Advertisement
Articles
Advertisement

Mario Paniccia: Conductor of Light

Sun, 11/30/2008 - 7:00pm
Paul Livingstone

At an Intel laboratory, R&D Magazine’s 2008 Scientist of the Year is designing the chips that are revolutionizing photonics and pointing the way to the terascale age of optical communications.

Mario Paniccia

Mario Paniccia was voted 2008 Scientist of the Year by R&D Magazine's readers.

The year 2000 was not a great one for starting a new research project, particularly one that defied conventional wisdom across two industries.

Had Mario Paniccia, Intel fellow and director of the Photonics Technology Laboratory, Santa Clara, Calif., been paying close attention to the markets, he might have questioned his own wisdom instead. On March 10, 2000, the dot-com bubble burst, sending the NASDAQ into a free-fall that didn't halt until 2001, by which time it had nearly halved in value.

Intel Corp. was hit hard by the sudden glut of high-speed servers and chips. Pentium-4 had just entered the marketplace, wireless technology had yet to enter laptops, and the company was in the process of re-examining its corporate culture and approach to research.

Paniccia, who started with the company in 1995, was unfazed by the convulsions in the marketplace. His thoughts had already veered into an unconventional direction, one that led him to propose a project that most scientists assumed was not possible or practical. But in the face of skeptics, uncertain budgets, and technical hurdles, he realized his vision of siliconizing photonics, an achievement that has led R&D Magazine to select him our 2008 Scientist of the Year.

Since establishing the Photonics Technology Lab in 2001, Paniccia has become one of the world's most recognized corporate researchers. His original premise—that silicon could form the basis for optical links for computing—was validated in 2004 when he and his team announced the demonstration of the first gigahertz silicon modulator. The discovery struck a chord with the scientific community and wider world—instantly more than 150 news outlets hopped on the story, putting Paniccia in the science celebrity spotlight.

“It surprisingly turned out to be a huge thing. It was amazing,” says Paniccia of the unexpected attention. “People were saying if you can do this in silicon, maybe we can build all the devices out of silicon and replace this big bulky expensive optical devices with integrated optical devices made out of silicon.”

Questions remained—could we really build advanced photonics devices on cheap silicon? If so, could we use these silicon devices in place of communications devices made from exotic materials that are expensive and hard to make? Could it be used to build an entire optical communication link and be inserted into a computing system?

Paniccia's goal is to answer these questions, because they have profound ramifications on the future of the semiconductor industry, which is built on silicon. If devices such as a silicon-based transceiver are possible, then the entire communications industry can leverage the tremendous capacity of CMOS fabrication, greatly reducing overhead. Better yet, data rates would rise into the oft-imagined terabit realm.

As Paniccia believed, the excitement in 2004 was just the beginning. In just four years, speed records have been broken and new devices have been pioneered. And in 2008 silicon photonics devices are knocking on the commercial sector's door.

Cascaded Silicon Raman Laser

One milestone at the Photonics Technology Lab was the first cascaded silicon Raman laser (2007).

Silicon determination
Endicott, N.Y., is a long way from Silicon Valley, but Paniccia knew early on that the famous hotspot for technological innovation would be in his future. The son of Italian immigrants Antonio and Domenica Paniccia, neither of which went past the fifth grade and who spent their careers working at Endicott Johnson (a shoe factory), Paniccia was an accomplished and diligent student.

“We didn’t have a lot of money growing up,” he says. “My parents stressed the importance of an education, though, and I was especially good in math.” His passion for the sciences followed him to the State Univ. of New York, Binghamton, where he earned a physics degree. A professor there, Robert Pompi, prompted him apply to grad school.

Paniccia planned to attend Cornell Univ., Ithaca, N.Y., but changed his mind after a friend attending Purdue Univ., West Lafayette, Ind., invited him to a big ten football game vs the Univ. of Michigan. The experience of 60,000 people cheering at once put a bug in his head. That, coupled with reality of having to pay his own way at Cornell, led him to join the graduate physics program at Purdue.

Paniccia's first exposure to corporate research occurred at Bell Labs in N.J. and IBM in Yorktown Heights, N.Y. While getting his masters degree at Purdue, he landed an interview at Intel Corp. They gave him an offer and, intrigued with the opportunity to go to California and Silicon Valley, he accepted and in 1990 flew out West.

What shocked his parents and friends, however, was that he turned down a job at IBM for what was then a relatively little-known company. Worse, the failure analysis position he accepted at Intel was not very fulfilling.

“I unfortunately quickly got bored with the job. I wasn’t motivated,” says Paniccia. “I met the head of research at the time, John Carruthers, and ended up putting together a proposal to measure electrical migration in wires using a new technique at that time called STM and AFM. I realized I wanted to go back to get my PhD.”

Paniccia contacted Purdue about coming back. With professor Ronald Reifenberger's help he wrote a proposal that would allow Intel to help fund his PhD. It was accepted.

“My parents were shocked that I would leave a high paying job and go back to school,” he says. After completing his dissertation, he was ready for the corporate world, but despite recruiting for Intel during the interim he was never approached for a job himself. Unable to accept a position at Texas Instruments or IBM because of his interest to return to Intel, he took a drastic step

“I wrote this letter and sent it to CEO Gordon Moore and everybody,” says Paniccia. “I literally wrote individual letters to all of the heads of management.” The young scientist expressed disappointment by the a lack of initiative on the part of the company to hire him back.

The letters worked—he received calls for interviews just two days later. Years later, after being named an Intel fellow, a position of distinction within the company, he was sent of copy of the flurry of emails prompted by his unsolicited correspondence.

Bringing photonics to the masses
Paniccia's first major contribution at Intel was the laser voltage probe (LVP) which acquires timing waveforms on flip-chip packaged microprocessors directly through the silicon backside. The technology, which eliminated the need for a vacuum system for testing, has become an integrated circuit (IC) industry testing standard.

As a result of this project, Paniccia was introduced to optics. Specifically, he learned about photonic devices used in the growing fiber optics industry. Fiber optics is fast and dependable. Unfortunately, it's also expensive. The hardware that manages optical data is usually individually crafted from so-called III-V materials, such as indium phosphides and metals that are difficult to process or obtain.

Driven by the desire for speed, the fiber optic infrastructure nevertheless grew exponentially. Intel researchers, including Paniccia, began examining these pricey optical links: modulators, photodetectors, waveguides, and others.

“I was shocked at how much these components cost,” he says, referring to transceiver components used to interpret photonic information to electrical signals and back again. “These modules were selling from $5,000 to $10,000. I remember thinking to myself ‘This is doesn’t make sense. Why can’t we make these devices out of silicon?’”

From his experience with the LVP, Paniccia knew light could be manipulated with electrons. He felt he could use these abilities within silicon. Not many people at that time would agree, however.

“This is one of these things that people said would never happen. Even in the early years here at Intel, the skepticism was very high,” says Justin Rattner, Intel senior fellow, chief technology officer, VP and director of the Corporate Technology Group., and R&D Magazine’s Scientist of the Year in 1989. The conventional wisdom was that if you wanted to do something with light, he says, there were far better materials.

But Paniccia saw things differently. He was well aware of Intel's legacy and saw silicon photonics as another step in the direction pioneered by Intel's founders. Specifically, he was thinking of the simple graph penned in a notebook by co-founder Gordon Moore in 1965. By 2000, when Paniccia began his silicon photonics R&D, Moore's Law had become a mantra in the semiconductor industry. In short, it states that the number of transistors on a chip will double about every two years. For the last 43 years, Moore has been right. However, Intel took greater meaning from his statement—they attribute the increase in transistor count and processing power with reduced cost, largely at the same rate. This law helps explain why Intel's entire product line turns over every 24-30 months, and why it has pushed so hard to introduce 32-nm semiconductor platforms and hafnium-based high-k metal-gate transistors.

Paniccia eventually convinced Intel executive and employee # 3 Leslie Vadasz to supply him with seed money and a small team of young scientists. The program quickly morphed from demonstrating a single device to a much broader goal of siliconizing photonics. Paniccia's goal totalled six components: a laser source, a modulator, passive devices such as splitters and routers, a detector, a way to package these devices, and, of course, electronics.

“So we started to say if we can build all of these devices in silicon, just like transistors, we would be able lay them out any way we want to produce entirely new optical devices. This would change the whole way think about optical communication,” says Paniccia. His physics experience gave him confidence during the initial project—the 1 GHz silicon modulator. A fundamental component of a photonics system, optical modulators rapidly pulse a light beam to digitally encode a high-quality data signal. At the time Paniccia’s team was tasked to meet its first goal, the preceding attempt at a silicon modulator had achieved all of 2% of that speed.

“Everyone on the team said this [goal] was too aggressive,” says Paniccia. But he went even further by taking this on as a CTO-level goal, one which helped put the pressure on his small group of researchers. Fending off a near-mutiny from his team, Paniccia encouraged them to push the limits of what could be done. For some time, his project teetered on the brink of cancellation. But the leadership of Paniccia and the energy and creativity of his team drove progress forward.

Kevin Kahn, Intel senior fellow and director of the Communications Technology Laboratory, says Paniccia’s ability to be flexible and leverage resources throughout Intel while at the same time retaining his vision allowed his project to survive.

“This stuff was getting started at the tail end of the dot-com bust, and business was kind of cruddy. It wasn’t a great time for projects with a dubious product intercept potential,” says Kahn. “Early on, there seemed to be an almost limitless number of ways in which it should have been killed. It survived, I think, due to sheer willpower on Mario’s part. At every stage along the way, specifically with his own team, he put down goals his team thought were absolutely nuts. And he’s managed to exceed those goals along the way.”

Favorable results were achieved by mid-2003. When Paniccia and his team went public, they reported 2.2 GHz. It was a double-take not just for Intel, but also for an industry which had previously shunned silicon as a basis for photonics.

“When they built that first modulator, it was fundamental in that, first of all, no one imagined you could build a silicon oscillator at anything close to that frequency,” says Rattner. “But then it provided that confidence you need to take the next step.”

Shortly after announcing their results, Paniccia’s team felt they should publish their findings in a scientific journal. This was a major step for Intel, whose innovations had typically been kept under wraps for security reasons. They recognized, however, that siliconized photonics was a different animal and would spur innovations for everyone's benefit.

340 GHz Avalanch Photodectector

One milestone at the Photonics Technology Lab was the first 340 GHz avalanche photodetector (2008).

A photonic avalanche
Emboldened, in 2005 the photonics team created the first continuous wave silicon laser. The waveguide-like device, which relies on the Raman effect to multiply photons and essentially act as its own pump, was another fundamental scientific breakthrough. The team demonstrated that silicon could be in fact be used as an optical gain material which forms the basis of a laser.

Now that Paniccia’s team was able to achieve lasing on a silicon chip, their research branched out even further. In 2006, they combined the light-emitting properties of indium phosphide with the light routing capabilities of silcion to create the world’s first hybrid silicon laser in collaboration with the Univ. of California, Santa Barbara.

In 2007, Paniccia’s team combined silicon with an IR-absorbing material, germanium, to create a low-cost CMOS-compatible photodetector. With considerable effort to reduce the strain due to the lattice mismatch of germanium and silicon they produced a waveguide-based SiGe photodetector. They achieved it in part by using a common P-I-N (positive-negative) diode. The photons create a negative charge and positive "hole" when it strikes the detector. Voltage applied across the detector pushes the electrons and holes apart, allowing the photodetector to read flashes of light at up to 40 Gbps, which matched commercial devices.

Innovations continued. With the first-ever cascaded silicon Raman laser, the team dumped the mirrors in the earlier laser and instead used ring resonators to greatly reduce the lasing threshold and boost output. Concurrently, the team also broke new ground on the modulator front. The team demonstrated that a silicon modulator is capable of transmitting data up to 10 Gbps, and in 2007 they achieved 40 Gbps, matching the best commercial, non-silicon-based modulators of the time. By June of 2008, Intel had announced a single chip capable of 200 Gbps through the integration of eight modulators each operating at 25 Gb/s.

Earlier this month, they unveiled an evolution of the SiGe photodetector: the 340 GHz avalanche photodetector (APD). True to its name, the device receives light and multiplies it—as many as a hundred electrons to every photon—while converting it to electrical signals for distribution. Achieving a frequency of 340 GHz at 1300 nm, it is the highest ever gain-bandwidth measured for an APD device. These speeds are required to realize the commercial terabit optical links that so far Paniccia has achieved in the laboratory (some have as many as 25 modulators and 25 hybrid silicon lasers on a single chip).

The new silicon frontier
At Intel, as Kahn puts it, to stop running is to die. It’s a place where R&D is pursued but managed carefully—the company has no choice but to make more right decisions than wrong ones. In this light then, it’s remarkable and fortunate that one man’s small research project survived technical pitfalls and budget alligators to emerge as one of the precious elements in the crown of Intel’s research portfolio.

“Many thought that building optical devices out of silicon was impossible,” says Paul Otellini, president and CEO of Intel Corp. “However, the results from Mario and the Intel team have proven otherwise. ”

Rattner says Intel is actively commercializing Paniccia's inventions, and the intention is there to place high-speed photonics in billions of connected devices. “This really can be viewed as extending the benefit of Moore’s Law to an entirely new area,” he says.

Gordon Moore says Mario’s contributions are likely to be most important in fields far removed from the areas he was familiar with when he co-founded the company.

“I have been surprised at the results that his group has been able to achieve using silicon as an optical material,” says Moore. “His earlier contributions using light to help test silicon devices were creative and useful, but not surprising. On the other hand, making Raman lasers in silicon was a real surprise to me, and the switching speeds that he has achieved with optical silicon structures are impressive, unexpected (by me and by other experts in the field), and potentially useful.”

As a result of his work, Paniccia, 42, has received two Intel Achievement Awards, has published more than 100 papers, including three book chapters, three Nature papers, three Nature Photonics papers, and has more than 67 patents issued or pending. Paniccia commends Intel for sticking with the research when there were so many opportunities to kill it. He also credits his parents for intilling in him a strong work ethic, as well as the support from his family, including his twin brother, Anthony, his sister, Loretta, and especially his wife, Rachel.

“I have been fortunate in my life and have had many people help and mentor me along the way. My advice has always been to follow your dreams with passion and remember to always give back. Someone helped you along the way; remember to help others.”

If Moore’s Law and Intel’s track record are any indication, says Paniccia, silicon photonics may revolutionize future communications. He hopes to see it in products in the next three to five years.

“We are just at the beginning. The fun is just starting. I can’t imagine what 10 years from now we will be building.”

Published in R & D magazine: Vol. 50, No. 7, December, 2008, p.16-19

Topics

Advertisement

Share This Story

X
You may login with either your assigned username or your e-mail address.
The password field is case sensitive.
Loading