Designers can take advantage of the latest fabrication techniques for and the expanding array of materials used in nonlinear devices (see “Use Nonlinear Devices As Linchpins To Next-Generation Design,” July 8). Yet they should also explore the potential applications for components that exhibit negative differential resistance (NDR).
Perhaps the most obvious application is the oscillator. Designers can choose one of many different approaches for building an oscillator. A specific NDR device isn’t required, but even the simplest of inductor-capacitor or crystal oscillators works via a negative resistance, or more broadly speaking, negative impedance, element. Many commercially available components offer the effective negative resistance needed to assemble these circuits.
When it comes to the fundamental importance of negative resistance in producing oscillations, oscillation in simple single-device circuits always requires the existence of negative resistance, in some frequency ranges, at either input or output terminals.1 Capacitive feedback is one simple way to produce negative resistance by means of an out-of-phase voltage.
Negative-resistance oscillators are known by familiar names such as Van der Pol, Barkhausen, Kurokawa, Pierce, and Colpitts. Other possible applications in this arena include highly frequency-stable crystal oscillators, voltage-controlled oscillators, and mixers.
Many NDR devices truly excel in microwave frequencies. Gunn and impact ionization avalanche transit time (IMPATT) diodes are used in millimeter wavelengths (see “Use Nonlinear Devices As Linchpins To Next-Generation Design,” again). Consequently, 10-GHz oscillator and mixer applications have become standard.
Now, engineers are solidifying apps at the 700-GHz plateau and are looking to extend performance into the terahertz range. As so often is the trend in electronics to higher-frequency operation, such is the case in the world of NDR devices as well.
DIGITAL LOGIC AND MEMORIES
If negative resistance can produce oscillation between two states, it’s also possible to hold either state for some length of time to function as a latch or switch. Indeed, NDR-based bistable devices often find their way into digital memories. It’s been shown with tunneling-based SRAM, and research is underway involving the optimization of silicon-based cells compatible with CMOS.
Of particular interest in digital applications is the potential to go beyond conventional binary logic. Ongoing research involves NDR technology being used to produce multiple stable logic states. In fact, three and four logic states have been demonstrated in gallium-arsenide (GaAs) resonant tunneling structures. The figure shows the extension of a single NDR peak and valley to three in succession.
The use of multiple NDR regions to achieve multiple logic states may result in highly dense devices and fewer circuit elements than prior generations of transistor-transistor logic (TTL), emitter-coupled logic (ECL), or CMOS logic. More than three states have also been demonstrated with the potential for an unlimited number of NDR peaks observed in a device containing a single-electron transistor and a MOSFET.2
Peak-to-valley ratio (PVR) is often mentioned as a figure of merit for NDR devices, and it offers a way to aid in comparing the performance of different devices. But for logic applications, a PVR usually needn’t be more than 5 to 10. For large-scale memories, though, minimization of standby power consumption is critical and demands as high a PVR as possible. Still other researchers feel that capacitance and peak current density are more appropriate figures of merit for NDR switches than PVR.
As we have noted, up until recently, the adoption of tunnel and related devices has been hindered by the difficulty in obtaining this type of behavior in structures that are compatible with silicon-based systems, rather than III-V materials. So while progress in silicon implementation continues, another strategy can be to employ a three-terminal structure like a gate-controlled diode.
Continue to next page
Unlike a conventional diode that has no NDR properties, a gate-controlled type can produce an N-shaped I-V curve due to a sudden drop in leakage current when transitioning from depletion to inversion mode. Memories based on generic p-n-p-n devices, with construction similar to that commonly employed as thyrsistors, diacs, and triacs, are also possible.
Research in reducing device size to the smallest possible dimensions has shown that single-electron tunneling, or its complementary device, the single-hole transistor, can be effective for logic applications. NDR and conductance switching are now commonly found throughout the most advanced molecular electronic junction research.
In perspective, engineer John von Neumann and others proposed similar circuit functionality 50 years ago in the “parametron,” a complex arrangement of universal elements of oscillators and clocks. Such foresight converging with advances in molecular electronics highlights the potential for an almost unlimited number of switches that can be used as logic elements. We’re on the verge of achieving the dream of exceedingly small and fast versatile switches that are capable of multiple states.
Beyond oscillators and logic, NDR applications can also extend to high-frequency mixers and analog-to-digital converters. These applications have only begun to be fully explored, possibly because of the many other existing solutions that can meet today’s requirements. The future of NDR devices remains wide open for new applications.
EMERGING MATERIALS TO ENABLE NDR DEVICES
The extent of different materials that yield NDR behavior is vast. For instance, chalcogenides—Group VI elements like oxygen and sulfur, but most notably selenium and tellurium—have a strong propensity to yield nonlinear electrical behavior. Rather than discuss in detail the many possible material sets, Tables 1-3 list examples of the types of different material categories that can produce NDR.
The most sought-after application relates to new types of memory. The diversity of possible approaches to building memories is quite astounding. Recent research in NDR behavior has shown lots of promise in the realm of nano and molecular dimensions. Tailoring devices to specific material requirements and performance requirements is now possible.
Table 1 lists examples of devices made with inorganic materials that aren’t traditional semiconductors. The types of I-V behavior, achievable PVR, and aspects like frequency of operation illustrate the variety of ways to make these emerging NDR devices.
When conjugated to form large continuous bands capable of conduction or doped with additives that contribute to improved conduction, organic (carbon-based) compounds constitute a relatively new class of electronic devices. Some call them “plastic diodes” or similar names. Future uses may support the evolution of hardware from mechanically placed and soldered components to printed electronics. Table 2 offers a sampling of research in organic and molecular systems.
Research in carbon nanotubes and similar structures based on derivatives of fullerenes or buckyballs has also yielded many variants of NDR and switching behavior. Table 3 provides a sampling of devices built with these materials. It’s clear by now that NDR also is found in structures unrelated to traditional electronic devices, within the chemical and biological domains.
As an example of the latter, digital memory was reported in a virus combined with platinum nanoparticles.3 Though examples exist beyond the biological and even psychological sciences, they’re considered as outside the scope of this article.
We saved photovoltaic applications was saved for last because the mode of operation is unrelated to nonlinearity. Having begun the discussion with the tunnel diode, though, its importance in multi-junction solar cells must be mentioned.
To improve light-gathering efficiency at different wavelengths, some photovoltaic designs stack three or more cells that require electrical connections between them. One constraint is that the connections must not limit the transmission of light through the structure.
The choice of ohmic contact material has generally been none other than tunnel diodes. Tunneling behavior isn’t used here at all, since these junctions are operated in their linear, not NDR, region. They’re typically fabricated with heavy doping to pass as much current as possible. However, it’s a testament to the universality of NDR devices that they’re increasingly helping power our world—even while operating only in their linear region!
1. Glover, S.R. Pennock, and P.R. Shepherd, Microwave Devices, Circuits and Subsystems for Communications Engineering, 2005, p. 227; www.ebookmall.com
2. H. Inokawa, A. Fujiwara, and Y. Takahashi, “Multipeak Negative-Differential Resistance Device by Combining Single-Electron and Metal-Oxide-Semiconductor Transistors,” Applied Physics Letters, vol. 79, pp. 3618-3620, Nov. 2001
3. A.N. Androitis, M. Menon, D. Srivastava, and L. Chernozatonskii, “Ballistic Switching and Rectification in Single Wall Carbon Nanotube Y Junctions,” Applied Physics Letters, vol. 79, pp. 266-268, July 2001
Roger L. Franz, formerly an engineering manager in Motorola’s Mobile Device Business, holds a BA from Grinnell College, Grinnell, Iowa, and an MS from Northwestern University, Evanston, Ill.