Student Scholars: 2017

Polymer optical waveguides are being studied as a possible solution to the increasing demand for higher interconnect bandwidth. In this work we aim to model, fabricate, and characterize a single-mode optical waveguide made from the inorganic-organic hybrid polymers OrmoCore and OrmoClad. These materials are of interest because of their low optical loss at data communication wavelengths and high thermal and mechanical stability. The waveguides are first simulated using Fimmwave to calculate the optimal dimensions and the optical loss from coupling a fiber to the waveguide. The waveguide is fabricated by spin coating and flood exposing a layer of OrmoClad on top of a silicon wafer. Then, a layer of OrmoCore is added on top of the OrmoClad and the waveguide
is pattered using the MLA150 (maskless aligner) and developed. Finally, the upper cladding layer of OrmoClad is spin coated on and flood exposed, finishing the waveguide. Once the waveguide is fabricated, a fiber is coupled to the waveguide. Light travels through the waveguide and is received by a lens which focuses the light on a power meter. The data from the power meter is used to characterize the optical loss of the waveguide. Finally, the losses from coupling the fiber to the waveguide is measured and compared to the simulated calculations to quantify the accuracy and reliability of this model.

Silicon photonic ring resonators have an inherent performance variation due to their manufacturing process, requiring heaters to trim resonators to their intended outputs. The parameters of ring resonators that are susceptible to manufacturing perturbation must be characterized to determine the impact of variation on performance and yield, and the amount of power required to trim. Identifying susceptible variables also enables robust design of resonators to inhibit performance variation. We investigate the manufacturing effects on geometric ring resonator parameters and develop design suggestions to improve robustness. Finite difference time domain software is used to simulate and characterize extremal variations of ring resonator parameters. Systematic sampling (corner analysis and latin square) is used to simulate the effects of photolithography and wafer thickness perturbations with minimal computational cost. The width and thickness of the resonator are found to be the key parameters that alter performance when perturbed. Accounting for these variations with a proof of concept design for manufacturability shows a 300% reduction in performance variation due to width perturbation. Robust design of ring resonators can be achieved by increasing the width of the resonator and decreasing its thickness with a partial etch, reducing the need for trimming.

The integration of optical circuit switches in data centers has shown great promise in providing the larger and more energy efficient data centers that we need. Optical circuit switches are composed of multiple copies of a subcomponent that control the direction that light travels. Thermally tunable optical rings-- wavelength selective filters that are reconfigured with temperature by passing current through a nearby resistor-- can be utilized as an optical switching subcomponent. Unfortunately, it is both time consumptive and costly to determine the optimal heater design through trial and error. I simulated different configurations of optical ring and heater systems in order to optimize both the heater design and its placement relative to the ring. Using photonic simulation software suite developed by Lumerical, I generated thermal profiles of the optical rings. These profiles were then imported into the optical simulation in order to determine the effects of temperature on optical behavior. Simulations showed those silicon optical rings designed in a racetrack configuration, with a bend radius of 10.5 um and a straight section length of 12 um, and a heater of doping concentration of 10 19 cm -3 had the widest thermal tuning range with the lowest power. As a result, the optical rings and heaters in the next fabrication run will be iterated around these parameters.

Successfully simulating photonics components on chips is advantageous for future iterations of design. I carried out a complete set of simulations on the passive components of our first generation education and workforce training integrated photonics chip set. The results of these simulations will be integrated into content for education in photonics skill sets for two-year and four-year college students. Using Lumerical, a finite-difference time-domain modeler, I generated scattering parameters from the designs of components on our chip set. These scattering parameters allowed us to model the response of ideal components in our circuits. Preliminary experimental measurements will help to shed light on additional losses in our ridge and strip waveguides as well as our input and output grating couplers (e.g. inter-spatial losses or absorption due to sidewall roughness). Once we are able to accurately incorporate real losses into our ideal simulations, we can simulate new passive devices to include on our second-generation chip set. This second-generation chip set will include similar passive devices as well as active devices such as PIN diode detectors and interferometers.

Germanium has many uses as both active and passive photonic devices, acting as a semiconductor with an indirect band gap of 0.6 eV, and absorbing at telecom wavelengths, but it is costly to deposit. Amorphous germanium is silicon CMOS compatible, can be deposited at room temperature, and is both cheaper and easier to deposit than epitaxial germanium. I studied the electrical, optical, and material properties of amorphous germanium with the ultimate goal to create photonic devices, such as fabricating waveguides. I used a sputter deposition system to deposit thin films of amorphous germanium on a variety of substrate materials, including CaF2, Si, and SiO2. I used a variety of methods and equipment to then measure the properties of the thin films, including the thickness, the IR spectrum, the Raman spectrum, the XRD, and more, with properties differing from crystalline germanium. Additionally, I attempted to passivate dangling bonds found within the germanium samples by annealing in a forming gas of 5% hydrogen in nitrogen which would help the amorphous germanium samples better act as a waveguide and other photonic devices. There was preliminary evidence for hydrogen retention by comparing the FTIR spectrum before and after annealing and observing a difference in transmission where a Ge-H absorption peak would be, and by observing a shift in the Raman spectrum similar to the shift of crystalline Ge to Ge-H.

Sample Grating Distributed Bragg Reflector lasers (SG-DBR) exhibit a very large tunable range, some capable of over 60nm. However, they receive large phase fluctuations from the various sections of the laser and hence are not narrow linewidth. One way of reducing the linewidth is to feed the laser into a wavelength filtering circuit and provide a negative feedback to the phase section of the laser. We modeled the laser linewidth as a function of the free spectral range and steepness of the filtering circuit and were able to demonstrate numerically that the linewidth can be reduced if both the free spectral range and the slope of the filer response were decreased. Finally, we propose a novel external cavity circuit, which utilizes Fano resonances in a high quality ring resonator fabricated in Silicon-on- Insulator instead of sinusoidal resonances of a Mach-Zehnder interferometer, which will narrow the linewidth, by an order of the magnitude. Thus, one can realize a widely tunable narrow linewidth light engine for telecom and datacom applications.

The world of optoelectronics has a wide range of applications remaining to be fulfilled. By characterizing optoelectronic devices and representing their behaviors with compact models, a process design kit (PDK) can be constructed. A PDK will unburden the circuit designer with the intricacy of the physics of each individual device and instead allow more attention to the optimization of the overall circuit performance. A key requirement for building a PDK is being able to characterize a large amount of devices (to provide statistical insight to the device performance) in a time efficient manner. During the span of this project, we developed and implemented an algorithm that controls an automated test setup that can measure a large number of devices in a single execution with minimal human intervention. We tested the algorithm on a series of test structures for a photonic device called the multimode interferometer (MMI). The measurement results are reproducible and consistent with the device design. Another measurement that we performed was the determination of transparency current in epitaxially grown quantum dot lasers using the same test setup. We were able to convert the setup for different testing purposes in a timely manner and obtain valid data. These devices, after careful analysis and modeling, can be added to the PDK for future circuit design. With the PDK, users can build their circuits before fabrication in a clean room. This is essential in providing a straightforward and innovative approach to building photonics integrated circuits.

In today’s age, the Internet is a vital source of everyday communication. However, communication in data centers is extremely inefficient due to switching between optical and electrical power. An all- optical photonic switch is a solution to this problem. These switches require heaters to tune to an ideal wavelength for the system to operate efficiently. In my project, we explored what heater structures would be optimal in order to fulfill this task. For the optical switch to perform at its optimal efficiency, the heaters must tune to the ideal wavelength, which can be done through experimentation. I have tested heaters within a variety of lengths and widths to calculate the resistance. Ohm’s Law was used to calculate the resistance when plotting current vs. voltage. Optical fiber alignment is a method used to introduce light to the chip. Aligning fiber well minimizes loss to the chip. Data is then recorded along a broad range of wavelengths. I found that the smaller the heater was, the more resistance it generated. The amount of power was calculated from these measurements. Finally, I have determined which heater geometries yield the most efficient heaters.

Optoelectronic frequency synthesizers currently function in a variety of applications from telecommunications to metrology to 3D-imaging. However current synthesizers are room-scale setups with limited frequency ranges tuned for specific applications. The overarching goal of this project is to create a chip-level optoelectronic frequency synthesizer with a wide output frequency range, enabling far more potential applications in a low-power, miniature format.  My primary task this summer was to program a digital chip which interfaces between the custom electronics and a computer software used to control the laser.  By adapting the design from a previous chip, this new digital chip implements the same functionality in a smaller size, weight, and power.  Using various resources such as existing design files, example templates and device datasheets, I managed to modify each file in the Verilog language to be compatible with the new chip. In addition, I helped define a custom protocol between the chip and the PC, allowing flow of communication from the optical drivers and sensors all the way to the computer software.  The new chip is now fully functional for integration into the next generation of the system.

Lasers are typically grown with III-V materials because of the direct bandgap for lasing capabilities. III-V substrates, however, tend to be very expensive and are unavailable in large sizes (>4 inch diameter). A more commercially viable alternative is growing the lasers on a silicon wafer. Since silicon does not have a direct bandgap for lasing, a template of III-V material is epitaxially grown on top to form the laser. The problem with III-V growth on Si is that the crystalline mismatch between the III-V substrate and the silicon causes defects. These defects, which predominantly include threading dislocations, diminish the performance of the laser. In order for the silicon-based lasers to perform comparably to more costly III-V lasers, the number of threading dislocations should be kept to a minimum. Quantum dots consisting of InAs grown on III-V/Si have been shown to decrease the impact of of threading dislocations within a laser by locally confining charge carriers away from the dislocations. This, in turn, brings these lasers closer to the performance observed in equivalent III-V structures. Similarly structured GaAs lasers and GaAs/Si lasers with InAs quantum dots were analyzed using a modular integrating sphere and an optical spectrum analyzer to determine the light-current- voltage characteristics and the gain parameters, respectively. The quantum dot devices performed within the same range as the GaAs devices.

Switches are used in data centers to transfer large amounts of information, but they often consume huge amounts of energy. Wavelength selective switches are energy efficient while still offering high data transfer rates. To create a wavelength selective switch, filters are used to select for particular wavelengths. The type of filters used affect the functionality of the switch. I examined how contra-directional couplers can be used as filters. These contra-directional couplers consist of two asymmetric waveguides connected by a grating. The grating is a periodic perturbation of the waveguides’ refractive index and is used to reflect certain wavelengths while transmitting others. I’m conducting an experiment on a silicon photonic chip containing contradirectional couplers with waveguides of various dimensions. By supplying light of various wavelengths and measuring optical power, the wavelength spectrum for the filters can be constructed. The purpose is to analyze how different physical parameters will affect the resulting wavelength spectrum of the contradirectional couplers.