DARPA propels InP electronics towards a terahertz
Increasing the operating frequency of transmitters and detectors can produce a range of benefits. These include communication at higher data rates and identification of objects that are highly absorbing at these higher frequencies. Imaging systems can also be improved, because switching to higher frequencies can slash the size of these systems.
The latter benefit has encouraged the US Defense Advanced Research Projects Agency (DARPA) to fund two consecutive programs that require the re-defining of the state-ofthe- art for high-frequency transistors and associated circuits. The first, Sub-millimeter Wave Imaging Focal Plane Technology (SWIFT), kicked-off in February 2006 with the goal of building an imaging system operating at 340 GHz, a frequency window of low absorption from atmospheric gases. Imaging systems at this frequency promise to offer two significant benefits over existing equivalents at other frequencies. Imagers operating at far higher frequencies are of limited use in dusty, foggy and smoky conditions, and 340 GHz systems could deliver far more informative images, thanks to greater penetration of the radiation at this frequency. Imagers at the 340GHz frequency should also be far more compact than lower frequency counterparts operating at millimeter waves, because it should be possible to realize the same spatial resolution with an aperture that is an order of magnitude smaller.
This substantial reduction in aperture size is a very important consideration for many government applications, according to DARPA program manager John Albrecht. “Consider a helicopter attempting to navigate a dust storm. While millimeter-wave radiation offers a way to see through the dust, the size of the millimeter wave imager – generally about one meter – is too large to be practical. A sub-millimeter-wave imager can do the job with an aperture size that will fit onto the platform.”
The first phase of the SWIFT program focused on the development of 340 GHz amplifiers and detectors. Efforts were restricted to III-V technologies, because these can form compact devices that are easily integrated with other pieces of electronics. Most of the targets for the first phase were realized, with those involved pioneering the fabrication of amplifiers and oscillators operating at frequencies above 300 GHz. “In the process, new infrastructure, benchmarks and metrology were created that will open the sub-millimeter wave frequency domain for future designs and systems,” says Albrecht.
More than 60 publications have stemmed from these achievements, which have included the development of 35 nm InP HEMTs, 0.3 μm-emitter HBTs, and monolithic integrated circuits operating at sub-millimeter wave frequencies. The only elusive goal was that of a 50 mW output power for the power amplifier.
Exceeding expectations
Given the great successes in phase I of the program, one would expect the funding of the second phase to be a formality. But this has not happened because DARPA believes that the best way forward is to pursue even more ambitious targets. “The device technology achieved unexpectedly high gain of typically 8dB per stage at 340 GHz, and suggested that much higher frequencies in transistor electronics were feasible,” explains Albrecht. “This was the inspiration that led to a new program entitled terahertz electronics.” Its ultimate aim is to treble the operating frequency of the devices in the SWIFT program, while maintaining their impressive power, noise figure and bandwidth characteristics. However, it’s not just a follow-on to SWIFT, says Albrecht: “The objective is to develop the critical device and integration technologies to realize compact, high-performance electronic circuits operating at center frequencies exceeding 1.0 terahertz.”
One of the key contractors in both of DARPA’s programs is Northrop Grumman Aerospace Systems (NGAS). This company hit the headlines in late 2007 with the claim for the world’s first terahertz transistor. The results produced by its ground-breaking InP HEMT were presented at the International Electron Devices Meeting (IEDM), and Northrop Grumman now has an entry in the Guinness World Records.
“What we’ve accomplished since then is to build a suite of circuits that demonstrate all the major active functions that you would need in the 340 GHz window,” explains Richard Lai, the head of the company’s microelectronics products. This includes fundamental oscillators for creating sub-millimeter-wave frequency sources, power amplifiers for boosting their intensity, and lownoise amplifiers (LNAs) that increase the intensity of a signal before it is detected. The IEDM results were obtained on-wafer, and packages have now been developed to house these chips.
The IEDM presentation included details of a LNA that produced 15 dB of gain at 340 GHz. Since then this amplifier’s operating frequency has been increased to 350 GHz, and it has been packaged into gain blocks with waveguides for coupling radiation in to and out of the chip.
A more recent edition to Northrop Grumman’s product portfolio is a 10 mW power amplifier. Lai says that the fabrication of this amplifier was one of the biggest challenges of the program. The InP HEMTs used in this type of amplifier have a relatively low breakdown voltage, and consequently a relatively low power density, so fingers are added to boost the output.
NGAS embarked on the terahertz electronics program in April 2009, and it is now working towards the phase I targets, which are defined at 670 GHz (see table for details). Researchers are in the middle of performing some fundamental device studies, and they are looking to exploit HEMT and HBT technologies.
DARPA is also funding another team on its terahertz electronic program, which is being led by Teledyne Scientific and Imaging. This firm played a smaller role in the SWIFT program - building an amplifier that produced a couple of milliWatts at 340 GHz. In addition, it has been a key participant in the Technology for Frequency Agile Digitally Synthesized Transmitters (TFAST) program, which has helped it to hone its expertise in high-frequency electronics. The TFAST project demanded the development of transistors with cutoff frequencies of up to 500 GHz, and fabrication of complex circuits based on these devices. 100 transistors are needed in very fast dividers, and 5000 are used to make digital synthesizers. “They had yield goals as well,” explains Bobby Brar, executive director of the electronics division at Teledyne. “End-of-program goals were 50-60 percent yield.”
Teledyne produced these circuits with InP HBTs, a technology that it will use throughout the terahertz electronics program. According to Brar, one of the advantages of using the HBT, rather than the HEMT, is that it can be used to build all the circuits needed by the program. “If successful, we would have one circuit that would be monolithic. And that’s a big deal, because getting on and off a chip at those kinds of frequencies is non-trivial.” Another advantage of the HBT is that it can be manufactured with much tighter threshold voltages than a HEMT, which simplifies the design of digital blocks, dividers and mixers. In addition, the problems associated with parasitics are less of a challenge when scaling HBTs.
Teledyne’s partners
Teledyne carries out everything from the fabrication of devices to chip packaging, and many of its partners in the terahertz electronics project are there to provide guidance when needed. If the company runs into problems related to transistors or circuits, then it can turn to Mark Rosker at the University of California, Santa Barbara. And if it is experiencing difficulties related to the coupling of electromagnetic radiation on to the chip, or off of it, then it can consult with Gabriel Rebeiz from the University of California, San Diego. The Jet Propulsion Laboratory is another important team member in Teledyne’s effort, and its role is to measure the device and circuit performance. Brar says that the team at the Jet Propulsion Laboratory is a world leader for this type of measurement, which is essential for meeting the program goals. The final member of the team is Raytheon. Its expertise lies with systems, rather than components, and it is well positioned to advise on how to apply the technology.
To increase the operating frequencies of its HBTs, Teledyne is shrinking device dimensions. “Making smaller dimensions is a piece of cake these days,” claims Brar. “But you have got to make sure that you scale your transistors in such a way that you get good contact resistances, low parasitic capacitances, and low leakage currents.” Teledyne is able to draw on the work of others regarding approaches to scale InP bipolar transistors. “Mark Rodwell has essentially published a roadmap for scaling InP bipolar, so we know what the challenges are,” says Brar.
Reducing parasitic resistances is one of the bigger challenges. Preventing any increase in overall resistance during the scaling process requires improvements in contact resistances that are commensurate with reductions in area. Realizing this gets tougher and tougher as the devices get smaller, because the periphery dimensions increase relative to those of the bulk as the transistor is scaled. “If you have good junctions that are well-passivated at 0.5 μm emitter dimensions, at 125 nm the problem becomes at least four times worse,” says Brar. “The area has shrunk, but the perimeter-to-area ratio has increased.”
Another challenge facing the engineers involved in the terahertz electronics program relates to circuit design. Care is needed to avoid issues cropping up at higher frequencies that have no impact at tens of gigahertz. “Having done the work at 340 GHz, we are confident that we can handle those kinds of problems,” says Brar. Potential pitfalls include losses in the semiconductors, and issues that arise because the dimensions of the substrate are comparable to the wavelength of the radiation. The latter can lead to the excitation of modes that hamper coupling of radiation between the chip and the outside environment.
Teledyne is still to publish its best results, but Brar says that they are “pretty close” to the goals at 670 GHz. He believes that it will not be long before his team is able to conceive of making circuits at this frequency. Values for the HBT’s maximum oscillation frequency are closing in on one terahertz, and the cut-off frequencies are hitting half a terahertz.
Similar outlooks
Brar and Lai both believe that one of the biggest challenges of the terahertz electronics program is the fabrication of waveguides for coupling the radiation in and out of the chips. Terahertz systems already exist in research labs, but they are very limited in the amount of power they can provide, because of the high losses at these frequencies. “We want to take power and sensitivity and improve it by several orders of magnitude,” says Lai. Realizing this would redefine the types of systems and measurements that are possible. Teledyne’s approach, which Lai claims is similar to that of NGAS, involves fabrication of waveguides by silicon micro-machining. The radiation’s wavelength is a fraction of a millimeter, so tolerances of tens of microns are needed for the features. “But smoothness tolerance are very important too,” says Brar, as otherwise losses are unacceptably high.
Both teams are aiming to develop production processes with good yields, rather than embarking on a quest for a ‘hero result’ from a single, unrepeatable device. “We’re not a university,” explains Brar. “We want to grow a technology that goes into a product.” The efforts of him and his co-workers are already having an impact on lower-frequency devices, which can now be manufactured with greater control and uniformity. NGAS is also starting to reap the rewards of its 35 nm gate process that has been used to produce hundreds of wafers. Circuits have been demonstrated with 25-30 transistor cells and 50-100 fingers with good functional and RF yield. And institutions such as the Jet Propulsion Laboratory have adopted the 35 nm process to make circuits operating at lower frequencies. “That shows the maturity of our process,” says Lai. “We can take our device process, share the device models that we have, and others can go and design this stuff successfully.” That means that DARPA’s programs can be viewed as money well spent, an important virtue in tough economic times. Not only have these efforts improved transistor performance; they have started to impact the characteristics of systems operating at lower frequencies.
