Technical Insight
6.1 Å devices use less power
Set up to develop super-fast InAs-based devices, DARPA's antimonides program could open up a whole new range of applications where low power consumption is desired. Tim Whitaker reports.
As electronic systems become more complex, more mobile and more demanding in terms of frequency and bandwidth, there is a need to develop new technologies that offer improved performance, particularly in the area of power consumption. This is one of the key factors behind DARPA s antimonide-based compound semiconductors (ABCS) program, which began in July 2001 and has just under 18 months left to run. The major participants are HRL, Rockwell Scientific and Northrop Grumman Space Technology (NGST).
The program s original motivation was to develop devices with extremely high ft and fmax values. Although this remains important, the progress made on programs such as TFAST, which is developing super-fast InP-based devices, has shifted the emphasis of ABCS towards low-power operation.
"The potential impact of this technology could be very large," said DARPA s Mark Rosker, program manager of ABCS. "A tremendous number of applications are power-limited, and any time you can introduce a technology which can greatly reduce the power draw, this is highly attractive."
Antimonide-based ICs have already demonstrated the potential to offer a 10-fold reduction in power consumption compared with equivalent InP-based devices. The technology is very promising for low-voltage mm-wave MMICs, as well as very high-speed, low-power digital ICs. Among the many potential applications are digitally beam-steered phased array receivers, particularly those deployed in space; 80-160 Gbit/s optical communication systems; unmanned reconnaissance vehicles; and mobile, battery-powered systems in general.
Materials issuesAlthough the DARPA program name includes "antimonide-based" to describe the material system, the devices under development are HEMTs and HBTs that use InAs for the critical electron transport layers (the HEMT channel or the HBT base). InAs has electron mobility and peak velocity values that are about twice as high as In0.53Ga0.47As, the channel material used in lattice-matched HEMTs grown on InP.
InAs-based device structures are grown with a lattice constant of around 6.1 Å (see figure) and include antimonide-containing layers, such as the AlSb barrier layers usually used in InAs-channel HEMTs. A key problem with growing InAs/antimonide devices is the lack of a lattice-matched semi-insulating substrate. However, the problem has been circumvented using relaxed metamorphic Sb-containing buffers, grown on GaAs or InP.
"Early in the program, the main emphasis was on substrate technology, but this has proved less of an issue than originally supposed," said Rosker. "Despite the lattice mismatch, we have groups developing metamorphic devices on InP and GaAs without significant degradation in performance."
Device performanceAs well as overcoming the substrate issue, participants have concentrated on improving the performance and complexity of individual devices. The third phase of the program, currently under way, is to increase integration levels and develop MMIC technology.
Rockwell developed the first antimonide-based MMIC about a year ago, says Bobby Brar, manager of the company s ABCS program. "The MMIC was a HEMT-based, three-stage low-noise amplifier operating at 35 GHz with a power consumption of about 4 mW," he said. The HEMT had a 100 nm gate and ft and fmax values of around 300 and 200 GHz, respectively. Brar says that attempts will be made to reproduce this level of performance at frequencies up to 100 GHz, opening a whole new range of applications.
The participants have all reported ft and fmax figures in excess of 200 GHz, while at last December s IEDM, HRL reported InAs-channel HEMTs with ft = 308 GHz. The device structures were grown by MBE (a technique used by all the ABCS participants) on InP substrates using a metamorphic AlGaAsSb buffer.
Meanwhile, NGST and the Naval Research Laboratory reported AlSb/InAs HEMTs that use a MMIC integration process identical to NGST s space-qualified InAlAs/InGaAs/InP HEMT process. The AlSb/InAs HEMTs provided equivalent high-speed performance at 5-10 times lower power dissipation.
While HEMTs are primarily being used for analog applications, antimonide-based HBTs are providing performance advantages for digital ICs. The key parameters are the level of integration and the power-delay product (the power consumption multiplied by time delay for each logic operation in the IC). ABCS is targeting levels of 5 fJ per operation, as well as integration levels of a few thousand transistors. NGST has demonstrated a divide-by-four circuit with a clock frequency of 21 GHz that contains around 100 HBTs.
Ken Elliott, program manager at HRL, says that his company has made a 62-transistor ABCS HBT divider operating at 27 GHz, using a device with simultaneous ft and fmax values of 175 and 100 GHz. (With a different layout the peak ft was 215 GHz.)
Given the progress made so far, it looks certain that the performance advantages of the 6.1 Å family of compound semiconductors will see them emerge as an important class of device over the next few years.
The program s original motivation was to develop devices with extremely high ft and fmax values. Although this remains important, the progress made on programs such as TFAST, which is developing super-fast InP-based devices, has shifted the emphasis of ABCS towards low-power operation.
"The potential impact of this technology could be very large," said DARPA s Mark Rosker, program manager of ABCS. "A tremendous number of applications are power-limited, and any time you can introduce a technology which can greatly reduce the power draw, this is highly attractive."
Antimonide-based ICs have already demonstrated the potential to offer a 10-fold reduction in power consumption compared with equivalent InP-based devices. The technology is very promising for low-voltage mm-wave MMICs, as well as very high-speed, low-power digital ICs. Among the many potential applications are digitally beam-steered phased array receivers, particularly those deployed in space; 80-160 Gbit/s optical communication systems; unmanned reconnaissance vehicles; and mobile, battery-powered systems in general.
Materials issuesAlthough the DARPA program name includes "antimonide-based" to describe the material system, the devices under development are HEMTs and HBTs that use InAs for the critical electron transport layers (the HEMT channel or the HBT base). InAs has electron mobility and peak velocity values that are about twice as high as In0.53Ga0.47As, the channel material used in lattice-matched HEMTs grown on InP.
InAs-based device structures are grown with a lattice constant of around 6.1 Å (see figure) and include antimonide-containing layers, such as the AlSb barrier layers usually used in InAs-channel HEMTs. A key problem with growing InAs/antimonide devices is the lack of a lattice-matched semi-insulating substrate. However, the problem has been circumvented using relaxed metamorphic Sb-containing buffers, grown on GaAs or InP.
"Early in the program, the main emphasis was on substrate technology, but this has proved less of an issue than originally supposed," said Rosker. "Despite the lattice mismatch, we have groups developing metamorphic devices on InP and GaAs without significant degradation in performance."
Device performanceAs well as overcoming the substrate issue, participants have concentrated on improving the performance and complexity of individual devices. The third phase of the program, currently under way, is to increase integration levels and develop MMIC technology.
Rockwell developed the first antimonide-based MMIC about a year ago, says Bobby Brar, manager of the company s ABCS program. "The MMIC was a HEMT-based, three-stage low-noise amplifier operating at 35 GHz with a power consumption of about 4 mW," he said. The HEMT had a 100 nm gate and ft and fmax values of around 300 and 200 GHz, respectively. Brar says that attempts will be made to reproduce this level of performance at frequencies up to 100 GHz, opening a whole new range of applications.
The participants have all reported ft and fmax figures in excess of 200 GHz, while at last December s IEDM, HRL reported InAs-channel HEMTs with ft = 308 GHz. The device structures were grown by MBE (a technique used by all the ABCS participants) on InP substrates using a metamorphic AlGaAsSb buffer.
Meanwhile, NGST and the Naval Research Laboratory reported AlSb/InAs HEMTs that use a MMIC integration process identical to NGST s space-qualified InAlAs/InGaAs/InP HEMT process. The AlSb/InAs HEMTs provided equivalent high-speed performance at 5-10 times lower power dissipation.
While HEMTs are primarily being used for analog applications, antimonide-based HBTs are providing performance advantages for digital ICs. The key parameters are the level of integration and the power-delay product (the power consumption multiplied by time delay for each logic operation in the IC). ABCS is targeting levels of 5 fJ per operation, as well as integration levels of a few thousand transistors. NGST has demonstrated a divide-by-four circuit with a clock frequency of 21 GHz that contains around 100 HBTs.
Ken Elliott, program manager at HRL, says that his company has made a 62-transistor ABCS HBT divider operating at 27 GHz, using a device with simultaneous ft and fmax values of 175 and 100 GHz. (With a different layout the peak ft was 215 GHz.)
Given the progress made so far, it looks certain that the performance advantages of the 6.1 Å family of compound semiconductors will see them emerge as an important class of device over the next few years.
