Optoelectronic Devices PDF

OPTOELECTRONIC DEVICES Part I - Semiconductor light sources 1. World semiconductor laser market Semiconductor lasers are a particular type of lasers or also called diode lasers. They are very widespread also becau...

OPTOELECTRONIC DEVICES Part I - Semiconductor light sources 1. World semiconductor laser market Semiconductor lasers are a particular type of lasers or also called diode lasers. They are very widespread also because non diode lasers still usually include diodes as pumps. In the year 2000 semiconductor laser were experiencing a very rapid growth in the marketplace but soon suddenly decreased since smartphones were not as popular and there was no actual need for the services offered by the fiber optic communication although a lot of those sources were available: technology must meet the necessities. Laser type λ [nm] Pmax Dimensions Wall-plug Price [€] [W] [m] efficiency [%] 0.632 He-Ne (red) 10-3-10-2 0.1-1 0.1 100-2000 1064 (infrared) 25(green laser 200 CW* 1 (lamp)** 1 (lamp)** Nd:YAG Can be halved to 0,532 (green) 107 peak 0.01 (DPSSL) 25(DPSSL) pointer) Up to 50K 10 (mid/far-infrared) 104 CW* CO2 Typically used for material processing 107 peak 1 (very big) 10-20 50K Semiconductor 0.38-300 1-100 10-3 (tiny) 50 0.5-15K *CW = continuous wave; peak is referred to the one produced in pulsed regime. ** Values referred to the cases in which the laser is pumped through a lamp or by a diode (Diode Pumped Solid State Laser) With a comparison between other types of popular lasers, semiconductor lasers present some important characteristics: They can be used to obtain wavelengths in a very wide range, compared to other lasers which typically produce a single wavelength. Their dimensions are typically reduced, so even though they are able of producing an output power that is not so high, they can be easily integrated also to be able to build more powerful devices. They have a quite high wall-plug efficiency. They can be mass produced which can result in a significant reduction in cost, which means they can be integrated into more expensive applications. 1  𝐏𝐨𝐩𝐭𝐢𝐜𝐚𝐥,𝐨𝐮𝐭 Note: Wall-plug efficiency is defined as 𝐖𝐏𝐄 = and it is an important parameter 𝐏𝐞𝐥𝐞𝐜𝐭𝐫𝐢𝐜𝐚𝐥,𝐢𝐧 since it is needed to determine how much power is wasted and, since devices can be supplied by batteries, to determine the lifetime of such batteries. Bottom line : Semiconductor lasers offer the largest and most interesting technologies, therefore they are crucial for today’s applications. Most popular applications of semiconductor lasers include: Fiber-optic telecommunications Optical storage (CD, CD-R, DVD, BD) CD DVD BlueRayDisc Wavelength 780 nm 650 nm 405 nm NA 0.45 0.60 0.85 Capacity 700MB 1 layer 4.7 GB 25 GB Mainly used for videogaming, for copyright and since they are large and - 2 layers 8.7 GB 50 GB hard to download Track pitch 1.6 0.74 0.32 [μm] → The holes are λ/2 tall to produce destructive interference, since they are scaled based on the wavelength, the track pitch is limited by the selected frequency. Printers and medicine, for the precision they grant. Sensors, since they are quite small and cheap. Interferometry Solid-state laser pumping Entertainment, for example picoprojectors. 2. Light generation from direct-band semiconductor materials Principle: generation of light (photons) by radiative transitions in a semiconductor diode, where electrons and holes recombine. Electrical current (flow of electrons) is converted into light (flow of photons). There are two types of SC light-emitting devices, which present different characteristics: LED (Light Emitting Diode) SC laser Incoherent, non-monochromatic emission Coherent, monochromatic emission E(t)=E0cos(ωt + ϕ(t)) E(t)=E0cos(ωt + ϕ) Wavelength: from UV to Near Infrared Wavelength: 380-300K nm (NIR): 280-1600 nm High radiance*: 1 MW/cm2 sr Relatively low radiance*: 100 W/cm sr 2 Good reliability** Very high reliability** Low cost (less than 5€) Very low cost (less than 1€) * Radiance R = P / AΩ **Reliability is related to lifetime 2 Optical amplification for a two-level (0 and 1) atomic system is described by analyzing the processes that take place with matter-radiation interactions: The energy difference between the atomic levels ΔE = E1 – E0 = hν is equal to the energy of the emitted/absorbed photon. To guarantee efficient optical gain it is necessary to have a lot of radiative processes, likeable to happen. Spontaneous emission does not always happen, and it is used in LED since the outcoming photon has random direction and polarization. In semiconductors, the structure is a crystal with a huge number of atoms and the behavior is a little different: the orbital description of atoms is no longer valid, but we speak about energy bands whose distribution and ability to make electrons move is different. Their behavior is similar to insulators for low temperature and to conductors for higher temperatures. Crystals are characterized by many atoms forming a lattice and depending on the valence number of the atoms they are able to form combinations, therefore we use a description that is no longer down to the orbitals, where we consider their potentials in a system representing energy: The electrons can have energies in the conduction or in the valence band and can jump from one to the other trespassing the band gap if they receive an energy equal to Eg. If it is not enough, the energy received will be released in any other way, but it will never take the electron to the bandgap. Electrons on the valence band can be absorbed by the conduction band thanks to the interaction with photons having energy superior of that of the bandgap (similar to the phenomenon of absorption). The radiative transition can happen when electrons from the conduction band can decay to the valence band releasing a photon (spontaneous emission). There are also ways they can go back to the valence band without emitting radiation (for example if a phonon, which is the energy caused with vibration of atoms due to temperature, absorbs the photons) so we need to find materials and conditions that make radiation very likable to happen. 3 The process can be described in these steps: 1. The overall charge is 0. Normally in the valence band the electrons recombine without conducting current. 2. If an electron is charged to the conduction band, it can move due to electrical field. This happens through absorption. 3. In the valence band there will be a missing electron, the vacant place is called hole and can “move” if electrons recombine this means, virtually, a positive moving charge is created (electron-hole pair) if you apply an electrical field, electrons and holes will move and are able to conduct a current, separating the charge, as it happens in metals. Metals already have a lot of electrons in the conduction band. 4. If enough electrons are brought to the conduction band, we can reach a situation where we have all electrons in the conduction bands and all holes in the valence band. 5. Electron-hole recombination for an electron to go back down it would need to have a place to go in the valence band to occupy it and disappear from the conduction band. It can only go down by releasing some energy in the form of a photon (spontaneous recombination). For this to happen the electron and the hole need to be in the same place, so it's important to find methods to keep them in the same place. The emission of photons is not always there and more than one recombination at a time cannot happen. In conclusion, we want to favor electron-hole radiative recombination, so we need to: a. Bring a great number of electrons to the conduction band. This could be done through heating, since it gives energy to electrons that can jump to the conduction band, but it's inefficient since you need too high of a temperature to produce light emission. b. Have many electrons and many holes in the same place, enough to raise the probability of radiative recombination to happen. c. Induce the radiative process: the probability can be calculated and it's different from one material to another, so we need to choose one with enough probability. → To make these happen we need a P-N junction in a direct band gap. The PN junction is created by takin an atom out of a silicon crystal and inserting phosphorous (which has an additional electron), the extra electron will automatically go to the conduction band. The more atoms get replaced, the more electrons will be in the conduction band and this process is called N-doping. These electrons are not able to recombine in the valence band since there is no hole they can recombine with. Equivalently we can create more holes by replacing atoms of Silicon with Boron (which has only 3 electrons) and by applying electrical fields, the crystal is able to conduct current, process called P-doping. When joining two materials (N-doped + P-doped), you create a PN junction, a diode: the free electrons and free holes try to recombine, so there is a special charge region in the middle. 4  A typical doping is of the order of 1017 cm-2, to avoid altering the crystal properties. At thermal equilibrium the situation is equivalent to an open circuit. The area where the p and n parts join, represents a built-in potential that prevents current from flowing through the diode since there are no electrons and holes in the same place. To lower this built-in potential, a forward-bias needs to be applied so that the two chargers get to the same place. Quasi-Fermi levels (Ef) are the maximum value of energy an electron can achieve. The status of a carrier is determined by its momentum and energy. The highest energy level increases with the number of electrons in the conduction band. The concentration of electrons and holes has a distribution that decreases exponentially, having most electrons on the lower energy level and decreasing for higher levels. By doping, the energy of the lower level in the conduction band will decrease. To add a forward-bias, we can use two kinds of circuits: A current generator in series, which fixes the current flowing through the diode. A voltage generator in parallel, which fixes the voltage at the ends of the diode. The first option is the best one, since it allows to have a better control of the system. By applying the bias, a zone with both electrons and holes in the same place is created so recombination becomes possible. We still need to increase the probability for radiative recombination, so the right material needs to be chosen. For this choice, two types of semiconductors need to be considered: Direct Energy gap semiconductor: conduction band and valence band are both parabolas and are aligned. Indirect Energy gap semiconductor: the bands are not aligned, so the recombination (which needs to respect conservation for energy and momentum) can only produce nonradiative recombination: before and after the transition from (2) to the maximum of the valence band, the electron changes its momentum. A photon has p=hc and cannot change the momentum of the carrier by much, so after its absorption, it goes to (1) then naturally flows to the bottom of the conduction band. In order to recombine with a hole, it will need a shift in momentum since there are no holes with the same momentum and, for it to be conserved, no photon can be emitted. 5 If the bands are aligned, the transition does not need momentum conservation, so the energy is compensated by emission of a photon. Silicon has an indirect band gap, that is why it is not suitable for producing light amplification. Materials with direct bandgaps include: GaAs, InP and GaN. Silicon and germanium are the best semiconductors (4) so by combining Gallium (3) and Arsenic (5) in a 1:1 ratio, a crystal is built, behaving like a direct bandgap semiconductor. The same happens with In (3) and P(5) or Ga(3) and N(5). Groups from (2) and (6) can also be combined, as Cd(2) and Te(6). As for absorption, it can happen in both kinds of materials, through a vertical transition by supplying enough energy. Notes: A direct bandgap material and a forward bias are needed to build a semiconductor laser. Forward bias is equivalent to population inversion: it translates into having enough holes and electrons in the same place. The pumping in this kind of system is achieved by electrical supply not by optical pumping which is much more efficient and easier to fabricate and control. 3. LEDs (Light Emitting Diodes): materials; structures; operating characteristics; opto-isolators; Solid-state lighting; RGB LED and White LED lighting systems; LEDs for displays and video projection; OLEDs; analysis of RGB LED datasheets. LEDs are diode structures that produce light in all directions with emitted wavelength depending on the band gap of the semiconductor used for its production and can be chosen according to the necessities. Such structures need high concentration of particles rather than great number of particles. Concentration is increased with current, so the device is designed to have current flowing through a very small area: in particular, the two GaAlAs layers are placed to prevent current from flowing towards the outside so that the electrons are concentrated in the central area. In general red-infrared (GaAs) and green (GaP) emitting LEDs are much easier to build than blue LEDs which need GaN. Basically, the current light intensity characteristic is linear and they are used as replacement for lamps, since they are much easier to control and more efficient. 6 Some applications need these devices to produce directional light. In order to achieve it, there are two options: Placing a reflector at the bottom Using a converging lens to narrow the emission. These devices are used in many applications: Transmission of digital signals through optical fiber, since it can easily be done by turning on and off the leading current. (deep) Ultra-Violet (UV) LED are used as replacement for Mercury lamps, since they contain no toxic elements, do not produce hazardous waste, work with lower voltage (6-12 V), do not produce ozone and have very long lifetime. Bio-Photonic applications - @280nm: activation of anti-biotic molecules that attack tumor cells via covalent bond. Opto-isolator: to avoid disturbance they allow the continuity of the transmission of signals for short distance without having physical linking components, so keeping a galvanic insulation between two separate circuits. Solid stage lighting: before 2005, 22 % of electricity worldwide used for lighting, by incandescent lamps (radiation by heating a tungsten filament) which were very inefficient (about 1%). LED-based lighting can be 20x more efficient and 5x more efficient than fluorescent lighting, so the idea is to favor large deployment of solid- state light sources (materials: semiconductors or polymers) for lighting, displays or entertainment (RGB lasers for large image projection). The principle used is the combination of three light sources emitting each of the three fundamental colors: Red + Green + Blue = White (RGB LEDs) so that our eyes can make an average of what all the receptors can detect. There are three types of LED lighting: RGB semiconductor (“inorganic“) LEDs By varying the relative intensity of the light emitted by the R, G and B LEDs, it is possible to change the “color” of white light (and also to generate all the visible colors). Electronic control is needed to realize the different colors and to compensate for ageing (LED efficiency decreases with time) High overall efficiency: 20-30% (to be compared to classical light bulbs, 1%) Durability: 100000 hours and more (i.e., > 11 years) 7 The three colors have different spectral widths and altogether form a spectrum that is continuous from 400 to 700 nm. The relation between light and current is basically linear, although it decreases towards the end because of an efficiency decrease as the input power increases: a lot of the input power is converted into heat which must be taken away from the device, since radiative recombination is overwhelmed by other types of recombination which become more frequent for higher temperature and the effect of thermal resistance (that dissipates PR=Id2Rs) and increases with the temperature. White semiconductor (“inorganic“) LEDs Blue LED generates intense light with high efficiency. Phosphor layer absorbs part of the blue light and converts it into green, yellow and red (plus residual blue) → white solid-state light source. Compact, efficient, reliable → state-of- the-art for “white light” lighting 1x monochromatic LED emitter + light (wavelength) converter (dyes, polymers, phosphors, and semiconductors). Based on the fact that the photon energy of original emitter is higher than that of converted photons → blue emitter + yellow converter (phosphor). They emit a spectrum that is detected from the human eye as white. Organic LEDs (OLEDs) Carbon based, flexible panels, good for large areas, good efficiency, good durability. The conduction is obtained by free electrons, thermally excited that emit light similarly to what happens for two level materials. Advantages: cheap, emit light in all directions (uniform), flexible materials, large area. Drawbacks: low power, not good for lighting, reliability (they are sensitive to humidity) Regular displays have borders which produce light into a matrix of RGB filters and polarizes with liquid crystals, while an OLED display is made of RGB OLEDs each of which is driven by current, they produce better black since it is not a “dimmed white” as in LCD, but it needs different levels of colored OLEDs to get white. They are very used in automotive (taillights) and for light projection. 3D digital cinema projectors require high brightness sources which are still being optimized. 8  4. LUMENTILE European Project: Electronic Luminous tile Exploits frontier technologies in large-area and organic electronics and photonics to develop a new generation of modular luminous components for design-driven applications, namely an electronic luminous tile. The advent of new technologies, capable to integrate into large area circuits semiconductor and organic light emitters (LEDs and OLEDs), allow to fabricate large light emitting objects with increased luminous efficiency, that are perfectly suited to realize the electronic luminous tile. Functionalities: i) the possibility to change the tile color. ii) the possibility to be used as high-efficiency and energy-saving light source. iii) the possibility to equip the tiles with electronic sensors (i.e.: strain, pressure, temperature, etc.) in order to make the tile a sensing element to be deployed and used to detect the presence and the location and the posture (standing, laying) of human beings within a specific ambient, with applications to health/safety monitoring. iv) the use of the tile as individual pixels for the realization of very large indoors or outdoors video displays with unprecedented ruggedness and ease of installation. The integration between the photonics and electronic technologies on the one side, and ceramic materials and structures on the other side will pave the way to the development of revolutionary products that can find widespread applications in different areas and significantly impact everyday life in a few years. 9  5. Introduction to LASER: the laser as an oscillator; optical amplification; threshold condition; pumping methods; oscillation frequency. Light Amplification by Stimulated Emission of Radiation was invented in 1960. Characteristics: monochromaticity (high coherence) Ei=E0cos(2πνst+ϕ(t)), high directionality, high brilliance B [Power/(area)*(solid angle)], single polarization. It is an oscillator of the type: 𝑈𝑜 𝐴 𝐴0 = with A produced by the optical amplifier, of the form 𝐴 = and β produced 𝑈𝑖 1−𝛽𝐴 1+ⅈ(𝜔−𝜔0 )𝜏 by the mirror. 𝑈𝑜 𝐴𝑅 𝐴0 𝜏 = with 𝐴𝑅 = and 𝜏𝑅 = 𝑈𝑖 1+ⅈ(𝜔−𝜔0 )𝜏𝑅 1−𝛽𝐴0 1−𝛽𝐴0 so, when 𝛽𝐴0 → 1 then 𝐴𝑅 and 𝜏𝑅 → ∞, which means the spectrum is infinitely narrow around ω0. In reality, we also have gain saturation that avoids going to infinity and βA = 1 is the equilibrium condition. The output is obtained through a coupler. To produce A we need: a medium with at least three energy levels and a pumping mechanism to reach population inversion, at that point the output characteristic will have a threshold condition (where optical gain equals optical losses, βA = 1) and then a slope. With optical gain G = P(L)/P(0) = eg·L Pumping methods: Optical → absorption of photons Electrical discharge in a gas → collisions between atoms/molecules bring electrons onto excited (i.e., higher energy) states/levels. Electrical → for semiconductor lasers (injection of a forward current into a p-n junction), very efficient. The radiation in the cavity produces gain only if after one complete round-trip into the cavity, the amplified electromagnetic wave adds in phase to itself (constructive interference), so we have a condition for the frequencies oscillating:  = q·2 (q = integer) k·2nL = q·2 (n = refractive index of the active medium) 2·(/c)·2nL = q·2 q = q·(c/2nL) = eigenfrequencies (or longitudinal modes) of the cavity Frequency spacing between adjacent modes:  = c/ 2nL → The cavity contains an integer number of wavelengths. 10 The effective oscillation frequency is determined by combining the longitudinal modes with the frequency profile of the optical gain and a laser can operate simultaneously on several longitudinal modes, or on a single longitudinal mode (the latter operation is typically preferred). Laser characteristics: It emits a coherent electromagnetic wave Laser linewidth:   kHz – MHz, in opposition to a light bulb (  200 THz) It is an oscillator with a high quality factor Q = 0 / = 400 THz / 10 kHz = 4·1010 Types of temporal regimes for the optical output power: o CW (Continuous Wave) o Pulsed (duration from 10-9 s down to 10-15 s) Directional emission: angular divergence is due to diffraction  10-3 -10-4 rad. Using lenses, it is possible to focus the power emitted by the laser onto a very small area (minimum spot size   ) 6. Semiconductor laser basics: structure; electrical and optical confinement; single- and double-heterostructure; optical waveguides (rib, ridge, buried) The mirrors of this cavity are based on the interface between the semiconductor (n=34) and air (n=1) so R=35% which is not much but since the material can produce a very high optical gain in a short piece of material, it still works as an effective laser. They are also very easily parallel due to the crystalline structure of the semiconductor. To guarantee an efficient superposition of optical gain (electrical carrier) and lasing light we need Light (optical) and Carrier (electrical) confinement. z = travelling direction of light xz = junction plane y = perpendicular to the junction plane It took 20 years to realize this structure. 11  Confinement along y directions: The concentration of carriers is an exponential monotonic function of the difference between Ef and the higher/lower energy level of the valence/conduction band. N*P establishes the probability of having radiative recombination. Photons travel along the z direction (orthogonal to the picture) only get amplified if the product N*P is not zero in that area. There are two types of semiconductor junction: when the p-n junctions are made from the same material with different dopage they are called homojunction, while the heterojunctions are made with three different materials p-i-n where i≠p,n and the p and n materials have different bandgaps so the situation will be: So, in the case of the heterojunction the slopes of the difference in the potential barrier are different. The best way to use this advantage is to use two p and n substrates made of the same material and use a third intrinsic material which has a smaller bandgap even without doping so two barriers are created when the materials are joined. They both need to be reduced in a different way when forward biasing the junction: at some point the smaller barrier will change its slope inverting its sign so the concentration of the carrier in the region where the intrinsic material is (d) will increase. As for the optical confinement, to force light into a region with optical gain, physics allows a lucky condition to happen: the refractive index is connected to the bandgap (inversely proportional), so the central (i) part of the material has a higher index, which creates a waveguide, so light is naturally guided to travel into the region with the higher optical gain. This also does not happen for homojunctions. This also happens for LEDs and is therefore crucial for obtaining higher efficiency. It is very important to have a very small d (100-400 nm) since the important thing is the concentration which is smaller for bigger areas. Building such size is very difficult and took a long time for technology development. 12 In conclusion, a double heterojunction is the most efficient and advantageous choice, but it is also more difficult to build. Confinement along x direction We need to create a guiding effect having higher current flow in the center of the junction. This can be done by creating physical barriers that increase the refractive index. There are three solutions: !!! It is never good to expose the diode to air because oxygen is highly reactive and would change the electrical properties of the diode, same goes with water. A typical w is 2-5 μm (high power 100-400 μm) An accurate waveguide design allows single-transverse mode optical propagation, usually the first one. The rib case has a very highly different n situation which makes it very hard to get only one mode to propagate. To solve this, the ridge structure is created: light senses that the center n is higher because there is semiconductor above at the sides of the junction. The difference in diffraction index is small enough to allow single mode propagation. Buried waveguides need a regrowth of semiconductor material which is a pretty difficult process, so the ridge structure turns out to be the best option. Analyzing the Power - current / Light – current characteristic, we can observe that at the beginning we only have spontaneous emission, until Ith (threshold current) which corresponds to the situation in which the optical gain equals the losses, then the concentration of carriers can increase, since increasing the current the concentration of electrons and holes in the active region increases until the threshold, since at certain point (loop gain = 1) it remains constant. From the threshold on, the number of electrons converted to photons increases with the current, linearly. 13  7. Historical evolution of semiconductor lasers 1958 Schawlow and Townes (Bell Labs, USA) invented the concept of the laser. They shared the Nobel Prize. 1960 Maiman (Hughes Research Labs, USA) builds the first working laser (ruby laser). 1962 the first semiconductor laser is demonstrated almost simultaneously by IBM, General Electric and MIT-Lincoln Labs (USA). 1963 Alferov, Kroemer and Kazarinov invent the concept of double heterostructure SC laser. Alferov and Kroemer shared 1/2 of the Nobel Prize in 2000. 1970 first demonstration of double-heterostructure SC laser. 2014 Nakamura wins the Nobel Prize for invention of the blue LED / blue semiconductor laser. The first semiconductor laser 1962 had many problems: a threshold current of 19 A Was shot into tiny devices, so to avoid damaging it, had to be operated in pulsed regime which is not practical. It was operated at 77 K: it had to be immersed in liquid nitrogen for radiative recombination to happen more frequently respect to other processes. It had a very short lifetime. Semiconductor lasers are used to make optical fiber a mean for optical communication: initially it used λ= 850 nm, now use for memory reading, then 1310, 1550 became used. 8. Materials for semiconductor lasers: heterojunctions; lattice matching; III-V AlGaAs ternary system and InGaAsP quaternary system; II-VI material systems, active region: bulk, Multi-Quantum Well, Quantum Dots; optical gain To build a double hetero-junction we need to make sure that the two crystals are well electrically and physically connected. Crystals are characterized by their lattice constant (space between the atoms), when we connect two different materials, we need to avoid traps: discontinuities that translate in changes of the energy bandgap which make atoms and electrons recombine non-radiatively. This is a huge problem when two materials with different lattice spacing are connected, since it happens in the connection space. To avoid this, we need to find two materials with the same lattice constant, but still different bandgap, so we need lattice matching and direct band gap. We also need to take into account what wavelength we need the material to operate at. 14 They created a ternary compound with X concentration changing according to the desired bandgap, working on the first transparence window. For instance: GaxAlAs1-x GaAs GaxAlAs1-x To operate on the second or third windows they created a compound called InGaAsP: InGaAs (different X for the right wavelength with varying bandgap) + P (to have the same lattice constant) Since d must be 0.1-0.4 μm and since we have lattice constant of 100 nm we have approximately 200 atomic layers. When joining the materials, we must make a very sharp junction and we need to mix no more than one atomic layer to avoid damaging radiative processes, the growth of the crystals needs a huge precision. For larger wavelength (gas detection) lasers were very unpractical, they have been replaced by quantum cascades. Research made it possible to replace “bulk materials” for the active region with multi- quantum well structures by growing the material that looks like: Electrons fall from the area █, likely into the wells and holes tend to flow towards the top of █. This wells work as traps for charges that induce them to go in the very same position, so strongly increasing the probability of radiative recombination and the optical gain. Since the walls are tiny, the energy levels of atoms that are strongly confined or discrete in those areas. High efficiency, higher gain = Lower threshold current, higher emitted power, larger modulation bandwidth. In this case lattice matching is not necessary, however these are very hard to grow. 15  9. Technologies for semiconductor laser fabrication: epitaxial growth (Liquid Phase Epitaxy, Vapor Phase Epitaxy, Molecular Beam Epitaxy, Metal-Organic Chemical Vapor Deposition); growth of quantum dots; definition of optical waveguides; photolithography and electron beam lithography; masking; wet chemical etching and dry plasma etching; metallization; regrowth; details of transversal semiconductor lasers sections; bonding; cleaving; packaging. SC laser fabrication process Technology consists of three main steps: Growth of SC ‘wafers’ with the appropriate (hetero)structure Wafer processing, which comprehends lithography, etching, insulation, contacting, scribing (separation into single lasers) Packaging Material growth Requirements: low defect and impurity concentration (< 0.1 ppm) and ‘good’ interfaces at heterojunctions, so we need epitaxial growth: from an existing crystal, atomic layers are grown to create new materials on top of it and they need to be planar and consistent. Also, the yield must be very high so the wafers must have as many working devices as possible. The techniques used can be: Liquid Phase Epitaxy - LPE (1962, AlGaAs), single heterojunction. Molecular Beam Epitaxy - MBE (1971, AlGaAs. Growth of few atomic layers → double heterojunctions) Metal-Organic Chemical Vapor deposition - MOCVD (1980, also InGaAsP. Growth of Multi-Quantum-Wells) Wafer processing For a ridge waveguide these steps are followed: Note: it is important to allow heat to flow from behind to a high conductivity material to reduce thermal resistance (dissipation of energy). The layer should not be under 100μm to avoid having a material that is too fragile. Dry etching is a process in which ions are shot to remove atomic layers of some material. 16 Between each step, accurate cleaning and measuring steps take place, along with the large number of steps related to the production of the wafer, this makes these processes very expensive. For a buried waveguide a cladding layer is required which makes a regrowth after the waveguide etching necessary. The flow of the current through the central area is guaranteed due to the smaller bandgap, however current leakage is still possible for higher currents. These processes are higher in cost than the ridge waveguide. Laser scribing Once the wafer has been processed, many waveguides are cut with a diamond edge into bars and then into the diodes so that the cut surface remains perfectly sharp. This operation is called laser scribing.At the final stage they are so small they cannot be handled with hands and need very good care. These processes take place in clean rooms, in which there is a tiny concentration of particles, absence of dirt or dust to avoid any undesidìred particle to contamine the device, they are very expensive to run. Also the room needs to be lightened by yellow lights, to avoid having the photoresist reacting to the blue light. Packaging The final device needs to be inserted into a packaging that is easier to handle. For most SC layer, this step represents the higher produciton cost since it cannot be done in batches. Packaging is needed for: protection ease of handling heat dissipation, onto the metal structure monitor photodiode to avoid wasting light, used to control and know the power usage coupling to optical fiber direct large-bandwidth modulation Automated processes reduce cost but are not always possible. 17 A typical packaging process is: The materials are welded with strong laser beams which can melt some spots without moving the structure so that very precise fiber coupling is guaranteed. Heterostructures Quantum wells usually have two degrees of freedom, so electrons can move across a sheet. To have even stronger confinement, quantum wires (2D confinement) or quantum dots (3D confinement) can be built to produce the highest gain, although these structures are clearly not easy to achieve. A practical result of going to lower dimensions is the reduction of the minimum current required for a semiconductor laser to start lasing. This results in a higher efficiency and a better performance for the laser diodes. Crystal growth In general, the construction of crystals for semiconductor lasers follows Vergard’s laws: For bandgap energy: The bandgap energy Eg(x) of a ternary compound AxB1-xC varies with the composition x as: Eg(AxB1-xC) = xEg(AC) + (1-x)Eg(BC) – bx(1-x) where Eg(AC) and Eg(BC) are the bandgap energies of the binary compounds, and b is called the bowing parameter. 18 For the lattice constant: The lattice constant a of the ternary compound IIIx-III1-x-V varies with the composition x as: a (AxB1-xC) = (x)aAC + (1-x)aBC where aAC and aBC are the lattice constants of the binary compounds AC and BC. This law expresses that the lattice constant of a ternary alloy varies linearly between that of the two binary compounds it is composed of. For a quaternary compound: AxB1-xCyD1-y where A, B, C, and D are either group III or group V elements: Vegard’s Law for lattice constant: the lattice constant a of the quaternary compound varies with the composition x,y as: a = (xy)aAC + x(1-y)aAD + (1-x)yaBC + (1-x)(1-y)aBD Vegard’s Law for bandgap energy: the bandgap energy Eg of the quaternary varies with the composition x,y as (the bowing parameter is neglected): Eg = (xy)EAC + x(1-y)EAD + (1-x)yEBC + (1-x)(1-y)E It is customary to represent the GaInAsP, AlGaInAs, AlGaInP compound semiconductor system using the diagram on the right. The coordinates of each point in the diagram corresponds to the composition (x,y) of a ternary or quaternary alloy. All the semiconductor compounds in this diagram have a direct bandgap, except those represented in the blue shaded area: 19  Single Crystal Growth Techniques Czochralski Growth Method Most widely used for silicon. High purity pieces of polycrystalline material (called the charge) are melted in a quartz or graphite crucible. A seed crystal is lowered into the melt and drawn up at a carefully controlled rate. The charge and the seed are slowly rotated in opposite directions to ensure uniformity. The material in the melt will make a transition into a solid-phase crystal at the solid-liquid interface, replicating the crystal structure of the seed. The resulting single crystal is called the boule. Very large boules can be achieved (up to 12” for Si). The size and crystalline quality of the boule that can be realized depends on the thermal and mechanical properties of the crystal. Impurities can be intentionally introduced in the melt to n-type or ptype dope the growing crystal. Parasitic impurities can also be introduced by the crucible. Epitaxial Growth Techniques Liquid Phase Epitaxy LPE is a thermodynamic equilibrium growth process. Growth is through the precipitation of material from a supercooled solution. The substrate is placed on a horizontally sliding graphite boat. Excellent quality and purity material can be grown. Three parameters control the growth: melt composition, growth temperature & duration. Advantages: Simplicity High purity Disadvantages: Poor thickness control High surface roughness Melt back effect High growth rate & not abrupt interfaces. 20 Vapor Phase Epitaxy VPE is a thermodynamic equilibrium growth process too. There are three major chemical reaction zones maintained at three different temperatures. 1/ The group III precursors are pure metal elements in a vessel. They are heated into liquids which then react with the flow of HCl gas to form volatile chloride compounds. 2/ The group V precursors are gases which are introduced and decomposed in the 2nd zone. 3/ The 3rd zone is the growth region where the group III and group V volatile molecules react on the substrate surface. Advantages: Flexibility for doping control Control of composition gradients Disadvantages: Difficult to grow very thin layers Hard to grow multi-quantum wells and superlattices Formation of hillocks and haze Interfacial decomposition Metalorganic Chemical Vapor Deposition There are four major parts of an MOCVD reactor. 1/ Gas handling system, which includes: alkyl sources hydride sources inert or “carrier” gases: H2, N2, Ar, He pneumatic and manual valves mass flow controllers pressure controllers clean & leak tight stainless steel tubing 2/ Reactor chamber. 3/ Exhaust system. 4/ Safety apparatus which consists of toxic gas monitors to detect & quantify the presence of toxic or flammable gases. Advantages: Highest quality semiconductor thin films Flexibility Very thin layers possible (a few atomic layers) Abrupt interfaces Disadvantages: Toxicity of gas precursors used 21 Molecular Beam Epitaxy An MBE reactor is composed of: 1/ Precursors, such as: solids (Ga, Al, In) heated above their melting points in effusion cells → the atomic beam fluxes are controlled by mechanical shutters gases (AsH3, PH3) connected through an injector → the flow of gases is controlled by pneumatic and manual valves, and mass flow controllers 2/ Reactor chamber: leak tight made of stainless steel encloses a rotating block and the substrates block is heated and temperature is monitored by a pyrometer maintained at ultra-high vacuum (pressure~10-9 Torr), with walls cooled by LN2 (77K) allows easy in-situ characterization technique (RHEED, Auger, XPS, LEED, SIMS, ellipsometry…) 3/ Exhaust system. The principle consists of the fact that Atomic and molecular beams are projected at the substrate surface. 22  Wafer Processing Lithography transferring a pattern into the wafer, can be done through: 1\ Indirect write: optical techniques 2\ Direct write: non-optical techniques. 1) Conventional Optical Lithography Technique Photolithography Any optical system includes a light source, reflector and a lens, all of the geometries are of the order of 1 cm or larger. Once the feature size on the mask approaches the wavelength of light, one must consider properties such as diffraction and interference. The smallest feature to be imaged with a contact exposure system is limited to: 𝑊𝑚ⅈ𝑛 ≈ √𝑘𝜆𝑔, where k= const. λ =Wavelength of light, g= gap The resist layer is deposited by spinning resist: the wafer is mounted on a vacuum chuck, a predetermined amount of resist is dispensed on the surface of the wafer and a uniform layer is formed as a torque is applied to the chuck, rapidly 𝑘𝑃 accelerating it. The final thickness is given by 𝑧 = where P √𝜔 is the percentage of solids in the resist and 𝜔 is the angular velocity of the spinner. The basic steps of photolithography are: These processes take place in clean rooms, which make them hugely expensive to run. A photomask is a plate of glass with alternate clear and relatively opaque regions. They can be of different types and generally are fabricated with different etching and cleaning steps. Aligning a mask implies three degrees of freedom between mask and wafer (x,y and angle θ) and it is crucial for the creation of the device. It can be aided with alignment marks. 23 2) Non-Optical Lithography Techniques Electron Beam Lithography To produce extremely fine features, electron beams need to be used because: Electrons can be formed into a beam with a shape as small as ≤ 100 Å, as opposed to 5000 Å for light. Electrons can be deflected and modulated with speed and precision by electrostatic or magnetic fields. The energy and dose delivered to the electron-sensitive resist on a wafer can be controlled precisely. Electron beam lithography (EBL) systems may be used for either mask generation or to directly write pattern on the wafer. All electron beam systems have in common the need for an electron beam source with a high intensity (brightness) (A/cm3-sr) high uniformity small spot size good stability long life. Electrons can be removed from the cathode of the gun by heating the cathode (thermionic emission), applying a large electric field (field aided emission), a combination of the two (thermal field aided emission), or with light (photoemission). Etching Pattern transfer from the mask to wafer. Image transfer from resist into a layer under the resist by: 1) Wet Chemical Etching Wet chemical etching consists of three processes: movement of etch species to the surface of the wafer chemical reaction with the exposed film that produces soluble byproducts movement of the reaction products away from the surface of the wafer. Wet etching is a purely chemical process. Disadvantages: Lack of anisotropy* Poor process control Excessive particle contamination Limited by removal rate of the reactants Advantages: High selective etching Often does not damage the substrate. Wet etching depends also on the crystallographic orientation of the semiconductor crystal, which determines the atomic packing density of the different planes exposed to the etching chemicals. Etch planes and profiles when the resist is oriented along various directions on a (001) GaAs wafer. 24 2) Plasma (Dry) Etching A plasma contains mostly unreacted gas molecules, electrons, ions and radicals. Types of Plasma Etching: 1. High Pressure Plasma Etching (Process pressure of between 102-10-1 Torr) 2. Reactive Ion Etching (Process Pressure of 10-1-10-3 Torr) (emphasis on high directionality) 3. Sputter Etching (Process Pressure of 10-2 Torr) 4. Ion Milling (Process Pressure of 10-3-10-5 Torr) Advantages of etching in a plasma environment compared to wet etching: Plasmas are much easier to start and stop than simple immersion wet etching Plasmas etch process are much less sensitive to small changes in the temperature of the wafer High anisotropy for small features. 1. A feed gas introduced into the chamber must be broken down into chemically reactive species by the plasma 2. These species must diffuse to the surface of the wafer and 3. must be absorbed 4. Once at the surface, they may move about (diffuse) until they react with the exposed film, 5. desorbed from the surface and diffuse away from the wafer and 6. Be transported by the gas stream out of the etch chamber. The reactive ion etching (RIE) is similar to plasma etching. In the RIE system the emphasis is primarily put on the directionality of the etch. Note that RIE system operates at much lower pressure of 0.01-0.1 Torr. Advantages: High anisotropy and directionality Disadvantage: The stage needs to be cooled in order prevent the temperature rise caused by the plasma Bottom Line: Wet chemical etching: contact with a liquid bonding with reactive species Dry (plasma) etching: flux of high energy particles, plasma is a mix of ions, radicals and gas molecules which act as high-power coils varying strong electrical fields. 25  Metallization The metal material used in the metallization must satisfy a number of properties such as: Have a good electrical current-carrying capability A good adhesion to the top surface of the wafer A good electrical contact with the wafer material Be easy to pattern (etch or lift-off) Be of high purity Be corrosion resistant Have long-term stability Normally, a multilayer metal film is required in the circuit since one metal material is often not sufficient to satisfy all the properties. There are three most common metal deposition techniques: 1) Evaporation The deposition of thin metal films on a semiconductor wafer is commonly accomplished through vacuum deposition. A typical vacuum deposition system consists of: bell jar (quartz or stainless steel) including: metal sources (placed in crucible) wafer holder shutter thickness rate monitor ion gauge to monitor the chamber pressure pumping system(e.g. diffusion pump) 1\ Filament Evaporation The metal can be in the form of wire wrapped around a high temperature coiled tungsten OR stored in tungsten boats that can be heated electrically by passing current through it Disadvantages: Not very controllable due to temperature variation along the filament Source material can be easily contaminated Mixture of metal alloys containing metals like Ti, Ni and Au are difficult to achieve due their different evaporation rate at a given temperature 2\ Electron beam Evaporation consists of A copper holder or crucible with a center cavity which contains the material A beam of electrons generated and bent by a magnetic flux so that it strikes the center of the charge cavity Water cooling system to cool the crucible and to keep the edges of the metal in solid state Limitations: Can only evaporate one metal at a time unless multiple guns are used 26 2) Sputtering Deposition A slab (or target) of the desired metal which is electrically grounded and serves as cathode is bombarded by positively charges Ar ions accelerated toward the cathode. The sputtered metal atoms leave the surface and evaporate on the wafer located below the cathode. Advantages: Composition of the deposited film is precisely determined by that of target material (the sputtered atoms are deposited on the wafer with no chemical or compositional change) Step coverage is improved Sputtered film has a higher adhesion This is the best method even if it is more complex. 3) Electrodeposition Due to the Electric Potential caused by the Bias Voltage, the electrolyte solution splits into its positive and negative ions. The Positive ions (the metal particles) will move towards the Cathode and deposit themselves onto the substrate. The negative ions in solution will move to the Anode. Packaging Dicing After the wafer is processed it normally contains a large number of equivalent integrated circuits (die chip) that need to be separated from one another. This is accomplished by using: (a) A diamond saw (b) A scriber (c) A laser beam for higher accuracy Wire Bonding After the wafers are diced and inspected under microscope, they are ready to be wire bonded to the rest of the circuitry. It is necessary to link the metal interconnects with microscopic sizes to macroscopic electrical connectors. 27 Encapsulating in a package After the device is fully wire bonded, it is ready to be packaged. The TO package utilizes A pie shape header where the die is mounted to the center of the gold plated header Wires are connected from the die pad to the Kovar lead posts that protrude through the header A glass-to-metal cap is sealed over the die chip to protect the device 28  10.Operating characteristics of Fabry-Perot semiconductor lasers: light-current curve; emission spectrum; temperature effects (on threshold and emission wavelength); output beam characteristics (spot size, angular divergence, polarization); laboratory instrumentation for the characterization of semiconductor lasers; constant current electronic circuit; high- frequency direct current modulation; laser safety (at a glance); reliability (at a glance); analysis of semiconductor lasers datasheets P-I curve also depends on the temperature since, with the increase in temperature, the probability for non-radiative recombination processes to happen increase as well. Mode-hopping of Fabry-Pérot laser Lasers are supposed to produce monochromatic light, however this does not actually happen, but we obtain something that looks like (d). That is because FP cavities only allow some 𝑐 modes to propagate, separated by 𝛥𝜈𝐹𝑆𝑃 = 2𝑛𝐿 which, since the production of SC laser is approximately 100-400 GHz which is very large compared to other types, however it has a very broad optical gain (80nm → some THz) so it still presents superposition of more than one longitudinal mode that make the condition gain=loss possible: they propagate and produce multi modal spectral range. This is far from the ideal monochromatic light emitter, which leads to the conclusion that even though FP lasers are easy, they are not always the best solution. The P-I curve is linear above threshold so it can be used for power modulation through the input of current carrying the digital signal, for this reason SC lasers are widely used for digital communication. To measure the speed of the response, the sign function is applied at different frequencies and the result is that it depends on the bias current, however modulation is possible up to 10 GHz. 29  Laser beam properties The properties of the laser beam are the worst characteristic for SC lasers. The shape of the emitted mode is elliptical since both in the ridge and in the buried structure, the width W (of the waveguide) and the thickness of the heterojunction D are quite different: W ≈ 1-3 μm and D≈ 0,1-0,3 μm so in the near field, the shape is an ellipse with axis 1x3μm. This implies some issues for coupling optical fiber and the fact that the gaussian beam in free space produces divergence angles that are different in the two directions, and we have higher divergence for smaller intensity in the near field. To solve this, some optics can be used, although the emission area is small, so the beam divergence is intrinsically strong. Output optical spectrum Usually, the spectrum covers four lasing modes (within 3dB from the main), some lasers only have this behavior just above threshold but become monomodal for higher power. For homogeneous gain, each mode takes away some population inversion from the others, for a broadband gain, modes do not take away population inversion since they interact each with different sections of electrons and holes. Homogeneous gain SC lasers become monochromatic only for higher power. Usually having a single mode is necessary for sensing or telecommunication, but in most other cases having four modes is just fine. Temperature Variation A temperature increase causes two effects: 1. Thermal expansion of the cavity (L increases). Consequences: Change in longitudinal mode spacing  = c/2nL [DECREASES] Change in frequency of order “m” lasing mode: m = m· = m·(c/2nL) [DECREASES] Change in wavelength of order “m” lasing mode: m = c/ m [INCREASES] CONTINUOUS red-shift of the Fabry-Perot mode pattern 2. Red-shift of semiconductor peak gain (bandgap shrinkage) (with a higher rate with respect to the F-P mode pattern). Hence: We observe mode-hops (order “m+1” mode experiences a higher gain, and it is above threshold; while order “m” mode is below threshold). An increase of pump current always produces a temperature increase due to Joule effect (electrical power dissipation occurs on the series resistance of the laser diode; typical value Rs = 2-10 ). 30 Hence, mode-hops occur also when the current is increased, even if the heat-sink temperature is kept constant (the junction temperature always increases). Variations in term of temperature lead to jumps in the linear increase and different jumps in the decrease. This can be a problem in some applications. This happens due to the fact that Lasers are non-linear devices and tends to hold the modes a little longer, even when the following is the one corresponding to the highest optical gain: it remains in lasing conditions for as long as possible, so this results in a small latency. Another effect linked to temperature increase is the increase of thermal resistance dissipation which allows more current to flow through the resistance as the temperature increases. Typical values are 2-10  for FP and DFB lasers and 300-500  for VCSEL. Since both I and T produce a change in wavelength, it is not predictable: whether this is a problem or not, depends on the application. Use of SC lasers A current generator is required: laboratories use modular instrument dealing with the field through dedicated electronic circuits. Power stability is guaranteed at constant current P, depending on T, to solve this there are two solutions: feedback from monitor photodiode or a temperature control (thermistor + Peltier cell) Thermistor = NTC (Negative Temperature Coefficient) resistor TEC = Thermo-Electric Cooler The transistor represents an high impedance for the creation of an ideal current generator. The laser needs to be short circuited since it is very sensitive to electrical discharges. 31  Safety Lasers can be dangerous because of their high radiance and high intensity. Collimated or focused beams are dangerous for the human eye even at low power (few mW). Class 1 → high-very safe: the retina does not absorb those wavelengths. Class 2 → like looking at the sun, we have the automatic reflex that makes us close our eyes, so it is not extremely unsafe. Class 4 → dangerous even for the skin, can dig holes in metal. Reliability Modern SC lasers are reliable devices and have a lifetime of over 35000 hours (1 year = 8760h). Severe reliability testing for TLC lasers needs to be done for DFB sources for WDM and pump lasers for EDFA (submarine installation). SC laser degradation analysis is lead at high optical power for higher temperatures since defect increase temperature and absorption, therefore break the crystal (atoms start to move around and increase defects). Other degrading conditions can be high currents, electrostatic discharge, Catastrophic Optical Damage (COD) at mirrors / end-facets. Tests are carried out in ESD, which is safe areas for example with conductive floor and shoes that do not make good insulators, at high temperature. With aging, there is a decrease in efficiency which can be fundamentally damaging for pump lasers. To solve this problem, Facet passivation is installed: an insulator at the interface prevents oxygen from entering and oxidizing the material. Applications Summary SC lasers low cost (typically) Wide wavelength range (380 nm - 300000 nm) Small size  high efficiency (35-50%) high reliability easy modulation of emitted power Integration with other optoelectronic devices. 32  11.Simplified theory of the operation of semiconductor lasers: rate-equations; spontaneous recombination (radiative and non- radiative); stimulated recombination; waveguide losses; equivalent mirror losses; threshold current calculation; differential and quantum efficiency; frequency response; relaxation frequency; relative intensity noise To study the operation of semiconductor lasers, a phenomenological approach is used: some equations are used that do work on what happened but do not come from principles. To study them, some assumptions are made: The active region is neutral N and P have the same concentration, so we consider N The geometry is given by V= d*w*l The state variables we consider are: N → volume carrier density S → volume photon density They can change in time, but we consider them constant throughout the active region. The results provide a quite accurate behavior, but it is not what happens in reality where, for example, S grows exponentially since they are amplified. The non-uniformity is actually not that big, of a factor 2-3. We consider the spatial average as a good approximation which makes the calculations a lot easier. We want to establish Rate Equations: using time derivatives, they state how physical quantities vary over time. We can compare the system to a water basin, where the basin is the active region, the water is the carrier density, the tab is the injection current, the water is the carrier density, the tab is the injection current, the holes on the right represent emitted photons and the ones on the left represent the losses. 33 Below the threshold we need to increase the water flow and the number of photons increase slowly, while above the threshold if we increase the water flow the level remains constant and the number of photons emitted increases quickly, almost as if the water that comes in is immediately out. We define the carrier rate equation with A, B and C vary according to the type of semiconductor. Also, C depends on the temperature and increases as the temperature goes up, we can assume temperature increasing the size of the hole to the left of the basin. As for the Photon Rate equation we can consider gmat and α as the optical gain and losses per unit length of the material and we obtain: where vgα represent the total loss. (usually 𝜏𝑃 ~5-3ps) In general g never crosses 0, since there will be absorption so when N=0 and g becomes negative, there will be a lot of holes and no electrons. Some losses generate electrons, so they are not completely bad, but some photons are lost forever. The total laser cavity losses are given by: αtot=αW + αm where αW is due to the waveguide since the walls may not be perfectly straight so might produce some cavity losses and αm is due to the fact that 1 1 some are lost going out of the mirror and is given by 𝛼𝑚 = 𝐿 𝑙𝑛 (𝑅). In general, the material gain depends on the carrier density. 34 The final rate equations obtained are: They are not linear and are barely ever solved, normally they are discretized and constructed by points. They carry no information on the spectrum, so densities should be replaced by electrical fields and some boundary conditions should be imposed. It is interesting to study the steady-state solutions: To obtain a smaller current: N0 and a depend on the material, so one with bigger a should be chosen. Reduce the losses related to the waveguide q and 𝝉𝒄 cannot be changed w should be chosen for a single mode waveguide and not too small l also appears in the waveguide losses so reducing it would increase losses. d can be as small as possible to increase carrier density, it is usually a fraction of a micron. To analyze the output power (emitted by both faces of the laser), the rate equation above threshold should be considered. Quantum efficiency only depends on losses, it is not efficiency in general since it does not depend on the input power: it states how many incoming electrons turn into photons above threshold. 35  Frequency response and relaxation frequency To analyze the frequency response a perturbative analysis is led: in putting a small signal. The so-called laser dynamics depends on the coupled carrier and photons rate equations. The frequency response for a small current modulation [ Pout ( )/Imod ( )] shows a resonance, and it depends on the DC emitted power. The resonance frequency is called relaxation frequency, and it is a combined effect of the energy exchange between carrier and photon populations. 3-dB modulation bandwidths of few tens of GHz can be achieved. However, this huge bandwidth cannot be effectively used in long-range telecommunication systems due to the interplay between frequency chirp and fiber dispersion. The response for step current modulation shows damped oscillations at the relaxation frequency. By analyzing the response, it shows that with low bias current (45 mA), the 3dB cut-off frequency is a few GHz and the resonance is quite high. When the bias current increases, the cut- off frequency also increases but the resonance is almost null. Both fRO2 and γ increase linearly with the current above threshold (I-Ith). 36  RIN - Relative Intensity Noise Quantum noise associated with the photon flux emitted by an ideal laser (coherent radiation) is called shot-noise. In power (spectral density): SP(f) = 2·h ·P0 [W2/Hz] (P0 is the power emitted by the laser) In photocurrent (spectral density): SI(f) = 2·q·I0 [A2/Hz] since a fluctuation of photons is converted into the fluctuation of electrons. Noise generated by a real semiconductor laser is larger than shot-noise SP(f) = 2·h ·F · P0 where F > 1 is the excess noise factor. To measure this, the RIN (Relative Intensity Noise) is defined as: 2 ⋅ 𝑆𝑝 (𝑓) 𝑅𝐼𝑁 = [1/𝐻𝑧] 𝑃02 In an ideal laser (P0 = 1 mW = 0 dBm;  = 1550 nm) → RIN = 5.1·10-16 [1/Hz] = -153 [dB/Hz] The RIN spectral density of a semiconductor laser is not white (i.e., constant over frequency). Rather, it exhibits a peak in correspondence with the relaxation frequency. 12.Experimental techniques for the characterization of semiconductor lasers: internal losses; optical spectrum; linewidth; relative intensity noise; modulation bandwidth; linewidth enhancement factor Experimental techniques for the characterization of semiconductor lasers can be done by extracting certain parameters: P-I curve and internal losses Optical spectrum Linewidth Relative Intensity Noise Frequency response Linewidth enhancement factor (“-Factor”) Differences in performance for lasers with different active materials. P-I curve and Internal Losses (W) The method would usually consist in lighting a laser through the waveguide and measuring the power, then dividing the waveguide and then measuring again. In SC lasers this cannot be done since the material produces a gain, so in this case the efficiency is calculated with the formula *, so first1/ηq is fixed, then m is extracted and consequently W is calculated. 37  Optical Spectrum A Grating-based Optical Spectrum Analyzer (OSA) is used. It has a resolution: 0.01 nm - 1.25 GHz (@ 1.55 m). After the diffraction grating, each wavelength is diffracted with different angles and by turning the exit slit in time. To have a better resolution the slit must be as thin as possible, but still enough to get enough light to the photodiode. A Fabry-Perot Filter has FSR = c/2L; Resolution FWHM = FSR/Finesse For example: Finesse = 60; Resolution FWHM = 1 MHz → L = 25 cm → difficult to align while keeping Finesse high. To have a narrow FSR two cascaded F-P filters must be used. Linewidth Hypothetically a laser emits one single frequency so its spectrum should consist of one main peak. Due to spontaneous emission, the other frequencies still correspond to a constant, so the spectrum looks like the picture. The smaller peaks represent longitudinal modes except the two next to the main peak: they are due to the relaxation oscillation. However, the main peak is not a single line since in a real oscillator there still are some phase jumps, so the main line has some width. The width of the main peak is an important parameter in the frequency domain, but it cannot be calculated with the OSA. If we consider τc coherence time (how long it can be 1 considered a pure sine) and Lc coherence length we can obtain the linewidth as: 𝛥𝜈 = and 𝜋𝜏𝑐 Lc=c τc and from theory   1/P, where P is the emitted power. To measure the linewidth there are three methods. Fringe visibility This method is based on an unbalanced (all-fiber) Mach- Zehnder interferometer: the upper path has a PZT that makes the path longer, so the two paths recombine with a different 𝐼 −𝐼 phase. Considering 𝛾 = 𝐼𝑚𝑎𝑥 +𝐼𝑚𝑖𝑛 𝑚𝑎𝑥 𝑚𝑖𝑛 and τ length of the second arm, we can calculate τc after having measured 𝛾, since 𝛾𝑎 (𝜏) = ⅇ −𝜏∕𝜏𝑐 = ⅇ −𝛥𝐿/𝐿𝑐 38 Heterodyne method Beating on a high-speed photodiode of the laser under test and a narrow linewidth reference laser (ECL = External Cavity Laser, tunable). ECL and SRL have a similar oscillation frequency, optical spectrum is down converted to RF frequencies, where we obtain the convolution of the spectra of the two lasers oscillating at the frequency difference between the ones of the two lasers. It is then converted to the frequency domain to measure this difference. Self-Homodyne / Self-Heterodyne methods Self-Homodyne → beating of laser field with a delayed replica of itself, the delay must be much longer than coherence time to obtain the auto-convolution of the laser spectrum, down converted around DC frequency. Self-Heterodyne → optical frequency shift (via acousto- optical modulator), the beating signal is moved away from DC to achieve better accuracy in the determination of the spectral shape and the FWHM. RIN measurement To measure the RIN two methods can be used: Using a photodiode connected to an electrical spectrum analyzer → the current flowing through the photodetector is measured along with the fluctuation of the optical power. Using a High resolution OSA → resolution of approximately 10MHz to see the RIN. Frequency response measurement Conventionally, a high-speed modulation of the injection current is used. A Network Analyzer is needed and is calibrated by replacing the laser with a large-bandwidth electro- optical modulator (to take into account the frequency response of electrical cables and photodiode). The measurement is (negatively) influenced by electrical parasites within the semiconductor laser. Optical Modulation Method eliminates the effects of electrical parasites → true (intrinsic) frequency response of the laser. For this method ECL + Electro- optical modulator is needed so that an amplitude-modulated ECL modulates the carrier density within the DUT 39  Linewidth Enhancement Factor (-Factor) Around 1975, Semiconductor lasers operating CW at room temperature became available, then experimental characterizations showed that the linewidth of SL was 20-40 times larger than that of other lasers (gas, solid-state) for the same emitted power. Hence in 1982: C. H. Henry explained this by introducing the -Factor → a variation in the carrier density produces: a variation of the gain a variation of the refractive index → carrier density fluctuation causes refractive index fluctuation, so instantaneous fluctuation of the lasing frequency which results in linewidth broadening. 𝜕𝜒 ⁄𝜕𝑁 4𝜋 ⅆ𝑛  SL =  Schawlow −Townes (1 + 2) 𝛼 = − 𝜕𝜒𝑟⁄𝜕𝑁 = − 𝜆 ⋅ ⅆ𝑔 𝑖 N is the carrier density, dg the differential gain and dn the differential refractive index. A non-SC laser (gas laser) has  = 0. The -Factor influences: linewidth, chirp under amplitude modulation (additional frequency modulation), mode stability, laser dynamics, optical injection, optical feedback, filamentation in broad–area lasers (with a ridge of 200μm). A low -Factor is generally good → typical values:  = 2.5  6. SUB-THRESHOLD (Hakki-Paoli) there is a refractive index change (dn) and differential gain (dg) are measured from the frequency shift and valley-to-peak ratio of longitudinal Fabry- Perot modes as bias current is slightly changed in sub-threshold operation, this is OK for F-P lasers but difficult to apply to DFBs (due to single-longitudinal mode operation) and is not applicable to VCSELs (due to large longitudinal mode spacing). CURRENT MODULATION (FM/AM) consists in direct current modulation of semiconductor laser, either Amplitude Modulation (AM) or Frequency Modulation (FM ) allow to see the chirp induced by -factor: CURRENT MODULATION (Fiber Transfer Function) consists in a periodic conversion AM  FM due to dispersive propagation in optical fiber. The Fiber Transfer Response (vs. fmod) shows minima and fitting gives  and fc. 𝜒 must be so that  guarantees the principle of causality. Applying an electric field to generate a polarization it must satisfy the ⅆ 𝐼𝑚(𝜒) Kramers-Krönig relation: 𝑅ⅇ(𝜒) =. ⅆ𝜒 For gas lasers, due to symmetry it is 0. In a semiconductor, before the peak there is no radiation and is transparent, after the peak photons towards the end do not find electrons and holes to be radiated so they are absorbed. The derivative is then not null on the peak and changes with the carrier density. 40 Multi-quantum well lasers, though they seem to solve this problem, present some non- idealities and have  = 0 only around the threshold. They however present advantages. In comparison: Bulk active material Multi-Quantum Well Quantum Dots active material Standard SLs Lower Ith Lower Ith, less sensitive to homogeneous gain Higher power temperature increase broadening Higher speed → larger Inhomogeneous gain Fabry-Perot lasers may modulation bandwidth broadening (due to scattering in operate on a single (due to lower non-linear dots size and spatial separation) longitudinal mode gain) Fabry-Perot lasers operate on 15-30 longitudinal modes low -Factor (?) 13. Effects of optical back reflections in semiconductor lasers (at a glance): Lang-Kobayashi equations; weak feedback regime; coherence collapse; chaotic regime; master-slave injection and locking Like many other oscillators, the operation of SLs can be strongly perturbed by the occurrence of Optical Feedback (Back-reflections) which causes linewidth broadening, instabilities, optical chaos (“coherence collapse”). This implies the need to protect the laser with an optical isolator. The Lang-Kobayashi equations are rate-equations (for complex field, not just photon number) that include the effects of the delayed feedback: where τ is the round trip of the external cavity and corresponds to the delayed feedback. These equations describe complex systems like the behavior of social groups and with SL practical solutions to these equations are observable. Feedback parameter (C) discriminates between different regimes, and it depends on the amount of feedback and the target distance. A is the round- trip optical power attenuation which in the ideal case is as large as possible since it reduces k. Longer s corresponds to smaller tolerated reflection. 41 Different regimes I: unperturbed II: Mode-hopping (oscillating frequency moves a bit) III: instabilities IV: “coherence collapse” (Chaos) V: stable regime (lasing on external mirror) Optical feedback can be exploited for sensing what happens outside of the cavity, for example with a self-mixing configuration: The final mirror is perfectly perpendicular to the beam, and it cannot have R=1 to avoid destroying the laser. The surface produces a variable attenuation and changing its position, the amount of feedback can be changed (intensity). The mirror moves back and forth to produce phase changes in the optical feedback and is usually made by attaching it to a loudspeaker moved by a sinusoidal signal. By applying a small signal, this configuration allows to measure what happens to the modes of the laser. Moving the target λ/2 → A goes from 1010 to 106. With moderate feedback we reach A=105 → Bistable that has three possible operating points, one unstable, since it tends to diverge and could take 100ps to transition and two stable points. This produces a jump each λ/2 so it can be used for laser sensing. In general, optical feedback is to be avoided, since it can perturb the normal operation of the laser, but, in some cases, controlled optical feedback can be exploited for useful purposes such as Chaotic encrypted optical communications and Interferometric measurements, the so-called Self-Mixing Interferometry. // Chaos = variables seem to evolve randomly but are actually evolving deterministically. 42  Optical injection MASTER laser is injected into SLAVE laser. Frequency is varied by changing the temperature and the slave is amplified but bounces back and forth by moving the frequency of the master, by the multiplying properties of the master laser this can be used for radio-frequency signals and studying the laser dynamics. Advanced semiconductor laser structures Conventional double-heterojunct