A brief history of light sources 1 Introduction and practical arrangements This lab session will provide an overview and timeline of the light sources used throughout history, mainly focusing on lasers. The sun is the first light source in history and is still our most important source of energy. Sunlight warms us, allows plants to manufacture oxygen and our food from carbon dioxide and water, and it allows us to find our way around during daytime! Of course, mankind has found other sources of light over the course of history. Fire is obviously the earliest of these: from the camp fires of our cave-dwelling ancestors to the spirit lamps still used where there is no electricity. But electricity is the source of artificial light today, starting with the invention of the incandescent light by Joseph Swan and Thomas Edison and progressing via fluorescent lighting to modern light emitting diode (LED) lights and lasers. Daily activities, from scanning items at a grocery store to playing your favorite CD, require the very precise light lasers provide. Some lasers have a well-directed, very bright beam with a very specific color; others emphasize different properties, such as extremely short pulses. The key feature is that the amplification makes light that is very well defined and reproducible, unlike ordinary light sources such as the sun or a lamp. In this exercise, you will learn about the most common laser types and typical characteristics of these sources will be demonstrated. The lab consists of three parts, each ± 15 minutes: 1. Sunlight, incandescent light,... 2. Gas lasers and solid state lasers 3. Semiconductor lasers Attendance at the lab is mandatory. The laboratory will take place on Friday 5 May 2017 from 2:30 PM to 5:15 PM. Please have a look at the file LaserLab Timeslot.pdf on Minerva to see at which timeslot you are expected to come. We will pick you up at the elevator on the fourth floor of the igent tower, Technologiepark-Zwijnaarde 15. More details about how to reach us can be found on http://photonics.intec.ugent.be/contact/contact.htm It is assumed that the students have read the chapter 13 and 14 on Laser and Optoelectronic Components of the Photonics courses note before the lab session. A report is not expected. 2 The history of lightsources In this part of the lab we will take a closer look at some of the most important light sources throughout human history. We will do this by looking at the spectrum of these different light sources. The spectrum will tell us something about the mechanism responsible for the light emission. Identifying these mechanisms will allow us to understand why the colour of a candle flame differs from that of an LED. But also allows us to understand why LEDs are more energy efficient than classical light bulbs. 1
2.1 Sunlight and incandescent light bulbs The sun has been the first light source used in history and is probably the most important. Its spectrum can be identified as black body radiation of an object at 5800K. This spectrum contains all wavelengths in the visible part of the electromagnetic spectrum. When diffracted by a prism we will see a broad continuous spectrum. Figure 2.1: Plot of the solar spectrum. Sunlight contains not only visible light but also a large amount of infrared radiation and some UV radiation. This infrared and UV radiation cannot be detected by the human eye, therefore it is not visible in the spectrum created by a prism. Classical tungsten lightbulbs are thermal or incandescent light sources. The spectrum of such an incandescent light bulb is very similar to that of sunlight. For example in a classical light bulb, a tungsten filament in a glass bulb filled with a inert gas is heated. When this filament reaches a temperature close to that of the sun surface it gives a bright white light. This is also the reason why a light bulb has an afterglow, as the filament cools down the black body radiation is centered on lower and lower wavelengths. This also explains why a light bulb is not energy efficient, as a large part of the radiated electromagnetic radiation is not visible to the human eye. 2.2 Gas-discharge lamps A very different kind of light source is the gas-discharge lamp. Well known examples of these are neon lights and sodium-vapor lamps. Sodium-vapor lamps can be recognized by their distinctive orange color and were used to illuminate the Belgian high ways. Their spectrum differs quite a bit from that of incandescent lamps and sun light, they only contain portions of the visible spectrum. A gas-discharge lamp consists of a glass tube in which a plasma is created. This ionized gas emits photons with energies corresponding to a set of discrete electronic transitions, 2
Figure 2.2: Picture of a gas-discharge lamp. known as emission lines. Each gas has its own distinctive emission lines, which can also be used to identify the gas. Fluorescent lights, commonly referred to in Flanders as TL lampen, are a type of gas-discharge tubes in which the glass wall is coated with a phosphorous material. The gas that is most often used is mercury vapor which has strong emission lines in the UV. This UV light excites the phosphorous material which then emits light in the visible part of the spectrum. 2.3 LED lights One of the most recent light sources are LEDs or light emitting diodes. These are solid state devices in which a layer of acceptor doped semiconductor is put into contact with a donor doped semiconductor layer. The contact between these two layers creates a so called pn-junction. In semiconductors with a direct bandgap the recombination of electrons and holes causes the emission of photons. As recombination happens mostly between holes at the edge of the valence band and electrons at the edge of the conduction band, the emitted photons have an energy corresponding to that of the bandgap of the material. This implies that a LED has a specific color and does not emit in the entire visible spectrum. However for modern lighting applications blue LEDs are used in combination with phosphorous materials to create a more natural light. Figure 2.3: Plot showing the spectra of three different colored LEDs. 3
3 The invention of the laser In 1917, Einstein proposed the process that makes lasers possible, called stimulated emission. He theorized that, besides absorbing and emitting light spontaneously, electrons could be stimulated to emit light of a particular wavelength. But it would take nearly 40 years before scientists would be able to amplify those emissions, proving Einstein correct and putting lasers on the path to becoming the powerful and ubiquitous tools they are today. Not many people know, but the first realization of Einstein s theory was performed with microwaves. Working with Herbert J. Zeiger and graduate student James P. Gordon, Charles H. Townes demonstrated the first maser (microwave amplification by stimulated emission of radiation) at Columbia University in 1954. The ammonia maser obtains the first amplification and generation of electromagnetic waves by stimulated emission. The maser radiates at a wavelength of a little more than 1 cm and generates approximately 10 nw of power. Townes received the Nobel Prize for his pioneering work in the field of masers and lasers in 1964. The grandfather of the laser recently passed away on January 27, 2015 at the age of 99. In May 16 1960 Theodore H. Maiman, a physicist at Hughes Research Laboratories in Malibu, California, constructs the first laser using a cylinder of synthetic ruby measuring 1 cm in diameter and 2 cm long, with the ends silver-coated to make them reflective and able to serve as a Fabry- Perot resonator. Maiman uses photographic flashlamps as the laser s pump source. From this point onwards there is a boom in the development of different laser types. Historically, we start our journey with the protypical gas laser. Then we move on to the solidstate lasers. Even though many of these lasers have been around for decades and many new types have been developed recently, they still deserve a place today in many applications. The main protagonists will be outlined below and demonstrated in the lab. Non-linear effects in optics is a very exciting topic and this lab would not be complete without a small demonstration. The working principle of a green laser pointer will be explained; a simple but yet interesting non-linear optical device. This will give a feeling on how one can still make light sources in wavelength ranges otherwise unattainable with the means described so far. 3.1 Helium-Neon laser In December 1960 Ali Javan, William Bennett Jr. and Donald Herriott of Bell Labs develop the helium-neon (HeNe) laser, the first to generate a continuous beam of light at 1.15 µm. However, people were much more interested in the operation at a visible wavelength. Other Neon transitions were identified and 18 months later the best-known and most widely used laser operating at 632.8 nm was built for the first time. Even now this relatively simple laser is being used for industry and research applications due to its low cost, ease of operation, high spatial and temporal coherence and output beam quality. You will typically encounter this laser being used for alignment. Gas lasers are useful and fun but the main drawback is their lack of wavelength tuning. Therefore, dye lasers were typically used which have very broadband gain curves covering the full visible spectrum up to the near- Figure 3.1: Energy diagram of the He- Ne system. 4
These drawbacks led to the develop- infrared. However, they are inefficient and the gain medium chemically degrades over time rather quickly. ment of the next famous laser, the Ti:Sapphire laser. 3.2 Ti:Sapphire laser In 1982 Peter F. Moulton of MIT s Lincoln Laboratory develops the T i 3+ doped sapphire (monocrystalline Al 2 O 3 ) laser, used to generate short pulses in the picosecond and femtosecond ranges. The Ti:sapphire laser replaces the dye laser for tunable and ultrafast laser applications in the tuning range 650-1100 nm. The Ti:Sapphire laser is most widely used for pumping other lasers because it is quite convenient to choose the right pump wavelength. Other notable applications include IR or Raman spectroscopy applications in the visible and near-infrared range. An interesting household application of solid-state lasers appeared in the form of laser pointers which we will briefly explain in the next section. The green laser pointer, a seemingly simple device is a result of many years of research activities in the laser domain. 3.3 Nd:YAG laser Neodymium (Nd) is a chemical element belonging to the group of rare earth metals. In laser technology, it is widely used in the form of the trivalent ion Nd 3+ as the laser-active dopant of gain media based on various host materials, including both crystals and glasses. The most common neodymium-doped gain media is Nd:YAG = Nd:Y 3 Al 5 O 12 (yttrium aluminum garnet. The output wavelength is typically 1064 nm, but is also usable at 946 nm and 1319 nm (and a few other lines). One of the most known applications is the green laser pointer. These pointers appeared on the market around 2000, and are the most common type of DPSSFD (diode pumped solid state frequency-doubled) lasers. They are more complicated than standard red laser pointers, because green laser diodes are not commonly available. The green light is generated in an indirect process, beginning with a high-power (typically 100 to 300 mw) infrared AlGaAs laser diode which is a solidstate laser operating at 808 nm. The 808 nm light pumps a crystal of neodymium-doped yttrium orthovanadate (Nd:YVO 4 ) (or Nd:YAG or less common Nd:YLF), which lases deeper in the infrared at 1064 nm. This lasing action is due to an electronic transition in the Figure 3.2: Schematic of a green laser pointer. fluorescent neodymium ion, Nd +3, which is present in all of these crystals. The Nd-doped crystal is coated on the diode side with a dielectric mirror that reflects at 808 nm and transmits at 1064 nm. The crystal is mounted on a copper block, acting as a heat sink; its 1064 nm output is fed into a crystal of potassium titanyl phosphate (KTP). This unit acts as a frequency doubler because of a nonlinear interaction of the electric field with the material. 5
Simply said, two 1064 nm photons combine their energy to produce one photon at 532 nm. The resonant cavity is terminated by a dielectric mirror that reflects at 1064 nm and transmits at 532 nm. An infrared filter behind the mirror removes IR radiation from the output beam. This last element is omitted in the cheap pointer-style green lasers such that the output beam has a large infrared component. This makes this laser pointers even more dangerous as compared to the red ones. 3.4 Fiber lasers Figure 3.3: Schematic setup of a simple erbiumdoped fiber amplifier. With the booming of the telecom market, fiber lasers and amplifiers have gained a lot of attention. In most cases, the gain medium is a fiber doped with rare earth ions such as erbium (Er 3+ ), neodymium (Nd 3+ ), ytterbium (Yb 3+ ), thulium (Tm 3+ ), or praseodymium (Pr 3+ ), and one or several fiber-coupled laser diodes are used for pumping. Although the gain media fiber lasers are similar to those of solid-state bulk lasers such as the Ti:Sapphire laser, the waveguiding effect and the small effective mode area usually lead to substantially different properties of the lasers. Closely related to fiber lasers is the Erbium Doped Fiber Amplifier or EDFA. They are by far the most important fiber amplifiers in the context of long-range optical fiber communications; they can efficiently amplify light in the 1.5-µm wavelength region, where telecom fibers have their loss minimum. The setup is the same as for Er-fiber laser but then without the resonator mirrors. 4 Semiconductor laser diodes Coherent light emission from a gallium arsenide (GaAs) semiconductor diode (the first laser diode) was demonstrated in 1962 by two US groups led by Robert N. Hall and by Marshall Nathan, although their lasers would only work properly at very low temperatures (around - 200 C). The dominant challenge for the remainder of the 1960s was to demonstrate continuouswave lasing at room temperature from a diode laser. This was achieved for the first time in 1970 by Alferov and his team. For their accomplishment and that of their co-workers, Alferov and Kroemer shared the 2000 Nobel Prize in Physics. Ever since, laser diodes became numerically the most common laser type, with 2004 sales of approximately 733 million units, as compared to 131.000 of other types of lasers. One major difference between diode lasers and other lasers is their size: whereas gas, solid-state and fiber lasers are typically tens of centimeters in length, diode laser chips are generally about the size of a grain of salt, although the mounting and packaging hardware increases the useful component size to the order of a cubic centimeter or so. The diode lasers are mass-produced using wafer scale semiconductor processes, which makes them really inexpensive compared to all other types of lasers. The semiconductor origins of diode lasers allows for semiconductor integration techniques to be applied, and for multiple building blocks to be defined along the common waveguide, yielding functionally complex devices and opening a new field of photonic integrated circuits. They are used in a wide variety of applications ranging from the readout sources in DVD and Bluray disk players, laser printers, mice and pointers, to the complex multi-wavelength transmitters 6
and receivers in optical fiber communication systems that carry hundreds of gigabits per second of information. 4.1 Characterization of a laser diode Characterization of a laser diode starts by measuring the laser light intensity versus drive current (PI) characteristics. Consequently, important parameters such as threshold current (in ma) and quantum efficiency (in percent) are determined. A good laser diode is characterized by a low threshold current and a high quantum efficiency. The threshold current is defined as the value of the laser injection current at which the laser starts lasing. The quantum efficiency of the laser is defined as the ratio of the extracted number of photons in the cavity and the number of injected electrons which is determined from the PI curve. Another important characteristic to mention is the optical spectrum which shows the optical power of the laser as a function of wavelength. Simple Fabry-Perot laser diodes have a comb-like spectrum with a number of longitudinal laser modes. For telecom applications, it is desirable that the signal is single mode, i.e. the optical spectrum has only one strong peak at a desired wavelength. Temperature has a big influence on the characteristics of a laser diode. That is why most lasers are assembled with a Peltier cooling unit. Without this cooling unit it is impossible to operate the laser in different environments and temperatures with the same performance. When the temperature increases, this will also increase the threshold current. Quantum efficiency will decrease and the wavelength of the emitted light will shift. During the lab three different laser diodes will be characterized using an optical power meter and an optical spectrum analyzer. 4.2 A commercial DFB laser Firstly, we look at a commercial 1.55 µm DFB laser made for telecommunication applications, as is shown in figure 4.1. The laser is mounted on a Telecom laser diode mount so we can easily control the temperature and injection current of the laser. 4.3 Lasers on a silicon platform By the end of 2002, about five Exabytes of data had been produced by mankind. In 2003 humanity jointly generated the same five Exabytes of data again, but Figure 4.1: A 1.55 µm laser module within a single year and by 2012 already 2500 Exabytes packaged in a 14-pin butterfly package. of data were produced per year. As a result, the amount of data handled in data centers is growing faster than Moore s law. 1 Bandwidth hungry applications such as real-time video processing, cloud computing and social networks have created the need for more powerful data centers that can sustain much higher bandwidths than a few years ago. This trend requires high-performance interconnects that can sustain this high bandwidth without consuming excessive energy. Since this evolution will probably not stop anytime soon, solutions that can be easily scaled up to higher 1 Since 1970, Moore s law is dominating the world of electronics. Every 18 months the number of transistors on a chip doubles. 7
bandwidths are preferred. There is a trend, and also a strong need, to switch from electrical to optical links at increasingly shorter distances, and employ them for rack-to-rack, board-to-board, chip-to-chip and intra-chip interconnects. Silicon photonics, leveraging the well-developed CMOS fabrication infrastructure and its economy of scale, provides a distinct cost advantage over other optical technologies such as InP based photonic circuits for transceiver applications. At the same time it allows lower power consumption and enables the scaling of the aggregate bandwidth of transceivers to the Terabit/s range. Silicon photonics is a rapidly evolving research field that has recently attracted a lot of interest from both academia (MIT, Ghent University, UC Santa Barbara,... ) and semiconductor industry (Intel, IBM, HP, Luxtera,... ). A transceiver, a combination of a transmitter and a receiver, contains various components: ˆ Lasers that generate the light that serves as information carrier, with wavelengths typically around 1.3 µm or 1.55 µm. ˆ Low-power, high-speed optical modulators to transfer the data signal to the carrier light. ˆ Thermally stable wavelength division multiplexing (WDM) multiplexers (MUX) / demultiplexers (DEMUX) to combine / filter out different wavelengths. ˆ Highly-sensitive, high-speed optical detectors. Today, optical modulators, multiplexers and photodetectors suitable for high-performance optical transceivers have already been demonstrated on a silicon-on-insulator (SOI) platform. Depending on the component characteristics, there exist multiple ways to create a high aggregate bandwidth transceiver. A first method uses an array of directly modulated distributed-feedback (DFB) lasers, as can be seen in figure 4.2a. Since external modulators are no longer needed, this has the advantage of a high output power, lower power consumption, compact size and simple operation. However, some applications require extremely high bandwidths, thereby making it necessary to replace the DFB lasers with mode-locked lasers (MLL). Figure 4.2b shows transmission using the generation of multiple frequencies in a mode-locked laser. This system offers a fixed channel spacing and reduced channel crosstalk due to the coherence of the modes. (a) (b) Figure 4.2: The structure of a WDM link with two different transmission schemes: (a) Directly modulated distributed-feedback (DFB) lasers, (b) one mode-locked laser (MLL) generating a lot of different wavelengths. Despite their differences, both configurations have two important things in common. To handle the massive amount of information that needs to be transmitted in current and next-generation optical interconnects, they both use wavelength division multiplexing (WDM). In this 8
technique multiple optical carrier signals are multiplexed on a single waveguide by using different laser light wavelengths to carry different data signals. Through this, the aggregate bandwidth scales linearly with the number of wavelength channels. Secondly, both solutions are scalable, meaning that the number of channels can easily be increased with the demand for more bandwidth. However, the realization of an efficient, electrically-pumped laser on silicon remains a serious challenge because of silicon s indirect bandgap. This requires the integration of III-V semiconductor lasers on the silicon photonic platform. The next two lasers are fabricated in the UGent cleanroom facilities by integrating InP on the silicon platform. 4.3.1 A DFB laser A distributed feedback laser is a laser structure that is designed such that light with only one wavelength is emitted. A cross section of such a laser structure is shown in figure 4.3. We can distinguish different components needed for proper laser operation. Figure 4.3: SEM picture of the longitudinal cross section of a III-V/Si distributed feedback laser. Firstly, we need a laser cavity, since laser action requires the resonance of a cavity. Only light with certain wavelengths, also called longitudinal modes, will survive in the cavity when the optical field is traveling back and forth. The available wavelengths depend on the cavity length and refractive index. Secondly, we need a III-V optical gain medium, that amplifies the laser light. Obviously this requires some form of input energy, so the gain material must be electrically pumped. The gain medium is wavelength dependent, hence only certain longitudinal modes will be amplified. The gain material of the demonstrated laser is made of InGaAsP quantum wells. In the end, to create single wavelength lasers, a grating is used to select one specific wavelength, called the Bragg wavelength. This grating can act as one or both of the cavity mirrors and provides the feedback, reflecting light back into the cavity to form the resonator. The short-period grating will be etched in the silicon layer, as is illustrated in figure 4.3, leveraging the high resolution of the state of the art deep UV lithography that will be used for the fabrication of these circuits. 4.3.2 A mode-locked laser Mode-locked lasers (MLL) are capable of generating stable short pulses which have a corresponding wide comb of optical modes. Most of the building blocks needed to realize a mode-locked laser are similar to the ones needed for DFB lasers. In contrast to the DFB laser, a MLL needs a cavity that can support a lot of longitudinal modes. By making use of the low-loss silicon waveguide structures that can be part of the cavity, a variety of laser cavity designs is possible, like ring-cavity configurations, a Fabry-Perot (FP) laser (figure 4.4), etc. Secondly, a broad and flat gain spectrum with sufficiently large gain value to achieve lasing in the whole required wavelength range is crucial. In the end, we also need a way to lock the modes of the laser in phase with each other. When the different longitudinal modes are not mode-locked, the output of the laser fluctuates in time 9
gain saturable absorber gain DBR III-V-on-silicon optical amplifier and SA silicon spiral cavity mirror cavity mirror silicon spiral Figure 4.4: Schematic overview of a linear cavity colliding pulse mode-locked laser. due to fact that the power distribution over the different longitudinal modes of the laser and their phase relation fluctuates in time. If, on the other hand, the phases are locked to each other, the modes will be in phase at fixed time intervals, resulting in a train of pulses which have a corresponding wide optical spectrum of phase correlated modes. The traditional way to implement this, is by means of a saturable absorber. This absorber is formed by electrically isolating a section of the III-V material in the laser cavity from the gain section and applying a reverse bias voltage to it, so that it absorbs some of the light in the cavity. The mode-locking occurs from the dynamic interaction between the gain and the absorber section. 10