a result. A variety of technologies for growth of semiconductor epitaxial layers, such as molecular beam epitaxy (MBE) and organomellatic chemical vapor deposition, have been developed

0018-9383/$20.00 © 2005 IEEE




and these technologies permit the growth of epitaxial layers of precise thickness and impurity doping concentration. It is now possible to fabricate solid-state devices with layer thickness of only a few angstroms, and this level of control permits devices with frequency performance well over 100–300 GHz to be fabricated. A summary of some of the semiconductor material properties most important to electronic device performance is listed in Table I for several semiconductors. Desirable material properties include a large energy gap, (eV), a low value of dielectric -cm , and constant, , high thermal conductivity, high critical electric field for breakdown (V/cm). Wide bandgap energy generally translates into an ability to support high internal electric fields before electronic breakdown occurs, and also provides for improved radiation resistance. For vacuum devices electric field breakdown is established by electrode spacing and does not occur in the vacuum in a similar manner as for the semiconductor materials. Most semiconductor device fabrication has been in Si, GaAs, and InP and related compounds and virtually all devices commercially available are fabricated from these materials. Recently there has been interest in the development of devices from wide bandgap materials such as SiC and GaN. These materials have energy bandgaps about two to three times those in the conventional semiconductors and show significant potential for fabrication of high power, high-frequency devices. The dielectric constant is an indication of the capacitive loading of a device and affects the terminal impedance. Generally, for solid-state devices a low value for the semiconductor dielectric constant is desired, and this permits a solid-state device to be larger in area for a specified impedance. Increased area permits larger RF currents and higher RF power to be generated. Since a vacuum has the ) transit-time lowest possible dielectric constant (i.e., dominated triode vacuum devices that operate in an analogous manner to solid-state devices have the lowest capacitive loading and can be made with larger geometries than semiconductor devices. However, most vacuum electronic devices make use of distributed structures and the dielectric constant does not affect device performance in the same manner as for solid-state devices. Vacuum electronic devices make use of bunched electron and distributed circuit concepts that permit electron beam/electromagnetic wave interactions that enable very high RF power and very high-frequency performance to be obtained. The thermal conductance of the material is extremely important since this parameter indicates the ease with which dissipated power can be extracted from the device. Poor

Fig. 1. Electron velocity versus electric field characteristics for several = 10 cm ). semiconductors (


thermal conductivity results in device operation at elevated temperature with degraded performance. Compound semiconductors such as GaAs and InP are, in general, poor thermal conductors and this introduces complexity in device design for devices designed to operate at high power. Diamond and SiC are excellent thermal conductors and are often used for heat sink applications. Vacuum devices are not limited by material thermal conductivity in the same manner as solid-state devices since most thermal dissipation in the vacuum devices occurs in the metal electrodes, where effective heat sinking techniques can be employed. Finally, the critical electric field for electronic breakdown should be high. This parameter is an indication of the strength of the electric fields that can be supported internally in the device before breakdown. High electric fields permit large terminal RF voltages to be supported, and this is necessary for the generation of high RF power. One of the attractive features of the wide bandgap materials is a high value for the critical field for breakdown, which is typically an order of magnitude greater than for conventional semiconductors. Basically, a current is defined as the movement of charge and expressed as the product between the charge density and transport velocity. Therefore, the dc and RF currents that flow through a device are directly dependent upon the charge carrier velocity versus electric field transport characteristics of the semiconductor material. Generally, for high currents and high frequency, high charge carrier mobility and high saturation velocity are desirable. A comparison of the electron velocity-eleccharacteristics for several semiconductors is tric field – shown in Fig. 1. The – characteristic is described in terms of charge carrier mobility , (units of cm V s) defined from the slope of the – characteristic at low electric field, and the saturated velocity (units of cm/sec), defined when the carrier velocity obtains a constant, field-independent magnitude, generally at high electric field. The high value for electron mobility cm V s) is the main reason of GaAs (typically, that FETs fabricated from this material have such excellent low noise and high-frequency performance. The – characteristics shown in Fig. 1 are for transport through semiconductors doped cm , which is a typical impurity concentration at used in device fabrication.



Fig. 2. Basic amplifier configuration.

III. AMPLIFIER FUNDAMENTALS The basic configuration for an amplifier is shown in Fig. 2. The amplifier is a two-port network that consists of a source that feeds the input with a load connected to the output. The network has gain and thereby amplifies a signal passing through it from the source to the load. RF power can only be generated from a real source (i.e., resistance) and delivered through a network to a real load (i.e., resistance). Since electronic devices and networks, as well as most microwave sources and loads, also include reactance it is necessary to employ reactive tuning to obtain optimum power transfer. Conjugately tuned output and load impedances deliver maximum RF output power from the source to the load. from the network can be The power delivered to the load written as (1) and are the voltage and current at the load where is the real part of the load resistance. The impedance, and power delivered to the load can be written as a function of the reflection coefficient at the load (2) where is the RF power available from the network and is the reflection coefficient at the load. Maximum RF power transfer occurs for no reflection from the load (3) This condition occurs when the load impedance is set to the conjugate of the network output impedance (4) The amplifier power-added efficiency is (5) where is the RF power into the network, is the dc power dissipated in the network, and is the network gain, expressed as (6) The dynamic characteristics of the amplifier are illustrated in Fig. 3, which shows dynamic load lines (i.e., current–voltage (I–V) characteristics) for three conditions: linear operation; db in compression; and db in compression. The dynamic load lines are superimposed upon the dc – characteristics for the active device. For the situation shown in Fig. 3 the active

Fig. 3.

Dynamic and dc load lines for a transistor amplifier network.

device, in this case a transistor, is biased with a drain/source V, and the network is tuned for maximum voltage of power-added efficiency for each dynamic load line. Since RF power can only be generated by a real source and delivered to a real load, the dynamic load line would be a straight line oscillating up and down the dc load line for the network. However, since the device has capacitance, the dynamic load line shows elliptical behavior. While the device is operating below saturation the load line is confined within the dc – characteristics. As the device is driven into saturation the dynamic load line shifts and extends outside the dc – characteristics on both the high current and low current portions of the RF cycle. The average value of the RF current also increases, indicating that the device dc current increases as the device is driven into saturation. The extension of the dynamic load line outside the dc – characteristics is possible due to the complex nature of the network. The total RF current consists of conduction and displacement components and although the conduction current is limited by the – characteristics, the displacement current maintains current continuity at the terminals. That is, as the device is driven into saturation the conduction current is clipped by the – characteristics for the network, but the total RF current continuity is maintained by displacement current. Network capacitance increases as the device is driven into saturation and inductive tuning is necessary to obtain optimum RF performance. Optimized inductive tuning results in the reversal of dynamic load line direction, as shown in Fig. 3. Under optimum tuning conditions the network is essentially a resonant circuit with the reactive energy shifting between the capacitive and inductive fields. As the network is driven further into saturation the current clipping behavior increases, with a net increase in both dc current and device capacitance. The dynamic behavior of the amplifier network defines the factors that determine the RF performance limits of the device and the materials from which it is fabricated. The power delivered to the load is a product of the RF voltage and RF current that can be established at the load, and this is determined by the active device. Semiconductors are limited in the bias voltage that can be applied by the critical electric field for breakdown of the semiconductor material. Therefore, semiconductors that have high critical electric fields for breakdown are desirable for power device applications. The critical field for breakdown is a



Fig. 4. Average RF output power versus frequency for various electronic devices (courtesy of the Naval Research Laboratory).

function of bandgap energy and wide bandgap semiconductors are desirable for power applications. Semiconductors, such as SiC and GaN, show significant potential for these applications. Electronic devices designed for microwave and RF applications operate in a transit-time mode and are scaled in size by frequency considerations. Under normal operation the electric fields within the devices vary from low magnitude near the electron injection location to magnitude sufficient to produce electron velocity saturation in the charge control/modulation region. Therefore, large current capability requires semiconductor materials that have high electron velocity. In general, both high mobility and high saturation velocity are desirable for high RF current. Traditional semiconductors such as Si and GaAs have electron saturation velocities that are limited to about cm/s, and this limits both the power that can be generated and the frequency response of the device. Wide bandgap semiconductors have electron saturation velocities that are a factor of two higher. The combination of high current and high voltage capability make wide bandgap semiconductors very attractive candidate materials for fabrication of high-power and high-performance electronic devices [4].

IV. SOLID-STATE ELECTRONIC DEVICES A large variety of semiconductor devices for high-frequency applications have been proposed, fabricated, demonstrated, and used in practical applications. A review of these devices has been presented [5]. The present state-of-the-art RF power performance of microwave solid-state devices is compared to that for microwave tubes in Fig. 4. Solid-state devices produce RF power of about 100 W in S-band and 1 W at 100 GHz. As indicated, the RF performance of solid-state devices is significantly lower than obtained from electronic vacuum

tubes. The reduced RF power capability of solid-state devices is due to 1) lower bias voltage that can be applied; 2) reduced electron velocity in the semiconductor which produces reduced current and; 3) a thermal limitation caused by the semiconductor thermal impedance. The relatively low bias voltage at which solid-state devices operate permit high reliability due to reduced electric field stress, and the ability to use lithography technology permits low fabrication costs. Solid-state device RF output power can be increased by using power combining technology, although system RF output power is generally limited to the 10’s to 100’s of kilowatt level in the microwave region. For megawatt systems it is difficult to efficiently combine the large number of devices necessary for practical systems. 1) Transistors: Three-terminal transistor structures can be fabricated with a variety of geometries and operating characteristics. Fundamentally, they all function by controlling the conductivity of a conducting channel by establishment and modulation of an electrostatic barrier. However, the various transistor structures differ regarding the details of how the electrostatic barrier is formed, and how it modulates the channel conductivity. Field-effect transistors are majority carrier devices, where the modulation region controls majority carrier current, and bipolar transistors are minority carrier devices where the modulation region controls minority carrier current. The operation and performance of these devices are reviewed. The first practical work on the development of FETs was reported in patents by Lilienfeld in 1930 [6] and 1933 [7], and reports by Stuetzer [8], [9] in 1950 and Shockley in 1952 [10]. The early devices demonstrated limited performance due to relatively poor semiconductor material quality and an inability to fabricate a gate electrode with fine line geometry. The realization of high performance devices needed to wait for the advancement of epitaxial semiconductor growth technology and the development of optical lithography to produce gate lengths


with micron dimensions. Development effort on FETs was delayed following the successful demonstration of the bipolar transistor by Bardeen, Brattain, and Shockley in 1948 [11], [12]. In the years following demonstration of the bipolar transistor most development work focused upon advancement of the pn junction, and a variety of transistors and other devices with performance adequate for practical application at microwave frequencies were reported. Initially, Ge was the semiconductor of choice. However, Ge suffers from relatively low bandgap energy and low thermal conductivity, and this produced devices with high leakage currents and poor thermal performance. There was a search for semiconductors with improved properties and Si soon replaced Ge as the semiconductor material of choice. By the early 1970s, material quality and fine line optical lithography were advancing rapidly [13] and high-performance FETs became practical [14]–[16]. The development of III–V compound semiconductors such as GaAs, InP, and related ternary compounds permitted microwave and millimeter-wave devices with excellent noise and power performance to be developed [17]. Today, RF performance of FETs extends well into the millimeter-wave region, and frequency response greater than 300 GHz has been reported for InP-based compound semiconductor high-electron mobility transistor (HEMT) devices. The bipolar transistor was invented by Shockly, Bardeen, and Brittain in 1948 [11], [12]. Since that time the device has undergone continued development and improvement and is now in wide use for microwave and millimeter-wave applications. Most bipolar transistors are fabricated from Si and can be produced at low cost. These devices are useful for both low-frequency RF applications, as well as microwave applications up to X-band. Although the advantages of utilizing a wide bandgap semiconductor for the emitter of a bipolar transistor were discussed by Shockley in his transistor patent, the modern heterojunction bipolar transistor (HBT) was proposed in 1957 by Kroemer [18], who also discussed the advantages of HBTs over conventional bipolar transistors. The HBT has been extensively developed and has found wide useage in applications, such as RF amplifiers in cell phones. The most common HBT structure makes use of the AlGaAs–GaAs semiconductor system. Submicrometer scaling has now pushed the s to the range of 300 GHz [19], [20] using InP-based semiconductors. The development of HBTs using SiGe as the base were reported in 1987 [21], and these devices advanced rapidly and now produce RF performance essentially equivalent to AlGaAs–GaAs HBTs. The basic structure for a bipolar transistor is illustrated in Fig. 5. The transistor is a pn junction device formed from back-to-back junctions. The transistor current is controlled by minority carrier diffusion across the base region, and since electrons have higher velocity in the semiconductor than holes, the npn structure is preferred for high-frequency applications. Modulation of the current through the device is controlled by varying the amount of minority charge injected into the base region from the emitter by application of a forward bias to the base–emitter pn junction. The minority charge diffuses through the base region and is extracted from the device by the reverse biased base–collection pn junction. The frequency performance of the transistor is dependent upon the time it takes for an electron to transit through the entire device.


Fig. 5. Basic structure for a bipolar transistor.

Fig. 6. Energy band diagram for a bipolar transistor.

To improve the frequency response and RF performance of the bipolar transistor a wide bandgap semiconductor can be used as the emitter, as illustrated in Fig. 6. The wide bandgap emitter blocks the back injection of charge from the base region into the emitter region, thereby increasing the emitter injection efficiency. The back injection of charge into the emitter region is blocked by the discontinuity in the energy bandgap, and in order for this mechanism to be effective it is necessary that semiconductors that have the energy discontinuity in the valence band be used. The range of semiconductor heterojunctions that satisfy this requirement is limited, and the most common are the AlGaAs–GaAs and the InGaAs–InP materials systems. A large discontinuity in the valence bandgap energy is desirable since it produces much improved injection of minority charge into the base region and thereby increases device gain. Heterojunction bipolar transistors fabricated from the SiGe–Si materials system are also being developed. The SiGe alloy has a lower bandgap energy than Si, so for these devices the SiGe is used



Fig. 7. Basic structure for a FET.

as the base material. The resulting HBT transistors have the advantage of a wide bandgap emitter, and also the SiGe has high mobility so that the base region has much lower resistance than standard Si transistors, and this helps improve RF performance. The SiGe–Si HBT also has the advantage of being compatible with standard Si processing technology, which makes the device attractive from a cost perspective. The heterojunction bandgap can be chosen so that the HBT will have current gain independent of the base and emitter doping, and this permits the device to be optimally designed to maximize RF frequency response. Compared to a standard bipolar transistor the HBT has reduced base resistance, output conductance, and emitter depletion capacitance, and greatly improved high-frequency performance. The advantages of using a wider bandgap semiconductor for the emitter than for the rest of the device were noted by Shockly in his original bipolar transistor patent. A comparison of gain-bandwidth products for similar GaAs and SiGe HBTs has been reported by Ning [22]. A GaAs HBT 4.6 m emitter produced an of 140 GHz with a 0.6 whereas a SiGe HBT with a 0.35 3.55 m emitter produced of 130 GHz. The peak occurs at slightly lower colan lector current for the GaAs HBT compared to the SiGe HBT. These results indicated that the two HBTs have comparable high-frequency performance. The SiGe–Si HBTs also have excellent low noise performance due to high mobility in the SiGe material, which helps produce a low base resistance. The FET is a majority carrier device with the basic structure shown in Fig. 7. Since electrons have higher velocity than holes, the majority of FETs are fabricated using n-type semiconductor material. The source and drain electrode contacts are designed to have ohmic characteristics (facilitated by the N regions) so that when connected to an external bias source the device operates as a simple resistor. A third electrode (the gate) is located between the two electrodes and is designed to be a rectifying contact. By applying a reverse bias the channel region under the gate can be depleted of charge, thereby providing a gating function. Since application of a small RF signal to the gate permits control of the channel current, which is relatively large, a gain mechanism is established. The rectifying gate contact can either be a metal-semiconductor Schottky contact to form a metal semiconductor FET (MESFET), or a pn junction to form a junction FET (JFET). In general, due to reduced gate capacitance and higher transconductance, the RF performance of the MESFET has proven much superior to that of the JFET and development work with JFETs has been limited. Fabrication of high-performance FETs is dependent upon the ability to fabricate gates with very short dimensions. This technology has continually improved and optical and electron-beam lithogm raphy can now routinely produce gate lengths down to

Fig. 8. RF performance of commercially available power FETs.

Fig. 9. HEMT structure.

and less. These devices can operate to well over 100 GHz with good RF performance. The GaAs MESFET is also capable of excellent low noise performance and noise figures on the order of 1–2 db in X-band and 3–4 db in Ka-band can now be obtained. The GaAs MESFET can be designed for microwave power applications by using multiple gate fingers arranged in parallel. This permits large gate widths to be fabricated while maintaining the short gate necessary for microwave performance. Since channel current is directly proportional to gate width the multiple gate finger structure results in large RF power due to the increased RF currents that flow. By scaling gate length to the m and by increasing channel doping to the range of cm , state-of-the-art power GaAs MESrange of FETs have s greater than 100 GHz, and can operate with good RF power, gain, and efficiency at least through Ka-band (26.5 to 40 GHz). FETs can be optimized for maximum RF output power in X-band (i.e., 7.9 GHz–12.4 GHz) by using gate-lengths on m and by reducing channel doping to the order of increase gate-drain breakdown voltage. Power designs also require device and package designs that produce reduced thermal resistance. Power FETs that produce on the order of 80 W at S-band and almost 1 W at 40 GHz with good power-added efficiency and gain, as shown in Fig. 8, are now commercially available. FETs based upon heterojunctions can also be fabricated. These devices have a structure as shown in Fig. 9, and consist of a highly doped wide bandgap semiconductor grown on a



Fig. 11.

RF current and power gain versus frequency for 4H-SiC MESFETs.

the semiconductor. Similarly, the highest frequency at which the transistor has positive gain is defined as the maximum frequency and written in the form of oscillation and stated as the (8) Fig. 10. Formation of a 2DEG at a hetero-interface between wide bandgap and narrow bandgap semiconductors.

weakly doped, or undoped, narrower bandgap semiconductor. The gate and source/drain electrodes are placed upon the wide bandgap material. If the two semiconductor materials are selected so that the discontinuity in the energy bands is restricted to the conduction band, a quantum well is created in the conduction bands at the interface between the two semiconductors. As electrons from the wide bandgap semiconductor diffuse into the quantum well a two-dimensional electron gas (2DEG) is created, as shown in Fig. 10. This concept was demonstrated by Stormer in 1979 [23]. The 2DEG can be used to form the conducting channel region for a FET. The resulting transistor is called a high-electron mobility transistor (HEMT) and was demonstrated by Mimura in 1980 [24]. HEMTs have extremely high-frequency performance capability and very low noise performance, primarily due to the very high mobility characteristics of the 2DEG. RF operation up to 300 GHz has been demonstrated. These devices are also used for microwave and millimeter-wave power applications and above X-band are superior to MESFETs. The frequency response of a FET can be defined by the gain-bandwidth product and described by the device (GHz) as (7) where (cm/s) is the saturation velocity of the electrons in the (mS) is the device transconductance, conducting channel, (pF) is the gate-source capacitance, and m is the gate length. High current gain and high-frequency response require short gate lengths and high electron saturation velocity. The former is controlled by the lithography and process technology employed to fabricate the gate and the latter is a function of

where is the drain/source resistance, and is the requires a high and a large gate resistance. A high ratio. The represents current gain and the ratio represents voltage gain. Power gain can be obtained , but only by establishing suitable at frequencies above voltage gain and this requires large output impedance to input impedance ratios. This is difficult to achieve at microwave frequencies and high performance millimeter-wave devices require high . Wide bandgap semiconductors show great promise for advancing the state-of-the-art for high power microwave electronic devices. Recent improvements in the growth of wide bandgap semiconductor materials, such as SiC and the GaN-based alloys, provide the opportunity to now design and fabricate microwave transistors that demonstrate performance previously available only from microwave tubes. The most promising electronic devices for fabrication in wide bandgap semiconductors for these applications are MESFETs fabricated from 4H-SiC and heterojunction FETs (called HFETs) fabricated using the AlGaN–GaN heterojunction. The theoretically predicted [4] frequency response of a 4H-SiC MESFET is shown in Fig. 11 and the corresponding RF performance for an X-band class A amplifier is shown in Fig. 12. The theoretical predictions indicate that 4H-SiC MESFETs can produce RF output power on the order of 4–6 W/mm and should produce useful RF power through X-Band. The theoretical predictions are in excellent agreement with measured data, and the predicted RF output power is experimentally obtained [25], [26]. The structure for an AlGaN–GaN HFET is shown in Fig. 13. The heterojunction is characterized with a 2DEG that has very cm . high sheet charge density, and on the order of This is very high and a factor of five higher than obtained with the AlGaAs–GaAs heterojunction. The high sheet charge density helps produce high current capability for the HFET. The


Fig. 12. RF performance of a 10-GHz 4H-SiC MESFET amplifier (V 40 V, class A).



Fig. 14. RF performance for a 1-mm gate width AlGaN–GaN HFET class A amplifier.

Fig. 15. Predicted RF performance for an AlGaN–GaN HFET class amplifier at 100 GHz (V = 30 V, class A).

Fig. 13.

Structure for an AlGaN–GaN HFET.

wide bandgap semiconductors also permit high bias voltage to be applied and with a drain bias of V these HFETs are predicted to produce RF output power on the order of 10–12 W/mm of gate periphery [4]. RF output power greater than 10 W/mm has been obtained experimentally [27]. Recently, field-plates consisting of an extension of the gate metal to the region between the gate and drain [28], have been used to suppress gate breakdown, and these devices permit high bias voltages to be applied. Field-plate HFETs have been biased as V, and these devices produce RF output high as power greater than 30 W/mm [28]. Nitride-based HFETs should be useful through Ka-band, as shown in the predicted RF performance in Fig. 14, and should operate potentially well into the millimeter-wave region. Simulations indicate that these devices may provide good RF output power as high as 100 GHz, as shown in Fig. 15. The Class A V and proamplifier shown in Fig. 15 is biased with duces RF output power of 400 mW with 17% power-added efficiency and 5 db linear gain. Total amplifier RF output could be increased by power combining technology. Through X-band the RF power capability of the nitride-based HFETs compares

very favorably with the 1–1.5 W/mm RF power available from GaAs MESFETs and GaAs- and InP-based HEMTs. 2) Active Diodes: Two-terminal active devices (diodes) can also be used to generate and amplify RF energy. In particular, the IMPATT [29], [30] and Gunn [31]–[33] devices have found practical application in a variety of microwave and millimeterwave systems. The IMPATT diode has the highest RF power/frequency performance characteristic of any solid-state device and for this reason, was extensively developed for use in transmitter applications. The IMPATT diode generates a negative resistance as a result of a greater than 90 phase delay between the RF current and voltage. The delayed RF current results from a combination of impact ionization and transit time effects, as first described by Shockley [29] and further explored by Read [30]. Read proposed a complex, multilayered diode structure that separated the avalanche ionization and transit-time regions. These devices, when biased into avalanche breakdown and placed in a resonant circuit, produce oscillations. Proper design of the diode structure can produce efficient oscillations well into the millimeter-wave spectrum. In fact, IMPATT diode oscillators have the highest power-frequency characteristic of any solidstate source. The Read structure proved difficult to fabricate and the first oscillations were obtained from simple pn junction diodes


[34]–[36], which were much easier to fabricate. However, the RF performance obtained from pn junction diodes has reduced RF output power and efficiency due to an inability to confine the avalanche region and interest remained in fabrication of the Read structures due to higher predicted performance. The Read structure and related variants with high-low and low-high-low doping profiles were realized in the late 1970s and early 1980s with the development of MBE semiconductor growth technology, which permitted submicron thick regions of precise impurity doping to be realized. These IMPATT diodes permitted engineered avalanche and drift regions and the resulting devices were capable of tens of watts of RF output power with good efficiency throughout the microwave region and into the millimeter-wave spectrum. The Gunn diode is another two-terminal device with active characteristics that found widespread application in a variety of microwave systems. Gunn observed the active characteristics of GaAs bulk crystals in 1963 [31] while performing measurements on GaAs crystals. The phenomena was shown by Ridley and Watkins, and independently by Hilsum, to be due to a transferred electron effect [32], [33] that occurs in the complex conduction band structure of GaAs and other similar direct bandgap III–V semiconductors, such as InP, that have conduction bands with multiple conduction band minima separated by low energy. As the electric field is increased electrons transfer from the central valley to the nearby satellite valleys, which generally have less curvature in energy-momentum space, and therefore higher electron effective mass. The net result is that the electron velocity decreases with increasing electric field as an increasing number of conduction band electrons are transferred from the light effective mass central valley to the heavier effective mass satellite valleys. A region of decreasing current with increasing voltage is established, and this establishes the potential for extracting energy from the bias electric field. The effect can be modeled as a negative conductance. When the device is biased in the active region an unstable condition exists, and a perturbation in the dynamic I–V characteristic will grow with time, establishing a necessary condition for oscillation. Placing the device in a resonant circuit will provide a feedback mechanism that completes the necessary and sufficient conditions for oscillation. Energy in the growing oscillation is extracted from the bias field applied to the device. The effect can be modeled by an equivalent circuit with a negative real part. These devices were generally termed Gunn or transferred electron devices and they found wide use in microwave and millimeter-wave systems, particularly for local oscillator applications due to a combination of wide tuning bandwidth and moderate noise performance. Although the feedback mechanism that generates the active properties of the devices vary, depending upon the internal physics, all the devices exhibit a negative differential resistance or negative differential conductance characteristic. Such devices can be placed in a resonant cavity to construct an oscillator, or they can also be used as amplifiers if used in a reflection-type circuit. Since the devices have only one port, an amplifier requires a reflection-type circuit that is realized by using either a circulator to separate the input and output signals or by injection locking the device operating as an oscillator.


Fig. 16.

N-type negative conductance and equivalent circuit.

Fig. 17.

S-type negative resistance and equivalent circuit.

Whether a negative resistance or conductance is generated depends upon the internal feedback physics of the particular structure. Two basic type classes are possible, and these are classified as N-type or S-type, after the shape of the dynamic characteristic in the RF I–V plane. An N-type active element is illustrated in Fig. 16. The active characteristic is obtained when the device is biased in the region of the I–V plane were the RF current decreases with RF voltage. The resulting characteristic is multivalued in current and an expansion of the RF current about the operating point yields the series indicated in Fig. 16. The first order term yields a conductance with a negative magnitude. Using perturbation theory it can be shown that the dynamic characteristic varies in a counter clockwise direction, indicating that the device reactance is capacitive. The equivalent circuit is, therefore, a negative conductance in parallel with a capacitance. Devices that can be modeled as N-type elements include Gunn diodes and diodes based upon electron tunneling phenomena. An S-type active element is illustrated in Fig. 17. For this element the active characteristic is again obtained when the device is biased in the region of the I–V plane were the RF current decreases with RF voltage. In this case the resulting characteristic is multivalued in voltage and an expansion of the RF current about the operating point yields the series indicated in Fig. 17. The first-order term yields a resistance with a negative value. Using perturbation theory it can be shown that the dynamic characteristic varies in a clockwise direction, indicating that the device reactance is inductive. The equivalent circuit is,


Fig. 18. RF output power versus frequency for GaAs and InP gunn devices. Numbers next to the symbols denote dc-to-RF conversion efficiencies in percent [5].

therefore, a negative resistance in series with an inductance, as shown in Fig. 17. Devices that can be modeled as S-type elements include IMPATT and other transit-time based devices. The RF power delivered to the load for an N-type active element can be expressed in terms of equivalent circuit parameters as [5]

(9) . Since the active device reactance where and the RF power is a capacitance, it follows that generated by the device will decrease with the square of the db/octave. frequency, or at Active diodes can be operated under pulsed conditions as well as continuous wave. Therefore, there are two limitations for power generation capability. The first is electronic and the second is thermal. The electronic generation capability is limited by matching to the load resistance including the series resistance of the device. The thermal limitation is determined by the thermal resistance of the device, which depends on various parameters, including area, heat sink and layer structure. The RF performance as a function of frequency for Gunn diode oscillators is shown in Fig. 18. These devices are generally fabricated from GaAs or InP, and both types of Gunn oscillators are indicated in Fig. 18. Gunn oscillations have been obtained up to 300 GHz, although RF power is low, and on the order of 1 mW. This power level is adequate for use in local oscillator applications, but is generally too low for use in transmitters. Conversion efficiency tends to be on the order of a few percent.


Fig. 19. RF Power performance of resonant-tunneling diode oscillators. The numbers associated with the point refer to the dc-to-RF conversion efficiency in percent [5].

Tunnel diodes and resonant tunneling diodes are also N-type active elements that can be used to fabricate oscillators and amplifiers. The tunnel mechanism occurs due to quantum mechanical tunneling of electrons through thin energy barriers. In order to create an active element the tunnel mechanism is used to efficiently inject electrons into a semiconductor depletion region. The tunnel injection is in-phase with the modulating voltage, so by itself, the tunnel phenomenon is not sufficient to cause a greater than 90 phase shift between the RF current and voltage. However, by injecting the electrons into a depletion region and making use of transit-time phenomena, it is possible to make an active element. The tunnel mechanism has subpico-second injection response time and the device is capable of very high-frequency performance, limited primarily by the length of the depletion region, and resulting transit-time. However, since the tunnel injection occurs at low bias voltage tunnel-based diodes are not capable of supporting high dc or RF voltages and low RF power results. The RF performance of resonant tunneling diodes is shown in Fig. 19. Tunnel diode oscillators have operated as high as 712 GHz, which is the highest fundamental frequency achieved to date from a solid-state source. However, RF output power is low, and on the order of 0.3 W. The IMPATT diode has been widely used in applications where good RF output power and good efficiency are required. In general, IMPATT diodes are capable of much greater RF power and efficiency than Gunn diodes. Both pn junction diodes and complex, multilayer Read structures are used. The most common semiconductor used is Si, although GaAs is extensively used for high performance diodes. In general, GaAs IMPATT diodes can produce greater RF output power



up to about 200 GHz. Useful power can still be obtained up to about 400 GHz. V. SUMMARY AND CONCLUSION

Fig. 20. RF output power versus frequency for IMPATT and other transit-time diode oscillators. Numbers next to the symbols denote dc-to-RF conversion efficiencies in percent [5].

and efficiency than similar diodes fabricated from Si. A variety of other device structures similar to the IMPATT diode have also been developed. These include the TUNNETT, MITATT, TRAPATT, and BARITT diodes. All these devices are similar in operation, and make use of a combination of charge injection and transit-time effects to create an active device. The primary difference between the devices relates to the physical mechanism that is used to inject charge into the transit-time depletion region. Charge injection can occur by tunnel phenomena (TUNNETT), a combination of avalanche and tunnel injection (MITATT), or thermionic emission over an electrostatic barrier (BARITT), as well as avalanche breakdown. The avalanche ionization process is the most effective in generating a phase delay and IMPATT diodes produce the greatest power-frequency performance. IMPATT diodes can have the ratio between the ) approach 60% in materials RF and dc voltage (i.e., such as GaAs and InP, and this permits conversion efficiency on the order of 40% to be obtained. Also the current density in IMPATTs is very high and this translates into high RF output power. IMPATTs are the most powerful solid-state devices available. However, the IMPATT oscillator does suffer from relatively high noise performance. The avalanche ionization process is a high energy statistical process and is characterized by relative high noise. Injection locking techniques can be used to reduce the in-band noise. The IMPATT diode has found wide spread use in practical systems. The RF power performance for IMPATT and other transit-time diodes is shown in Fig. 20. The devices produce on the order of watts RF power in Ka-band, and good RF output power, on the order of tens of milliwatts,

The history of vacuum device and solid-state device development share much common ground. Electron tubes were the first practical electronic devices developed and most work with semiconductors consisted of attempts to repeat vacuum diodes and triodes in a solid-state environment. A variety of two-terminal diode and three-terminal transistor structures have been proposed and successfully demonstrated. These devices have had a major impact upon the development of microwave and millimeter-wave electronic systems. The development of high-performance solid-state devices has been closely linked to the availability of suitable semiconductor materials and related process technology, and considerable effort has been directed toward the development of advanced semiconductor materials improved parameters. This work started with early device demonstrations in Ge and quickly moved to Si, and then to GaAs, InP, and ternary III–V compounds such as AlGaAs and InGaAs. Heterostructures such as AlGaAs–GaAs and GaInAs–InP are also possible and have significant advantages for device development. Currently, wide bandgap semiconductors such as SiC, GaN, and the AlGaN–GaN heterostructure are being developed and demonstrate promise for fabrication of improved microwave and millimeter-wave devices. The basic properties, power generation capabilities and state-of-the-art experimental results of a variety of two-terminal and three-terminal solid-state devices were presented. These solid-state devices have provided active sources for use as oscillators and amplifiers from UHF to terahertz frequencies. New materials such as GaN and SiC have the potential of increasing the power output significantly and ultimately vacuum-based ballistic devices may be used for generations of significant power levels at terahertz frequencies. REFERENCES [1] F. Braun, “Uber die stromleitung durch schwefelmetalle,” Ann. Phys. Chem., vol. 153, pp. 556–556, 1874. [2] J. A. Fleming, “Instrument for converting alternating currents into continuous currents,” U.S. Patent 803 684, 1905. [3] L. DeForest, “Device for amplifying feeble electrical currents,” U.S. Patent 841 387, 1907. [4] R. J. Trew, “SiC and GaN transistors: Is there one winner for microwave power applications?,” Proc. IEEE, vol. 90, no. 6, pp. 1032–1047, Jun. 2002. [5] G. I. Haddad and R. J. Trew, “Microwave solid-state active devices,” IEEE Microw. Theory Tech., vol. 50, no. 3, pp. 760–779, Mar. 2002. [6] J. E. Lilienfeld, “Method and apparatus for controlling electric currents,” U.S. Patent 1 745 175, 1930. [7] , “Device for controlling electric current,” U.S. Patent 1 900 018, 1933. [8] O. M. Stuetzer, “A crystal amplifier with high input impedance,” Proc. IRE, vol. 38, pp. 868–868, Aug. 1950. [9] , “Junction fieldistors,” Proc. IRE, vol. 40, pp. 1377–1381, Nov. 1952. [10] W. Shockley, “A unipolar "field-effect" transistor,” Proc. IRE, vol. 40, pp. 1365–1365, Nov. 1952. [11] J. Bardeen and W. H. Brattain, “The transistor, a semiconductor triode,” Phys. Rev., vol. 71, pp. 230–230, 1948. [12] W. Shockly, “The theory of p-n junction in semiconductors and p-n junction transistors,” Bell Syst. Tech. J., vol. 28, pp. 435–435, 1949.


[13] S. Middelhoek, “Projection masking, thin photoresist layers and interface effects,” IBM J. Res. Develop., vol. 14, pp. 117–124, Mar. 1970. [14] J. Turner, A. Waller, R. Bennett, and D. Parker, “An electron beam fabricated GaAs microwave field-effect transistor,” in Proc. GaAs Related Compounds Symp. (Inst. Phys. Conf. Series), London, U.K., 1971, pp. 234–239. [15] P. Wolf, “Microwave properties of schottky-barrier field-effect transistors,” IBM J. Res. Develop., vol. 14, pp. 125–141, Mar. 1970. [16] K. Drangeid, R. Sommerhalder, and W. Walter, “High-speed galliumarsenide schottky barrier field-effect transistors,” Electron. Lett., vol. 6, pp. 228–29, Apr. 1970. [17] W. Baechtold, W. Walter, and P. Wolf, “X and Ku band GaAs MESFET,” Electron. Lett., vol. 8, pp. 35–37, Jan. 1972. [18] H. Kroemer, “Theory of a wide-gap emitter for transistors,” Proc. IRE, vol. 45, pp. 1535–1537, Nov. 1957. [19] J. I. Song, K. B. Chough, C. J. Palmstrom, B. P. Van der Gaar, and W. P. Hong, “Carbon-doped based InP/InGaAs HBTs with f t = 200 GHz,” in 52nd Annu. Device Research Conf. Dig., 1994, pp. 97–98. [20] M. J. Rodwell et al., “Submicron scaling of HBTs,” IEEE Trans. Electron Devices, vol. 48, no. 11, pp. 2606–2624, Nov. 2001. [21] S. S. Iyer, G. L. Patton, S. S. Delage, S. Tiwari, and J. M. C. Stork, “Silicon-germanium base heterojunction bipolar transistors by molecular beam epitaxy,” in IEDM Tech. Dig., 1987, pp. 74–76. [22] T. H. Ning, “History and future perspective of the modern silicon bipolar transistor,” IEEE Trans. Electron Devices, vol. 48, no. 11, pp. 2485–2491, Nov. 2001. [23] H. L. Stormer, R. Dingle, A. C. Gossard, W. Wiegmann, and M. D. Sturge, “Two-dimensional electron Gas at differentially doped GaAs–AlGaAs heterojunction interface,” J. Vac. Sci. Technol., vol. 16, pp. 1517–1519, 1979. [24] T. Mimura, S. Hiyamizuk, T. Fujii, and K. Nanbu, “A new field-effect transistor with selectively doped GaAs–n-AlGaAs heterostructures,” Jpn. J. Appl. Phys., vol. 19, pp. L225–L227, 1980. [25] H. G. Henry et al., “S-band operation of SiC power MESFET with 20 W (4.4 W/mm) output power and 60% PAE,” IEEE Trans. Electron Devices, vol. 51, no. 6, pp. 839–845, Jun. 2004. [26] R. C. Clarke and J. W. Palmour, “SiC microwave power technologies,” Proc. IEEE, vol. 90, no. 6, pp. 987–992, Jun. 2002. [27] Y.-F. Wu, A. Saxler, M. Moore, R. P. Smith, S. Sheppard, P. M. Chavarkar, T. Wisleder, U. K. Mishra, and P. Parikh, “30-W/mm Gan HEMTs by field plate optimization,” IEEE Electron Device Lett., vol. 25, no. 11, pp. 117–119, Nov. 2004. [28] U. K. Mishra, P. Parikh, and Y.-F. Wu, “AlGaN/GaN HEMTs-an overview of device operation and applications,” Proc. IEEE, vol. 90, no. 6, pp. 1022–1031, Jun. 2002. [29] W. Shockley, “Negative resistance arising from transit time in semiconductor diodes,” Bell Syst. Tech. J., vol. 33, pp. 799–826, Jul. 1954. [30] W. J. Read, “A proposed high-frequency negative resistance diode,” Bell Syst. Tech. J., vol. 37, pp. 401–446, Mar. 1958. [31] J. B. Gunn, “Microwave oscillation of current in III–V semiconductors,” Solid-State Comm., vol. 1, pp. 88–88, 1963. [32] B. K. Ridley and T. B. Watkins, “The possibility of negative resistance effects in semiconductors,” Proc. Phys. Soc. Lond., vol. 78, pp. 293–293, 1961.


[33] C. Hilsum, “Transferred electron amplifiers and oscillators,” Proc. IRE, vol. 50, pp. 185–185, 1962. [34] C. A. Lee, R. L. Batdorf, W. Wiegmann, and G. Kaminsky, “The read diode—An avalanching transit-time, negative-resistance oscillator,” Appl. Phys. Lett., vol. 6, pp. 89–91, Mar. 1965. [35] R. L. Johnston, B. C. Deloach, and B. G. Cohen, “A silicon diode microwave oscillator,” Bell Syst. Tech. J., vol. 44, pp. 369–372, Feb. 1965. [36] “Special issue on solid-state microwave and millimeter-wave power generation, amplification, and control,” IEEE Trans. Microw. Theory Tech., vol. MTT-27, May 1979.

Robert J. Trew (F’91) received the Ph.D. degree from the University of Michigan, Ann Arbor, in 1975. He is currently the Alton and Mildred Lancaster Distinguished Professor of Electrical and Computer Engineering and Head of the Electronic and Computer Engineering Department (ECE) at North Carolina State University, Raleigh. From 2001 to 2002, he was the Willis G. Worcester Professor of Engineering and Head of the ECE Department of Virginia Tech, Blacksburg. From 1997 to 2001, he was Director of Research for the U.S. Department of Defense (DoD), with management oversight responsibility for the $1.3 billion yearly basic research programs of DoD. During this time, he managed DoD’s University Research Initiative, which includes the MURI, DURIP, DEPSCoR, and HBCU/MI programs. He also served as a Program Manager in the Electronics Division of the U.S. Army Research Office from 1992 to 1997. From 1993 to 1997, he served as George S. Dively Professor of Engineering and Chair of the Electrical Engineering and Applied Physics Department of Case Western Reserve University, Cleveland, OH. He has written over 140 publications, 15 book chapters, and has given over 320 technical and programmatic presentations, and has seven patents. Dr. Trew served as Vice-Chair of the U.S. Government interagency committee that planned and implemented the U.S. National Nanotechnology Initiative (NNI). He serves on the IEEE Microwave Theory and Techniques Society Administration Committee (AdCom) and was President for 2004. He was the recipient of the 2001 IEEE-USA Harry Diamond Memorial Award. He also received an IEEE Third Millennium Medal Award, the 1998 IEEE MTT Society Distinguished Educator Award, the 1991 Alcoa Foundation Distinguished Engineering Research Award, and a 1992 NCSU Distinguished Scholarly Achievement Award. He received an Engineering Alumni Society Merit Award from the University of Michigan in 2003. He was Editor-in-Chief of the IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES from 1995 to 1997, and from 1999 to 2002, was founding Co-Editor-in-Chief of the award-winning IEEE MICROWAVE MAGAZINE. He is also a member of the Editorial Board of the IEEE PROCEEDINGS. He was an IEEE Microwave Distinguished Lecturer from 1997 to 1999, and is currently serving a second term as Microwave Distinguished Lecturer for the period 2003 to 2005. He is a Member of the Materials Research Society, the Electromagnetics Academy, AAAS, ASEE, Sigma Xi, Eta Kappa Nu, and Tau Beta Pi.