Advertisement

Blog

Nanophotonics: Promise for O/E/O Integration

A common theme here on Planet Analog has been that digital circuits and signals are really analog circuits and signals to which we’ve added an abstraction — a layer of significance or meaning. The digital abstraction has proven itself to be among the most powerful constructs the field of electronics has ever invented — up there, I’d assert, with superposition and communications theory.

Of course, as is true with all such constructs, the digital abstraction holds only so long as the underlying circuit’s parametric (read analog) behavior performs sufficiently to maintain the illusion. Perhaps the clearest demonstration of this fact is the eye diagram (Figure 1 ).

Figure 1

An open eye.(Image courtesy Tektronix)

An open eye.
(Image courtesy Tektronix)

Here, symbol uncertainty or, if you prefer, error-probability rises with a range of parametric challenges including increasing phase and amplitude noise, multipath interference, and limitations on channel bandwidth (Figure 2 ).

Figure 2

A stressed eye.(Image courtesy Tektronix)

A stressed eye.
(Image courtesy Tektronix)

Getting information rapidly on and off chip, be it for communication or data processing applications, is a significant concern because, in a world of deep submicron processes, I/O and associated media often form the throughput bottlenecks. I mention this because an announcement that slipped under my radar (and may have done the same with yours) last December concerns technology advances that can significantly broaden I/O bandwidth.

The technology — called silicon nanophotonics — integrates optical components with electrical circuits on a monolithic device. IBM has succeeded in fabricating silicon nanophotonic devices for the first time on a sub-100nm semiconductor process (Figure 3 ).

Figure 3

This image depicts a device IBM fabricated on a 90-nm process that integrates a photodetector (red feature on left), a modulator (blue feature on right), silicon transistors, and eight metal-interconnect layers.(Source: IBM)

This image depicts a device IBM fabricated on a 90-nm process that integrates a photodetector (red feature on left), a modulator (blue feature on right), silicon transistors, and eight metal-interconnect layers.
(Source: IBM)

The work builds on initial proof-of-concept efforts in 2010, but the more recent effort demonstrates IBM’s ability to transfer silicon nanophotonic technology to a semiconductor process running in a commercial foundry (Figure 4 ).

Figure 4

IBM silicon nanophotonic chip showing blue optical waveguides and copper conductors for high-speed electrical signals.(Source: IBM)

IBM silicon nanophotonic chip showing blue optical waveguides and copper conductors for
high-speed electrical signals.
(Source: IBM)

IBM described the work in a paper presented at the last IEEE IEDM (International Electron Devices Meeting) held December 10 through 12, 2012 (Reference 1 ). Enhancements to a standard 90nm high-performance process allow integration of optical modulators and germanium photodetectors. The resultant process is optimized for analog functionality and supports monolithic fabrication of power-efficient multichannel WDM (wavelength division multiplexed) 25 Gbps transceivers. The 90nm process with 63nm gate lengths produces NMOS devices with Ft greater than 150 GHz with high transconductance and good matching characteristics.

Amongst the passive nanophotonics features IBM integrated without additional mask layers were optical waveguides, waveguide crossings, directional couplers, lateral fiber couplers, wide-temperature range (±30°C) WDM filters, and vertical grating couplers. With minimal add-on mask layers, the technology can implement active nanophotonic functions including modulators, E/O (electrical-to-optical) and O/E (optical to electrical) converters, and waveguide-integrated photodetectors.

References:

  • Assefa, Solomon, et al., A 90nm CMOS integrated nanophotonics technology for 25-Gbps WDM optical communications applications , IEDM, Dec 2012.

Related posts:

33 comments on “Nanophotonics: Promise for O/E/O Integration

  1. goafrit2
    August 14, 2013

    In college, my professor told me that soon we can have waveguides that can move packets of light so that signal transmission be done in the optical domain inside chips. Yes, bring fiber optic cable inside silicon. Till today, I am yet to see that. Nano and its variants are not living to the promise to radical change to the world as they were heralded a decade ago.

  2. eafpres
    August 14, 2013

    As a follower of developments in Silicon Photonics, I saw this work when the early PR came out.  My impression of the “blue cube” figure is that it is more of a marketing conceptul illustration.  Nonetheless, the level of integration they have achieved is promising.

    While there are a host of issues to be fully solved yet for a really good integrated Si Photonic solution with, say, 100 Gb/s per channel, perhaps one of the most important is the photon source.  In most optical communication systems I think they are using some kind of laser, and making lasers in CMOS has been elusive.  This has bifurcated efforts where some are pursuing so-called hybrid systems which involve various forms of bonding somewehre in the process flow, and others that are seeking a true full CMOS process.  It remains open what technology will win out in the long run.  

  3. TheMeasurementBlues
    August 16, 2013

    Joshua, even some of the hard core signal integrity people from DesignCon realize that photonics on boards and on chips is inevitable. Case in point, there are two articles on DesignCon Community (links below) that describe the problem of skew on differential electrical signals caused by a change in dielectric constant in PCB materials as a result of the glass weave. We've down in the micorscopic levels here.

    At 25Gbps, the glass-weave skew is a serious problem that only statisticall comes up, but if it does, BANG you're dead.

    Glass Weave Skew Problems May Be Solved by Eric Bogatin

    Glass Weave Needs a Standard by Lee Ritchey

  4. Steve Taranovich
    August 16, 2013

    I love the idea of nanophotonics in the interface section of the IC. How about using TI's micro-mirror technology to act as a switch for the optical signals? Of course, thermal and cost barriers for nanophotonics need to be overcome, but I am sure that time will bring those issues under control

  5. Netcrawl
    August 17, 2013

    @Steve That's a good idea, using TI's technology enables innovative optical solutions that can disrupt existing end equipment-this one has great potential. TI's DMD( Digital Micromirror Device) is a great stuff, this one can be used for high speed, efficient and reliable spatial light modulation.  

    @Steve with TI working hard, I think they're in a process of solving the problem. TI has a proven semiconductor capabilities, I believe they can crack the barrier.

  6. goafrit2
    August 18, 2013

    >>As a follower of developments in Silicon Photonics, I saw this work when the early PR came out.

    Generally, the whole domain of nano and the rest has been driven by PR. In the venture capital industry, they are exiting the sector because it has underdeliverred despite the promises of innovation capability that is inherent in the technology. I think it will take another 20 years for us to see real impact provided people are still investing in the area.

  7. goafrit2
    August 18, 2013

    >>Case in point, there are two articles on DesignCon Community

    The problem is not if people are writing about this. Rather, it is if we are seeing products that are built on them. 

  8. fasmicro
    August 18, 2013

    There is still promise on what nanophotonics will do in the industry. The major challenge has to do with lack of government spending in driving fundamental research. With all the talks about government exiting the public-sphere, fields like nanotechnology will suffer. Why? You need a lot of money to incubate such ideas. That said – the integration of O/E/D will point always in the direction of nanophotonics. It may not happen immediately but it will surely come to pass. Data will explode and when we move from silicon, we will find a better way to interconnect inside chips.

  9. eafpres
    August 18, 2013

    @goafrit2–“In the venture capital industry, they are exiting the sector because it has underdeliverred despite the promises”

    That may be true, but the US OEMs have not exited and continue to fund at high levels, in addition there is consolidation going on which means some small companies generated value the big ones decided they wanted.  Most of these acquisitions are not at low valuations.

    I also disagree on lack of meeting promises.  The progress is absolutely phenomenal in both board level and chip level optical/electro-optical.

  10. eafpres
    August 18, 2013

    From my reading (and I'll admit I understand only a fraction of what is being published) the optical mirror/splitter/switching devices on Si Photonics and PICs are below the level of MEMs and micro-mirrors in terms of the size domain and probably the time domain.  You have microring oscillators that are being used in optical switching on Silicon that are in the 10s of microns in size:

    Image from High-Q micro-ring resonators and grating couplers for silicon-on-insulator integrated photonic circuits; Tong Xiaogang(仝晓刚), Liu Jun(刘俊)Ž, and Xue Chenyang(薛晨阳)Ž.  Key Laboratory of Instrumentation Science and Dynamic Measurement, North University of China, Taiyuan 030051, China

  11. Netcrawl
    August 19, 2013

    Its not just the funding but also some good competition abroad, I talking about the rise of asian players- China and India. Globalization is making a huge impact in the industry, we lost our focus because most our companies are taking a massive shift, they're moving abroad, they doing some cost cutting measures and decided that outsourcing could be a good idea and the best way to effectively compete in the global market. 

  12. Netcrawl
    August 19, 2013

    We're not looking good right now, I think we're losing our competitiveness in the market, we 're facing emerging threats- lots of them. Yes we still have our technological superiority, our education system still making good engineers, we still turning out good research works but the real issue is on the big market. The main focus is on emerging markets like Brazil, China and India, not on US.     

  13. SunitaT
    August 20, 2013

    The main benefit of a photonic connect is improved on- and off-chip bandwidth at reasonable power levels, which could not be accomplished by an all-electrical interconnect. This allows the application of a many-core processor for which the byte-per-FLOP ratio (i.e., the ratio of the communiqué bandwidth to the compute bandwidth) is 1 both for the on-chip worldwide interconnect and the off-chip memory links. This in turn allows shared-memory parallel architectures that are easier to program. In adding, having photonics on the chip enables new applications of broadcast busses and arbitration protocols that further increase performance by reducing latency.

  14. fasmicro
    August 20, 2013

    >> Most of these acquisitions are not at low valuations.

    The problem is that people are looking at nano investment from the lens of KPCB Silicon Valley VC which has disastrous track in that space and have not been kind to the sector. Facks be facts, I do not think the nano space has done well. We see these billion dollar valuations for websites and apps, I am yet to see anything close in the nano space. Maybe you can point me to one.

  15. fasmicro
    August 20, 2013

    >>Its not just the funding but also some good competition abroad, I talking about the rise of asian players- China and India. Globalization is making a huge impact in the industry, 

    I am not sure China and India are doing anything huge in nanophotonics. I think this is still the industry of the advanced nations. Maybe with time, we can see more penetration and diffusion in these new markets but not now. 

  16. fasmicro
    August 20, 2013

    @Sunita, that is the main theoritical  benefits and there are many side ones in addition. If we get on-chip communication to move in the domain of optics, it will revolutionize the industry. Everyone understands that we need a faster way to communicate and move data even in chips. Wall Streets will order the first processors built with it because it will help them take some milli-secs out of time to make money.

  17. eafpres
    August 20, 2013

    @fasmicro–I have followed Silicon Photonics for a couple of years closely and did some historical research as well.  What I'm seeing is that many Chinese universities are working a lot in these areas.  However, there is not as good a track record of moving innovation from Universities to Industry in China as in the West.  But I do think they will be a player in the eventual market, whatever that looks like.

    I do not see a lot of contirubtions from India in Si Photonics so far.

  18. eafpres
    August 20, 2013

    @fasmicro–Another factor to keep in mind is that if more of the optical components can be made in CMOS, the cost comes down.  Also, at the chip level I think it is as much about power consumption and heat as performance, per se.  In other words on chip you can do 100 GHz or 400 GHz all electrical but the more you can convert to photonic, the lower the power.  It is clear that at the 100G to 400G ranges everything of any distance will be optical.

  19. Brad_Albing
    August 26, 2013

    @goafrit >>Till today, I am yet to see that . Yes – and even as of today, the technology is not widespread, just experimental. So we will need to wait a little longer to see see this become a commercial product and become low-cost.

  20. Brad_Albing
    August 26, 2013

    @Netcrawl – the smart money is on India to make the big strides in this sort of research. Do keep in mind tho' that as India gets better at this sort of research (and all research in general), costs, expenses, and wages will rise sinificantly. So their cost savings advantage will fade away.

  21. Brad_Albing
    August 26, 2013

    @eafpres – the important question then is can the optical components be made in CMOS? Thoughts?

  22. fasmicro
    September 8, 2013

    @eafpres, I agree that China may not have the records of moving university inventions to the market. Nevertheless, China has been good is cloning any great idea from the West. They just have to figure a way to do anything you do cheaply and they will be fine.

  23. fasmicro
    September 8, 2013

    Also, at the chip level I think it is as much about power consumption and heat as performance, per se. 

    Speed will be a big factor when we see chips operating in the range of 500GHz. In that case, how fast you move the electrons or photons will be the game changer. The real question is if Silicon can cheaply support that material construct in the industry.

  24. fasmicro
    September 8, 2013

    >> So we will need to wait a little longer to see see this become a commercial product and become low-cost.

    Any prediction on how long we need to wait? And what is really the main technical challenge preveing us from getting there in time. In college, I did a lab work on this and thought it was primetime but when I got to the industry, I noted that it was far more challenging than my professor made it seemed.

  25. fasmicro
    September 8, 2013

    >> Do keep in mind tho' that as India gets better at this sort of research (and all  research in general), costs, expenses, andwages  will rise sinificantly. So their cost savings advantage will fade away.

    That is true except that it will take a while for India to have a high level of saturation. With its population and fairly decent few universities, we see a paradigm where a decent regular organic supply can suppress that wage increase at least for a decade. The supply side is healthy and wage in India may not rise very fast.

  26. fasmicro
    September 8, 2013

    >> the important question then is can the optical components be made in CMOS?

    Yes – you can made waveguides with Silicon by using SU-8 to create the patterns. However, whether that is economical for a product is what I do not know. Theoritical, that is possible.

  27. Scott Elder
    September 8, 2013

    @fasmicro

    “We see these billion dollar valuations for websites and apps”

    This is a 21st century problem whose magnitude hasn't yet been fully realized.  The most expensive sectors for innovation, in both money and time, are the least valued and out of favor for investment.  It will be interesting to see the price paid for innovating primarily in software.

  28. fasmicro
    September 12, 2013

    >> It will be interesting to see the price paid for innovating primarily in software.

    Software is easily scaled and that is the essence of investment. You can get your money quick and fast. Business is about making money and web does that easily.

  29. etnapowers
    September 12, 2013

    Microelectronics industry is working on the process to produce cheap silicon waveguides, at this time the reliability is an issue to solve to ensure the success of this technology

  30. etnapowers
    September 12, 2013

    @Brad: the research on silicon photonics is a project funded by an European joint venture of big microelectronics firms and the same happened in the US silicon valley.

  31. fasmicro
    September 18, 2013

    >> The main focus is on emerging markets like Brazil, China and India, not on US.     

    That seems like a good plan. You cannot go wrong with those three. U.S. is still dominant but you need to be really good to win markets in this economy.

  32. fasmicro
    September 18, 2013

    >> Microelectronics industry is working on the process to produce cheap silicon waveguides,

    Many prototypes are in academic papers – the problems remain the lack of mass production scaling at high quality and reliability. Optical waveguides are usually done with photoresists but they seem sub-optimal when benchmarked with copper in intra-chip interconnection.

  33. Brad_Albing
    September 25, 2013

    @fasmicro – not my area of expertise, so I won't hazard a guess.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.