The Optical Society Oral History Project Interview with Emmanuel Desurvire Conducted in 2007, by Gail Overton

Transcription

1 The Optical Society Oral History Project Interview with Emmanuel Desurvire Conducted in 2007, by Gail Overton OVERTON: Hello. I am here to interview our 2007 John Tyndall Award Winner, Emmanuel Desurvire, and Emmanuel was awarded this for his contributions in erbium-doped fiber amplifiers, specifically as they relate to DWDM systems in use today. And I ll start with a few questions for Emmanuel. DESURVIRE: Good morning, Gail. OVERTON: I ll just go into questions. I guess we should start by talking first about erbiumdoped fiber amplifiers. Can you tell us a little bit about the history of optical amplification? DESURVIRE: [00:59] All right. It s, to me, a history that starts almost with the invention of the laser back in the 60s, where Elias Snitzer, who is the inventor of glass lasers, demonstrated the possibility to have stimulated emission by doping glass with rare Earth ions, those elements you find in the next-to-last line in the Mendeleïev table. Neodymium is one of the most famous for industrial applications, and erbium was for a long time, I would say, neglected for the physical reasons of inversion a medium that is difficult to invert unless you cool it down to a low temperature, or you put a lot of power. So let's say it had the reputation not to be very efficient, and the 1.5 microns being the lasing wavelength, they didn't have in the 60s any specific interest for industry. And at Bell Labs in the 70s, there was a regain of activity of doping glass to make fibers, in view of realizing compact sources. But this, again out of luck, maybe went into oblivion because of the emergences of semiconductor laser source at the 800 nanometers, and then 1.3 microns, and then later 1.35 microns. So now we are at the beginning of the erbium dope fiber amplifiers so a field not so new, save for the idea of doping the fiber with erbium, as opposed to neodymium. And to make this fiber lase and then amplify signals with the 1.5 micron wavelength as the objective in order to achieve amplification of optical communication signals, which is an idea connected to, of course, Southampton University. And at Bell Labs, there was an activity of doping fibers, and this regain of activity on doped fibers certainly took my attention, because my background, my PhD work, was on fiber amplifiers, but based on the stimulated Raman scattering, which is an interesting application where there is no dopant in the fiber to amplify light; it s based on scattering by the Raman effect. So it s an undoped fiber amplifier, if you want. And I was naturally very interested in this erbium-doped fiber amplifier, because to me, it made every sense that this had to be what would be called later a killer application, but we were not yet in the 2000s; we were in the mid- 80s. So at Bell Labs, I took over this activity with a fair amount of support, but not more than that meaning that people were busy looking at semiconductor amplifiers, which were very much mature, and naturally, candidates to achieve optical repeater type of system applications. OVERTON: And what year was it that you came to Bell Labs?

2 DESURVIRE: [05:08] June So I immediately started-- that was the time when Southampton was also active with the first results on the-- I think we ll remove that later. The first amplifier result came, actually much later, because as I said, erbium glass is not a very good laser medium, but what changed is that the core of the fiber, being a few microns in diameter, you can actually achieve by sending the pump through the core, you can achieve extremely high pump intensities. By extremely high, I want to be specific. Let's say for one milliwatt of pump, you get a kilowatt per square centimeter. So it s six orders of magnitude of intensity with respect to power. So this conveys a sense of the power confinement in the core of the fiber that changes completely the property, because now with these high intensities, you can achieve medium inversion, which is the condition required to achieve gain, or amplification, of traveling signals by stimulated emission. OVERTON: I think before we discuss in more detail about your role at Bell Labs and beyond, can you tell us a little about your educational background and how you came to be interested in optical fibers all those years ago before they were even a commercial product? DESURVIRE: [07:13] Okay. To make it simple, I did courses in physics, and even to the point of doing theoretical physics nothing even close, really, by far to what we are talking about and feeling that theoretical physics was not necessarily my cup of tea after graduation, and being, I would say, the practical experimental type of person, yet with a taste for theories and modeling and physics, I came to what is now Thales which was Thomson-CSF at the time, and the topic was light amplification in single-mode fibers by stimulated Raman scattering. So this work allowed me to continue some research with an application on Raman scattering in fibers at Stanford University, where I did a post-doc, and there, I put together what was unnoticed, because the claim was not that, a recirculating fiber loop where I would inject a pulse of light and make it recirculate so many times. So I propagated one single pulse over, say, 800 kilometers. Without knowing it, I was achieving the first optically-amplified transmission in a recirculating loop, which has been used by several people after. But the claim of my work was not to investigate telecoms, but to try to achieve optical memories the idea of being able to do, for instance, fiber gyroscopes, and also memories for ultrafast digital signal processing, because light has this problem that it s very hard to store as information, unlike electrons for electronic devices. [09:35] Then, this work was deemed interesting for Bell Labs, who recruited me on campus literally and I had no difficulty to accept what would be considered at the time, and still today as being a phenomenal opportunity with the reputation of Bell Labs. So little did I know that I would stay this long in the United States, coming with a post-doc and staying, actually, for close to ten years altogether. So this is where I came from, and now I hope it s clear that having worked so many years on light amplification, I was immediately turned on with the dope fiber, and practically everything was to be rethought, because erbium-doped glass has nothing even close to Raman amplifyers except for stimulated emission and for the geometrical mode overlap issue. So I had a little bit of an advance on the modeling, except that we had with the erbium amplifier something completely different to analyze, where nothing was really done before. And this is where it started to be truly the exploration of the physics of EDFAs. OVERTON: [11:09] So in 1986, then, when you came onboard, I guess at that point, that s where your real work began with bringing this device from a concept to a functional working product. And I know that you talked quite a bit about the erbium amplifier as it evolved through

3 the years. It wasn t instantly, you know, working from the time of conception to a commercial product. So I guess tell us about the years after 1986 and the steps that you took to bring this to a working product with gain levels that were useful for communication. DESURVIRE: [11:48] Okay, yes, you re absolutely right to emphasize that it was not an overnight success, like ISDN. In hindsight, when you consider the time scale between mid- 80 and mid- 90, that makes it look long, but in the meantime it took only two or three years to prove the entire concept. So we are talking about the fast-paced, accelerated discovery, engineering, and implementation probably one of the fastest in the history of fiber optics, which normally has time scales of ten, 20, or more, considering the time it took to do the first low-loss fibers, and to achieve the first WDM. We re talking about half-century scales. So after I started to investigate, of course, we had to face-- when I say we, I refer to a small team of colleagues. That was the way we worked at Bell Labs at the time. We tried things with benevolent support of the management without having really a product concept in mind. I would say the concept in mind, as I mentioned earlier, is this gut feeling that an optical amplifier has to be a fiber amplifier simply because it s a fiber. And then you splice a fiber to a fiber. It s very simple. You splice it to fuse the glass from the doped to the undoped, and already you have solved the problem of coupling. Also, there s a symmetry. That s where the modeling comes in, in my mind the symmetry of the problem, cylindrical symmetry. It s beautiful. Nature wants that, as opposed to a planar wave guide, which has polarization modes, and which is asymmetrical with different media, with reflection and coating issues. I m not here to do the backward analysis of the drawbacks of the competitor called the semiconductor optical amplifier. But having that in mind, my colleague, Jay Simpson, who was preparing the doped fiber, we, and other people that joined successively in the course of a few years, the first task was to understand exactly what we were looking at, how much erbium had to be put in there. The first samples we over-doped, and when you put too much rare Earth in glass, it has an effect called quenching. So erbium absorbs light, and you pump in the fiber, and what you get is what is called bleaching, this absorption to the point of transparency. So the funny thing is that my director, at the time, was not sure I was doing amplification, but bleaching an absorbing fiber by putting watts of power into it. So it didn t look like I was going anywhere by just changing an absorber into a transparent. [15:17] But as we understood that we had to put the doping level to a reasonably small, meaning a few percent or less, which fiber is long, so you just kept the length to the point where you have enough gain to achieve what the traveling-- the gain coefficient is local, and its integration gives the gain from input to output. So the first experiment was, finally with a fiber that looked like it was producing amplification with the right doping level, to cut it back to see-- as you cut the fiber, the gain increased, because along the fiber, it goes up, and down, because the pump is absorbed, so it s also where you reabsorb the signal. So that was the first discovery at Bell Labs, was we had to optimize the length of the fiber for a given amount of pump power. And then, if you continue to cut, then the gain goes down, because, of course, the medium is inverted, but you don t have enough integration of length to have significant amplification. So we are talking about 20dB, 25dB with several hundreds of milliwatts of green light. I can still see that green today, so much it s a very, very bright green that you work hours with in the darkness. The guys at Southampton had the same thing with a krypton laser, red, and we had an argon laser in green with hundreds of milliwatts to achieve gain. [17:08] So that is to say, this was interesting for experts in telecomm and veterans of Bell Labs and the management, that we were trying to amplify signals with this thing what you call the high-power gas laser. But we were very far from any product concept. So let s call this

4 time an exploratory discovery stage, where it was just investigating the possibility of amplifying signal, and not to be focused on how to invert the medium, but what kind of fluorescent spectrum do we get, concentration level, as I mentioned the length, how the pump is absorbed along the fiber, and how the signal is amplified and then reabsorbed. Then came the second stage in mid- 87 let's say after a year of initial exploration. That was the focus on digital application. We would not just send a signal through the fiber; we would send coded data, still using the argon laser as a pump. And now, to investigate, if, by any chance, those data, while being amplified, could be corrupted by some effect. As we know in semiconductor, the gain dynamics are such that successive bits can talk to each other. That s called patterning effect. That is, one bit is taking the gain, the bit that comes after has less gain, and depending on the number of zeroes that you have between two ones, you could have funny patterning effects well known in semiconductor, which explains the approach of constant-envelope signaling or modulation format, which is known as FSK or PSK, as opposed to ASK. Then, we observed, we did some bit error rates and it was perfect no patterning effect. Some people said, Great, and so what? You know, you observe nothing, so what is the problem? [19:36] But then, something very important happened, which was noticed by many. My colleague, Randy Giles and I took the initiative to put several channels simultaneously in the fiber amplifier, and to do error rates on each of the channels. This was truly the first WDM transmission experiment, if you forgive transmission, because we transmit from this point of the lab to this point of the lab. It was not the matter of transmitting kilometers, but it doesn t matter. The importance is that the output data were not corrupting each other. In a nutshell, we had proved the inexistence of cross-talk as opposed to semiconductor, which was known to be a big problem, that channels that would be as close together as one or two gigahertz spacing would beat and interact with the gain medium in such a way as to exert cross-talk from each other. And we observed some penalty due to the saturation of having several channels. But the penalty didn t show any floor, meaning that there was no cross-talk, per se; it was attributable to an effect of gain saturation. And the nice thing is that in parallel, we brought the evidence of the gain dynamics, the constant of the dynamics, in the erbium amplifier was on the order of ten milliseconds. It s actually pump-power dependent. Let's say it s between one and ten milliseconds. We re talking about six orders of magnitudes slower than in a semiconductor. [21:35] Meaning what? Meaning that as long as the bit rate is greater than 100MHz to 1KHz-- which is not the bit rate, Kbit per second, sorry. As long as you re above 1Kbit per second, there is no patterning. The gain medium is completely constant over time; it s steady state. And that is tremendous. That has a tremendous impact on the quality of the signal, because if you saturate the gain by overloading with many, many channels tens of channels with relatively high power, the sum of which causing a substantial amount of saturation of the gain you still have the perfect amplifier. This is unlike an electronic amplifier. It works under saturation as 100% linear; it s a linear device when it s saturated completely counterintuitive from an engineering standpoint. And that s the beauty of it. It remains linear even under saturation. [22:55] So no cross-dock, slow gain dynamics. The next step was, how many channels can you put? And that means, how do we interpret this gain curve? Okay, we have put several channels a few nanometers apart a complete waste of bandwidth, according to a modern type of interpretation. Now, how do we explain that peak and those wiggles, and that came the physics of it. This gain curve is actually a fluorescence spectrum, which is actually partly overlapped by an absorption spectrum. So the absorption is bleached by the pumping effect, and then the stimulated emission has this spectrum, but it s never 100% achieved because for people

5 who know laser systems, the three-level laser system. And inside the glass there is an electrical field that remains between electric dipoles that causes the atomic levels of the erbium atoms to split by something that s called the Stark effect. It s a random effect, because the dipoles and their strength are randomly oriented. But altogether, these energy levels split, and that creates a broadening of the emission by a sum of as many actually there are eight transition down and seven up, and the sum of this creating the fluorescent spectrum. So it s homogenous, but it s Stark split. And in homogeneity because it s random. Every atom has its own Stark effect, so you combine all these billions of atoms per whatever unit of volume, and then you get a strongly inhomogeneous gain curve, which is very broad, but a little bit uneven. [25:10] So the next step was to understand how we could co-dope the glass, because glass is nice. You can put aluminum; you can put germanium; you can put phosphorus, boron, as you know to achieve index profiles. So it s really within the industrial capability. As it turns out, it was bigger luck that Jay Simpson was pioneering at the time the aluminum doping of fibers, not in view of changing the spectrum of erbium fiber amplifiers, but in view of achieving high-index differences in fibers, because if you do that with germanium, you have to affect exodiffusion of germanium doping into cladding, which creates a gap, and creates the effective step index profile. And second, germanium causes loss. So you cannot achieve high-index difference without a penalty. So he put aluminum, which had a nice quality to remain confined in the core, and we asked him the highest step index difference possible, because we are talking about short length. So the loss of the excessive aluminum was not an issue. We re not talking about making transmission fiber; we re talking about making a small, few-meter length, 40- meter length of fiber with a very high index, one order of magnitude higher than what is used in telecomm fiber. OVERTON: And what year was this that he was? DESURVIRE: That was started even before he arrived. So I got really spoiled. OVERTON: So you were aware of that? DESURVIRE: [27:00] I got spoiled, because we didn t have to start the doping with rare Earth. We had already done that. We didn t have to work on the aluminum co-doping, because it was already in his machine. And that was a lot of luck, and I will explain why: because today, all the erbium amplifiers are aluminum co-doped for several reasons, one being that the gain spectrum is very flat; it doesn t have a dip in the center as observed when you don t have this co-doping. It s more complex today. You can optimize it; you put also filters and equalizers to achieve something within a fraction of a db of difference over the entire available spectrum. [27:56] But the luck came also for the reason that we-- the next step, if you want, was to try to understand now, okay, we have a high index difference. I didn t say why: because we want to confine light. We want to have the pump mode diameter as small as possible of the order of the core size, in such a way that with a few milliwatts of pump, you can achieve very high intensities, as I mentioned kilowatts per square centimeter and a few milliwatts is what you can get from a laser diode. Now, we came from, in the beginning, a few hundred milliwatts of argon pump, like 514 nanometer, which is completely inefficient to pump the erbium glass because of some other effects called the excited state absorption, including also multi-mode propagation, including also loss, background loss.

6 [28:57] So the key was, okay, achieving confinement of the pump, designing the index difference, and the core size two degrees of freedom to achieve maximum overlap between the pump and the signal. And on top of this, choosing between the different energy levels of absorption, which one would be best suited for pumping with the semiconductor. We enter a second phase, which is a pre-qualification, the engineering of the EDFA. We came into some form of competition with Southampton that identified, and justly so, that 980 nanometer was the best pump wavelength, because you could achieve full inversion of the erbium glass, but there was no 980-nanometer pump available. OVERTON: So you were still using the argon laser right along through? DESURVIRE: [30:05] We put the argon laser on the side of the bench still useful for some funny physics experiment that I did at low temperature, but we ll talk about that some other day. That phase was very important. The team lenlarged with John Zyskind, with whom we worked an intense month of investigation. At the rate of one year, it was about 17 papers per year. We had so many things to do, and to publish that we had to take roles of measuring, writing, designing, modeling, and preparing samples, and why was that critical? Because now we were pumping the fiber with 1480 nanometer an idea from Elias Snitzer that came back in 1988 with the idea to pump the laser directly into the absorption band of the gain band, which was like pumping a laser a level two-level system, but because there is a difference between the absorption and emission, you pump in the absorption, and you can amplify in the emission in such a way that you don t fully invert the medium, but you have enough gain to exploit. [31:29] And why 1.48 microns? Because then, you can think of that where the average peak of the absorption or the average of the absorption is centered, because indium gallium arsenide laser could be the ones that are used at source with a little bit of design for a high-power version of it, would be the perfect candidate. And that was exactly what happened nanometer lasers were not available for several years because of the problem known as strainedlayer, lattice mismatch. To achieve this, these diodes would require really designing the epitaxial process to realize these semiconductor diodes. However, with indium gallium arsenide phosphide, I would not say it was straightforward to achieve several milliwatts as opposed to one milliwatt. And we observed, without having the laser diodes, that pumping with a color-center laser, something very big and quite impractical but experimental, we could achieve several dbs per milliwatt, dbs of gain per milliwatt, with a very steep slope. That is, with maybe two or three milliwatts, you were already into 20 or 30 db gain with a fiber length of a few meters, and that was the real proof of existence of the possibility of laser diode pumping very efficiently. Our colleagues from the other side of the ocean at Southampton achieved strictly the same thing with 980 pumping, but the advantage for our approach was we got closer to transfer, because of the laser diode pumping possibility. That was closer to industrial realization. And our colleagues, from the development of Bell Labs, worked right away on this, and this is also where the Japanese team and some laser diode manufacturer in Japan came with the first device, and the first, immediately in [34:11] And that s practically the end of my story as far as the early times are concerned, or the beginning of the big story. 89 was the year of all the record-breaking transmission distances using laser diode pumped EDFAs as inline amplifiers with 50km of transmission fiber in between 1Gb, 10Gb, 100km, 200km. People who worked in coherent transmission could not believe what they were seeing. Now you could transmit signals at gigabit rates over several hundred kilometer distances with laser diode pumped practical, compact, efficient laser diode

7 pump EDFAs. And then the rest is WDM, putting more channels, as I explained no change, no degradation of signals. You need a little bit more power to provide to each channel the required amount of amplification to go to the next step of fiber loss. So for one more channel you add, you have to increase 3dB of pump power, but that s like one milliwatt or two milliwatts versus four milliwatts, eight milliwatts. But the rest is straightforward. OVERTON: So the bulk of your work was between 1986 and 1989, coming up with these primary discoveries of optimum fiber length, optimum fiber design, and doping level. So that all occurred within a three-year period? DESURVIRE: [35:36] That s right. There was another serendipitous thing that was completely unexpected. I mentioned that the fiber amplifier push was nice because you could fuse the fibers, but if you just think of a high-index difference fiber with a transmission fiber, they are completely different. They are different doping profiles, and the erbium one has a different mode size, and of course, if you couple the mode of the doped fiber into the mode of the transmission fiber, there s coupling loss because of mode mismatch. And the nice thing about it is when you fuse with an arc, the dopants diffuse into each other, and if you do it carefully, with time and consideration, and a little bit of learning, you end up having a completely smooth transition of the mode size from one fiber to the other. So the end result being that you have virtually zero coupling loss. If you just fuse the two fibers, they re completely different in design, but fusing the glass together, the right way to do it, not too brutally, would make the most smoothly propagate from one medium to the other and back, because it has two couplings. OVERTON: Now that you determined these optimum fiber lengths, doping levels, mode profiles, pump characteristics, were there any other issues that arose in your research that turned out to be problematic? DESURVIRE: [37:18] Yes. I would say-- I mentioned the key one, that there were not many signs that we were going in the right direction, because the solution was easy to find. It was not instantaneous, not trivial, but within the course of a few months we would make significant progress. I didn t mention the polarization insensitivity, based on the fact that erbium-doped glass is anisotropic, so it doesn t matter if the pump is polarized, if the signal is polarized this way or that way; the gain is the same. Again, this goes in the right direction, like the cylindrical symmetry, and the amorphous OVERTON: This is the Mother Nature part? DESURVIRE: [38:14] Yes, the amorphous characteristic of the gain medium was really calling for an erbium amplifier. So Mother Nature did things extremely well to provide the ion erbium they put in glass is 100% radiative with no loss of the pump photon. There s a little bit of energy differential, but that effect of stimulated emission is 100% efficient. There s no thermal relaxation of the ions. And like other rare Earths that could fluoresce, but never provide efficient gain because of competing relaxation processes. So without going into those details of the physics and the understanding, the other problems we could think of is how do you couple the pump into the doped fiber while you have the transmission fiber for which the signal is coming? How do you achieve that with an optical component that will have low coupling loss? Do you pump forward or backward? Which configuration is better? Since there is a lot of power

8 involved, is this something that could be resistant to defect or to burning effect, or any type of degradation from the component viewpoint? [39:45] Another aspect is, if there is somewhere a reflective or reflection point in the path, since you have a gain medium, then you want to have a laser that is going to lase or do spurious oscillations, and that, you have to eliminate. So there was a need to develop a compact, low-loss, polarization-independent Faraday isolator. That is, light comes through, but doesn t go back through it. It's what is called high return loss. And this way you don't build oscillation. So it s attributed to the optical component industry not only to have been able to develop very rapidly in a matter of months or a few years to come to the full qualification, not only those pump diodes, as I mentioned, but also the so-called pump multiplexer to inject the pump into the doped fiber while the two fibers are spliced, as I explained, and also the isolator. So now, if you have the pump diode, the pump max [?], the isolator, the splice, and the doped fiber, maybe you put a little passive coupler to look at what s coming in and what s coming out and do an electronic gain control loop to ensure automatic gain control. We did that, too, in different patterns and configurations. And altogether, you have, basically, the EDFA as of today. Maybe today, you have gain-equalizing filters, and you have pump redundancy, dual-pump, bidirectional all kinds of designs. There are lots of things that have been thought about, which make a better amplifier with broader, flatter bandwidth and even more efficient. We had basically covered it all in terms of the basics that qualified this device to be passed to, as it turned out, the ssubmarine department in Holmdel [?], where my former colleague Niel Bergano has done these recirculating loops to study the long haul propagation over several thousand kilometers, which would qualify the EDFA for the submarine domain. [42:24] And that went very rapidly, because by 95, these EDFA repeaters were at the bottom of the sea. We re talking about end of the 80s, in five years it went from a research prototype to an industrial-class submarine repeater for the Internet and for traffic. That represents an enormous market for telecom. And then the transition to terrestrial was not as rapid, because terrestrial doesn t work. First, with the long haul, but the importance of EDFAs was rapidly, immediately exploited for terrestrial, and then WDM, and the WDM in submarine, in the back-and-forth complementary evolution more bits per second, more channels, and more traffic, more cables on land and undersea. And that s pretty much where we are now. We have those cables everywhere on the globe, and even coming to the home. OVERTON: How does it feel to have those accomplishments in your history and see how fiber optics has evolved from that point? You know, since you left Bell Labs, I guess, what are the technologies you see coming in the future, and what do you see as maybe a message to other scientists working in this area, what they should be thinking about when they do their research? DESURVIRE: [44:04] Okay, to take your question in that order, it feels very good to have been able to contribute one small but important piece of the early exploration and qualification phase. There are few people who have that privilege to have been able to be at that time, at the right time, to look at this device that was in its infancy that came completely unexpected, that was truly discovered, explored, and then exploited and applied. So it feels good to be one of the contributors. But this being said, there are lots of people that must feel good to have made it happen, because from the exploration in the lab to the submarine repeater and the DWDM EDFA module, there are still lots of engineering issues that lots of people have enormously contributed to.

9 [45:10] Now, in terms of the lesson learned in what could be given as an advice or encouragement to the generation, the young people, or not so young, or both that could come and wonder, is there anything that could be done as good as this EDFA in a lifetime that I could participate to, that I could think about? The answer is definitely yes. Because there s no law in physics that says that something is impossible, unless you try. Of course within an industrial environment, what you try has to reflect a purpose. If we are in telecom, we have to think of a crazy idea that, if it worked, would revolutionize telecom. That s exactly the state of mind that we had back in Bell Labs if this fiber amplifier would work with practical design and effectively, then that makes a big difference. [46:20] So having this objective in mind, the rest is just a matter to try to find the shortest path to reach the subject. So I would recommend that no idea is stupid; no idea is crazy, as long as there is this purpose not science fiction, but a reasonable purpose ahead in mind, because that gives you the strength to overcome the obstacles called indifference, adversity, competition with competing solutions, or mature solutions that are already well-established, or expertise that is, by definition, skeptical of anything that is going a little bit too far or too different from what is well-consolidated. The advice is, any idea that has potential to solve a big problem potential being not proven, but deeply felt if you know technology that has the capability to fit into this purpose, then it is worth to invest and to defend that to the management that is there. Good management is there to support the good ideas. It s contrary to popular belief, and you have to fight for it; and you have to prove it. That s not a picnic; it takes time. And there are lots of smart people that have exactly the same idea at the same time, so it s a matter of to go fast and to get the necessary teamwork and support to complete. As I explained today, no matter how motivated you are, there are things you cannot do. I don't know how to fabricate fibers. Some people don t know how to model. You need to complete experimental skills, engineering skills, and design skills and theoretical analysis, optimization. All these are different facets which you will find in your colleagues. If they are motivated the teamwork goes and the management supports your idea, and you get the funding or the necessary amount of push or encouragement that you need to go ahead. But it doesn t come overnight. It will make you lose sleep. But that s basically what good research is about work on weekends, and sleep short nights sometimes. [49:10] Now, the future, the future is, as I perceive it-- we have gone through phenomenal times over the last 15, 20 years since that story of today took place, where technologies like error correction, DWDM, Raman amplification, again, and receiver design came into the picture. But it has been, I would say, incremental. We haven t seen anything comparable to the EDFA. So the question is, is there anything yet left to invent? As I said yes. There s no law that says everything has been invented. Famous quote from the USPTO: Everything that could have been invented has been invented. Untrue. My feeling is that we are now in pre-edfa times, again, because the fiber bandwidth is going to rapidly saturate. Of course, we are putting more bits per second per Hertz, but that means that we are turning to more sophisticated modulation formats, and this looks very much like the conditions that we had back 20 years ago of big potential for moving very fast and very strongly into a new direction, which was not the case ten years ago or a few years ago even, because it was the post-bubble era. Research was a little bit looking for a purpose. Development was key. Cost reduction was key in terms of making your low-cost devices. Now that the market is back, and the bandwidth exploitation is exploding, there is a call for greater performance, for crazy ideas. That s why I call that the pre-edfa time and not the post-edfa time, meaning that it s something big that s meant to happen. It s not a prediction; it s just a feeling that something big, as big as the EDFA

10 or as big as the first low-loss fiber, is going to come in the picture within the next few years, because the industry needs it. So the young generation will work on that, and then hopefully I'll be watching that from a keyhole, and rejoice for the happy people that will be in the adventure of discovering something that would be of this importance. OVERTON: Good. Well, I d like to thank you very much, Emmanuel. I think that that s a great story, and I really appreciate the message to the researchers of today, that not everything is invented, and there are many more things out there that the next generation can invent for us. So thank you very much. DERSURVIRE: [52:27] Thank you very much. OVERTON: [52:32] Hi, I m Gale Overton with Laser Focus World Magazine, and I m here to interview Emmanuel Desurvire, who is the 2007 John Tyndall Award Winner for his early exploration of the theoretical and physical understanding of erbium-doped fiber amplifiers. So I just thought we would start with a question about the history of optical gain. Can you tell us a little bit about how optical gain was discovered, and how this phenomenon came to be? DERSURVIRE: Well--- OVERTON: So how does it feel to have contributed to the early developments of erbium-doped fiber amplifiers, and what message do you have for researchers today? DERSURVIRE: [53:23] There was-- a major part of it had to do with noise, because stimulated emission that amplified signals also goes with the amplification of spontaneous emission, referred to as amplified spontaneous emission, which is a parasitic random background signal, and we had to go through the understanding of this ASE and see if it was possible, for instance, to cross the Atlantic with maybe 100, 200 amplifiers, and if that noise accumulating through the chain, through the transparency there, would destroy the signal quality, referred to as the signal to noise ratio. So as it turned out, that was another happy discovery, is that the signal to noise ratio degradation due to the accumulated ASE was not that substantial to forbid any transoceanic or extra long distance application. And that was also a key proof. Another proof was using the EDFA as a preamplifier to overcome the thermal noise of a receiver. As it turned out, we broke the record of the photon per bit sensitivity at the time by a fraction of a db. But that was not the purpose; it was just to show an EDFA could be used as a very good-quality, top-quality optical preamplifier. And this is in use today. So EDFAs are not only used as power-boosters and line repeaters, but also optical preamplifiers because of this good noise quality that comes with-- it s a degradation, but it s not an impediment to the application we re talking about. But this was not an overnight realization. It had to be qualified theoretically and experimentally. And we derived a noise figure formula that is used by ITU-T today. So I m happy having contributed to a standard, and a small contribution, but I believe a practical one. [Joking with interviewer] DERSURVIRE: [56:00] Altogether, it was an incredibly intense and fun and exciting period of life. That s why I cherish those years of exploration and friendship and teamwork.