Comms chips

Out of the Lab: Erbium-Doped Chips?

What appears to be a significant breakthrough in the integration of optics and electronics on silicon was announced yesterday by STMicroelectronics NV (NYSE: STM) (see STM Sets Silicon Emission Record). Putting optics and electronics on the same chip could slash costs, while also boosting the performance of all kinds of semiconductor devices.

Thus far, the showstopper in silicon integration has been the fact that bulk silicon doesn't emit light unless it is supercooled to temperatures close to absolute zero, which is hardly practicable.

Now scientists at ST's research center in Catania, Sicily, claim to have overcome this problem. In fact, they are claiming that the silicon-based devices they've made have an efficiency that matches that of gallium arsenide and indium phosphide -- the most popular materials for making lasers and LEDs (light-emitting diodes).

To put a figure on this, ST scientists are claiming an external quantum efficiency of 10 to 20 percent. That means for every electron that goes into the system, 10 to 20 percent are converted into useful photons. A range is given because the efficiency largely depends on the packaging of the device (see Scientists Try a Solar Sell).

Previous "breakthroughs" in light-emitting silicon only achieved efficiencies an order of magnitude lower than this. Last year, for example, it was widely reported that researchers at the University of Surrey in the U.K. had created a silicon LED with an external quantum efficiency of 0.2 percent (see Out of the Lab: Light From Silicon).

However, ST may be stretching the truth a little by claiming to get light out of silicon itself. What it's actually done is produce a layer of silicon-rich oxide (SRO), which it has doped with erbium ions. The result is a material similar to erbium-doped fiber, which is compatible with standard silicon processing techniques. Light is really produced by erbium ions, not silicon.

Salvo Coffa, manager of the research team responsible for the work, says that light is produced from silicon in the sense that the electrical current needed to power the device is applied to silicon. Inside the device, the electrons travel to the erbium ions, which then give up energy as photons.

It's worth pointing out that other companies -- notably those working on erbium-doped waveguide amplifiers (EDWAs) -- have also come up with ways of producing erbium-doped layers of glass on a silicon substrate. But there's one important difference: These other companies can only get electrons into the erbium by pumping it with light from another laser. ST's main breakthrough was finding a way of injecting electrons into the material simply by applying an electric current.

Glass is normally an electrical insulator -- hence the need for optical pumping. However, silicon-rich oxide does conduct electricity. As the name suggests, SRO is basically glass that contains more silicon than is expected from the chemical formula SiO2. It consists of small agglomerates of silicon inside a matrix of glass, which conducts electricity by a process called Fowler-Nordheim tunneling, an explanation of which is a bit beyond the scope of this article (and the author -- so go look it up!).

The most interesting thing about all this is that ST's work could be close to commercialization. In the next couple of months, it expects to ready engineering samples of a device containing two separate electrical circuits that communicate using light. The device would be used in applications where the two circuits must be electrically isolated to prevent crosstalk, such as power control circuits. The chip integrates an LED, a waveguide, and a detector, as well as electronics.

The devices are being produced on the same pilot line that ST uses to develop new MOSFET (metal-oxide semiconductor field-effect transistor) and bipolar integrated circuits, the company says.

Beyond this initial product, key application areas are likely to be high-speed digital circuits and optoelectronics for telecommunications. In its press release, STMicroelectronics says it is investigating advanced CMOS circuits where clock signals are distributed through the chip at the speed of light, as well as low-cost integrated devices for DWDM transmission. Another intriguing possibility is the development of new types of optical amplifier that are electrically, as opposed to optically, pumped.

Coffa declined to give details of future applications, saying the industry would have to wait for product announcements. His company has applied for patents on the technology and plans to present technical papers in due course.

— Pauline Rigby, Senior Editor, Light Reading
isitso? 12/4/2012 | 9:23:34 PM
re: Out of the Lab: Erbium-Doped Chips? That is completely different story - of course you can amplify very fast signals in EDFA but the
erbium fiber lasers can only be turned on and off at very slow speed. In optical amplifier you do not transfer the information from the electrical signal to the optical signal hence the lifetime is irrelevant. You do not need to change teh gain of the amplifier to modulate the signal - it already comes in modulated! If you are trying to modulate the laser or LED you transfer the info from electrical to optical form - you modualte the gain of the laser medium meaning the number of excited atoms - the lifetime comes into play.
Think about it: it is precisely because the gain in Er-medium cannot be modulated the amplifier cross-talk is very low. It is just the opposite in semiconductor lasers - they can be modulated very fast but the crosstalk in SOA is huge.All because the lifetiem is less than 1 ns.
So, teh slow medium makes great linear amplifier but lousy source - teh fast medium is just the opposite. Therefore one makes amplifiers with Er and sources with semiconductors.
PS - in Er laser syou can get very short pulses using mode locking but once again it does not mean that you can send information at high speed.

Touch 12/4/2012 | 9:23:35 PM
re: Out of the Lab: Erbium-Doped Chips? Not exactly. The metastable lifetime of the 4I13/2 level of the Er3+ ion is 10 ms. This does not limit the ability of the medium to 100 Hz.

Erbium-doped fiber amplifiers routinely work with 10 Gb/ signals.
isitso? 12/4/2012 | 9:25:41 PM
re: Out of the Lab: Erbium-Doped Chips? To the best of my knowledge, the radiative lifetime in Er ion is 10ms - thus any LED
based on Er ion is bandwidth-limited to 100Hz - to go faster would require an external modulator - which would negate all the advantages
of the on-Si integration.
The laser would be able to operate at somewhat higher speeds all the way to 1MHz or so, which is also not something to be proud of. But it is a very long way from electroluminescence to lasing
because the gain is very low - in my estimate the cavity length of the laser should be about 1cm at the very least (see the waveguide amplifiers available today) - that would render the device useless for the on-chip interconnects, and, once agin would require an external modulator to operate at the speeds higher than 1MHz.
So, while from the pure science point of view the achievement is indeed admirable, from the practical point of view it amounts to little more than growing a lightbulb on a Si chip.
dwdm2 12/4/2012 | 9:26:16 PM
re: Out of the Lab: Erbium-Doped Chips? This was an interesting press release, thanks Pauline. I am interested to know more about the topic. Does anyone know of any references?

A fundamental problem in optical signal amplification via electron interaction is that the later is an activated process. By virtue of electronic interaction, it will "generate" heat and noise. In addition to low efficiency, one will have to deal with "on-chip" noise generated by the same process amplifying the signal. This different than EDFA or EDWA, where the noise associated with input signal is amplified, however, no additional noise is generated by optical pumping.

It would be quite interesting know about any further development on this issue.

Touch 12/4/2012 | 9:26:47 PM
re: Out of the Lab: Erbium-Doped Chips? Enough already! Since when does a press release constitute a serious scientific announcement? Am I the only one that remembers that Fleischmann and Pons announced their "discovery" of cold fusion by a press conference?

The usual tradition of announcing major scientific discoveries is to submit an article to a scientific magazine or journal (Nature, Science etc...). If the reviewers (selected by the editor)give a favorable review, then the article is selected for publication. The press conference then occurs when the article is published.

Excuse me if I don't get hot and bothered about this until the authors publish their work in a respected, peer-reviewed journal (no... SPIE doesn't count).

And... yes I have worked in the field of rare-earth spectroscopy and erbium-doped semiconductors.

beachboy 12/4/2012 | 9:26:55 PM
re: Out of the Lab: Erbium-Doped Chips? Knew it was only a matter of time before the big innovative semiconductor companies would produce something potentially significant. Intel have been using the BUY strategy, IBM tried and failed with Kymata. ST have been beavering away in the background and voila ! IC packaging is still a stroll in the park compared to optics. Now if ST or someone else could work miracles there.
desikar 12/4/2012 | 9:27:00 PM
re: Out of the Lab: Erbium-Doped Chips? Still remember the hoopla around porous Si in the '93-'94 timeframe,and the yellow / green luminescence from a Si wafer (with lights turned off in the lab). Unfortunately, that was an untamable surface effect, or so it seemed.

This one, though, does sound like a major breakthrough. For one, it should be much easier to set up waveguiding and other effective ways of herding photons and amplifying them in selected areas through current injection. Allows for some "regular" processing,albeit with materials related challenges to overcome. This also opens up the possibility of more integration on Si.

It will be interesting to know what current densities will be afforded by Fowler-Nordheim injection, and what the reliability issues may be.

Wonder if there are other dopants they could use to access non-Er wavelengths. If this technology could get down to the 600 - 950 nm wavelength range, it could be very useful in non-telecom applications as well.

Good to hear of something with promise. Thanks
for the article!

FastSwitching 12/4/2012 | 9:27:01 PM
re: Out of the Lab: Erbium-Doped Chips? I ripped the following out of yesterday's OND foreward.

"These are all companies that over a couple of decades, as a minimum, not only systematically came out with successively great technical innovations but also shone as something special in their management and development of people, their taste in buildings, their sense of social responsibility etc. Even in these exceptional cases much of the company's extraordinariness could perhaps be put down to early idiosyncratic (and maybe even plainly lucky) decisions by the company's founders, whose technical and commercial impact lasted for a generation or more, and because the company was thus able to dominate its sector, managerial skill sometimes seemed more extraordinary than it really was.

STMicroelectronics is a company not often quoted in the same context as the above, except perhaps in Europe, and particularly within the hermetic societies of France and Italy from which this particular company sprung, and whose politics and industrial management are somewhat opaque to the rest of the developed world by virtue of their antiquity. Against the company has been the fact that it was not American and also perhaps, more fundamentally, that for much of its life it has been partially government owned. Nonetheless the company today is a tribute to its manager of almost a quarter of a century, Pasquale Pistorio, a native Sicilian/Italian who initially rocketed up the sales hierarchy in Motorola from simple salesman in 1967 to General Manager of Motorola's International Semiconductor Division by 1978, and then in 1980 was persuaded to return to Italy to take over SGS ? a small technically feisty but erratic Italian semiconductor company with a number of Fairchild licenses.

Pistorio had an immediate impact, roped himself to the mast and clearly in his mind's eye, and sometimes in those of the public, made clear he was aiming for the top of the industry. SGS progressed steadily but at a rate that would have taken a century to get where Pistorio wanted to be.

In May 1987 his dream took a massive lurch forward via a merger with Thomson Semiconducteurs, another company with lots of interesting bits of technology but not enough coherent direction. Over the intervening 15 years Pistorio has stayed at his post and, bit by bit, the company, with its initially scrappy collection of products, technology and assets, has inched itself forward from somewhere that was well out of sight, and even more out of mind, in the international listings to just below the top 10, and then into the bottom half of the top 10, and last year, according to industry analysts, lurched into No 3 position worldwide behind Intel and Toshiba.

Remarkably this has been achieved without much involvement in bulk memories; indeed avoiding this high volume, low margin business may be partly what has saved STMicroelectronics - as it was rechristened in 1998.

What has really propelled the company forward has been Pistorio's own personal charisma and energy, and the company's systematically winning choices of participating in high complexity, volume dedicated, chip markets.

In terms of classifying STMicroelectronics as an extraordinary company, in the context of those mentioned above, might be tough, but there is little doubt that Pistorio qualifies on his own as an extraordinary manager. It seems unlikely that at 65 he has enough years left to oust Intel from the top spot, so his initial pledge, to make the company he joined in 1980 a worldbeater, might just not be fully met. But it has been a pretty good try in a market where massively financed and seasoned competitors and violent industry cycles have been around forever."
Sign In