Theses 2021

Data Driven Compuational Display: Computational 2D Display Using the Hogel Basis Screen

In the present thesis we analyze some aspects of the novel Hogel Basis Screen experiment. The ultimate goal is to produce a screen that can display a 3D holograhic video of a moving object.

The screen contains volume holograms with different ‚bases‘ images stored on them. The project’s main idea is that a linear combination of the bases will produce the display of the target images.

First, we study the optimization problem to determine the optimal bases that we should record. Later, we perform the wave propagation with numerical methods to simulate the experiment. Finaly, we develop a procedure to print these volume hologram with a nano-precision 3D printer.

Modeling of Grayscale Lithography and Calibration with Experimental Data for Blazed Gratings

Optical lithography facilitates the printing of smallest possible micropatterns on a Silicon wafer using visible or ultraviolet light. Grayscale lithography is one such genus of optical lithography dedicated towards the fabrication of three-dimensional microoptical structures. It is envisioned as an economical and high throughput lithography technique to revolutionize many lithographic advancements around the corner.

This master’s thesis explores grayscale lithography in a research and development lithography simulation software – Dr.LiTHO. It involves the modeling of a thin Kirchhoff based grayscale mask in Dr.LiTHO to emulate the grayscale exposure of blazed gratings generated from an LED based micro-image stepper operating at an exposure wavelength of 405nm and numerical aperture of 0.55. The resulting photoresist profiles of blazed gratings are obtained with a standard model for Diazonaphthoquinone (DNQ) photoresists. The objective of this master’s thesis is to develop a photoresist model calibrated with experimental data to simulate the fabrication of blazed gratings. The calibration of simulated and experimental profile data of blazed gratings is performed using Dr.LiTHO’s inbuilt optimizer – Pythmea. The Pythmea optimizer requires a merit function to measure the difference between the simulated and experimental profiles. A merit function named ‘areaFit’ is thereby defined in Pythmea to measure the difference in areas of both the experimental and simulated blazed grating profiles. The areaFit is then minimized by varying unknown model parameters having an impact on the simulated photoresist profile. The impactful model parameters along with their range of values are devised by executing a sensitivity analysis of Pythmea data using the methods of multivariate analysis, Latin Hypercube sampling and correlation analysis. A good calibration result has been conclusively accomplished post adapting the experimental exposure dose distribution. A further improvement in the calibration model has been achieved by implementing first steps towards a three-dimensional optical proximity correction (3D-OPC). This 3D-OPC technique and a Fourier transform method have been integrated with the Pythmea optimizer to increase the efficiency of simulated blazed gratings. The extendibility of the developed photoresist model to a larger range of blazed grating pitches and grayscale exposure settings for another microstructure, the microlens, is discussed.

This master’s thesis marks a successful development of a photoresist model calibrated with experimental data of blazed gratings generated using grayscale lithography in Dr.LiTHO. A crucial role of sensitivity analysis towards the development of a robust and time efficient photoresist model has been noted. It has been finally discussed how the introduction of 3D-OPC paves the way for many interesting applications in the domain of grayscale lithography.

Polarization of EUV Light via Multilayer Design

Several studies have been performed to investigate polarization effects in multilayer systems for extreme ultraviolet (EUV) wavelength. The impact of polarization was observed in EUV imaging simulations for high NA lithography. It is shown that polarized illumination is essential for improving the local contrast of images or NILS (normalized intensity log slope). The work investigates the possibilities to polarize EUV light by optimized multilayers. This characterization and simulation of multilayer structures has been performed using Dr.Litho.

A typical multilayer for EUV systems consist of 40 bilayers alternating of molybdenium (Mo) and silicon (Si). The reflectivity of the multilayer depends on the thickness of the individual layers in the stack, on the direction and polarization of the incident light.

The most efficient multilayer polarizers operate close to Brewster angle, where the reflectivity for TM polarized light (R_TM) is virtually zero. A mulitobjective optimization algorithm was used to identify the suitable multilayer configuration maximizing reflectivity of TE polarized light (R_TH) and fraction of polarization. Fraction of polarization (FoP) was calculated as the ration between (R_TH-R_TM)/(R_TE+R_TM) to obtain the appropriate multilayer with changing thickness and number of layers. The multilayer structure is optimized to have the highest reflectivity of THE polarized light and fraction of polarization at Brewster angle. It was found that MoSi multilayer can achieve 99.9% fraction of polarization by optimizing the thickness of Si and Mo. The Pareto Front demonstrate the tradeoff between high reflectivity and high fraction of polarization. Each point on the Pareto represents a combination of multilayer parameters that provide the indicated performance.

In reality, a multilayer polarizer has to operate over certain ranges of incident angle and/or wavelength ranges. Therefore, ranges of wavelength (13-14 nm) and incidence angles (37°- 47° ) were optimized. Additional simulations to investigate the impact of different optinos in the design of the multilayer (e.g., constant vs variable bilayer thickness) and materials (e.g., RuSi vs MoSi multilayers) on the achievable performance are included. The impact of intermixing layers, which result from diffusion between Mo and Si, is investigated. In addition, MoSi multilayer is investigated to operate as a quarter wave plate. This may be useful to convert linearly polarized light to circularly polarized light.

Studying Non-linear Effects in Tapered Optical Fibers

Tapered step-index solid core optical fibers are an interesting platform for the genertion of third harmonic (THG). They present a promising medium for the generation of photon triplets, the inverse process from THG. In this work I report the successful fabrication of submicron tapers via the sweeping-flame technique. I aim to generate third harmonic in the mode H E₁₂ at 532 nm. This is the mode with the highest nonlinear spatial overlap with the fundamental pump mode H E₁₁ that allows phase-matching between the two. The laser used for this experiment was a femtosecond Erbium-doped fiber laser centered at 1550 nm with a pulse duration of t_{FW H M} = 66 fs. The heat source used for the tapering was aplasma ring, a cleaner source than the widely emploxed oxybutane flame, and the fabrication was performed in low vacuum. The first tapers fabricated had linear transitions, which the machin could fabricte with an in-built program. A special technique to take the fibers out of the machine to study them without damaging them was implemented. Once the correct functioning of the machine was checked, tapers with custom adiabatic transition regions were fabricated, in order to minimize the losses. The transitions were designed by applying the adiabatic criterion deines by Love and implemented in the machine via user-generated tables. Given the time resolution of the machine, the walk-off of the plasma ring during the tapering process had to be corrected. Once all the fabrication problems were solved, adiabatic tapers that generated only the mode HE₁₂ were successfully manufactured with good reproducibility.

Quality Control of Glass Coating: Quantification of Residual Stress using Raman Spectroscopy

Research question, objective and structure of the paper: Effects of residual stress on carbon films  or silicon film, with Raman spectroscopy over several GPa range has been extensively discussed with deconvoluted peaks or no deconvoluted bands. This paper aims to develop a method to quantify the residual stress distribution of double glass system in elastic stress range using Raman spectroscopy backscattering geometry. This approach is based on Raman shift-stress calibration with uniaxial load frame equipped piezoelectric actuator and spectra components score variation.

A double glass system is a slice of thin glass plate being coated on substrate with heat treated in furnace. E.g. D263 borosilicate slice on aluminosilicate glass.

The investigated spectra range is Brillouin shift, and Raman shift band 250-610 cm^-1.

A mathematical analysis model from Zhang (2007) is used to calculate the stress distribution in single layer glass-coating system with different thermal expansion coefficient and compared with calibrated stress afterwards to give a hingt whether Raman calibrated method works well or not.

Section 2 clarifies the theoretical foundations, such as the cause of residual stress inside double glass system, the model mentioned above, brief introduction to Raman spectroscopy, uniaxial stress, and glass structure.

Section 3 is methodology, from specimen preparation procedure to data collection and analysis. Illustrations of equipment is also there to give an explanation how do they work. And the operationalization of variables is explained in details with figures.

Section 4 is the result and discussion, influence of uniaxial stress on Raman and Brillouin spectra of double glass, and the calibrated stress with BS and RS are discussed and compared with the calculated residual stress inside glass-coating system with the same physical characteristics as double glass material used in calibration.

Section 5 is contribution to practical and theoretical knowledge.

Section 6 lists the contributions and limitations of this work.

In Appendix, information about specimens, fitted curve parameters of all cylinders, characteristic bands for silicate glass, borosilicate glass, aluminosilicate glass, lithium silicate glass found in literatures are listed, and last the curve fitting methods Gaussian and Lorentzian are compared.

Cooling in Optomechanics

The aim of this thesis is to analyse and propose a few strategies for optomechanical cooling especially at the level of simultaneous refrigeration of many mechanical resonances. We extend known techniques such as cavity self-cooling and feedback cold-damping to the multimode case and propose a novel approach for cooling any generic resonator, which we dub, phase adaptive parametric cooling, in which the potential energy of the oscillator is modulated at twice its natural resonance frequency.
For the self-cooling technique, we propose an approach based on collective addressing of many mechanical resonances via a frequency comb source. In the cold-damping case, we show that a single feedback loop suffices to cool many modes even in the presence of feedback time delay.

Ultrashort Pulse Generation and Peak Power Scaling Systematic Study in Yb:YAG Bulk Oscillators with Separated Kerr-lens and gain medium

In this project a Kerr-lens mode-locked (KLM) Yb:YAG bulk oscillator with separated gain and Kerr media was developed. The objective was to investigate peak power scaling
mechanism in a femtosecond ( ) bulk oscillator. In 10 a standard KLM oscillator with a − 15 gain medium simultaneously serving as a Kerr medium, the average power is restricted by
balancing the mode matching in the gain medium and tight focusing in the Kerr medium. The tight focusing is necessary to obtain a strong Kerr- lens effect.
By applying different concepts in three stages the peak power of the oscillator is increased:

• The mode sized inside gain and Kerr medium are independently controlled and adapted to achieve mode matching in the gain and simultaneously tight focusing in the
  Kerr medium.
• Controlling the Kerr- lens strength by choice of different Kerr media (different nonlinear refractive index) and mode size inside the Kerr medium.
• Significantly decreasing the repetition rate of the oscillator by increasing the cavity length via a 4 f extension.

An increase of the pulse energy from 10 nJ to 56 nJ was obtained directly from a diode-pumped Yb:YAG bulk oscillator without external amplification. A repetition rate of
10.7 MHz was obtained by using a 12 m long 4 f telescope that extends the cavity length to a total length of 14 m. The oscillator delivers in that configuration 600 mW of average power
with nearly Fourier-transform-limited 94 fs long pulses with a bandwidth of 12 nm at a center wavelength of 1030 nm. Peak power scaling from 97 kW to 524 kW was achieved with the
above three mentioned mechanisms. To the best of my knowledge, this is the first systematic study of this peak power scaling concept in bulk femtosecond oscillators.

Interactions Between Photoreceptor-driven Signals in the Electroretinogram (ERG)

The delay change by the stimulus size can cause a dependent change in th cone processing. Also, the chromatic and the luminance phases decreased with increasing the stimulus size.

Luminance stimuli are common to be used in standard flash ERGs. This causes that ERGs show only luminance mechanisms. Also, flashes can decrease the informative value of ERG because the energy in flashes I compact in a short time and can change the operation mode of retina. Thus, continuous stimuli can bring more information about retinal pathways rather than flashes. In this thesis, continuous stimuli shat contained red and green modulation was used. ERGs were recorded at different fraction values and they provided information about chromatic and luminance mechanisms.

Finally, the results of the experiments in this thesis provided the knowledge of retinal processing.

Controlling Temporal Jitter in Direct Electric-Fields Sampling

Direct measurement of the electric field of an ultrafast laser pulse is commonly used for terahertz time-domain spectroscopy (THz-TDS). Electro-optic sampling is an electric-field measurement technique, which is successfully employed in the terahertz (THz) to mid-infrared (MIR). Since the electric field in THz and MIR has a smaller diffraction limit, in compare to the near-infrared region, therefore extending the electric-field measurement to the NIR region has high importance. The focus of this thesis is to build an electric-field sampling technique, which extends the electric-field measurement to the NIR region. The field sampling setup relies on a 10 fs NIR phase-stable field to be sampled that is coincident with a 5 fs sampling pulse in a nonlinear crystal, in which the sum-frequency of the sampling pulse and the NIR pulse is generated. The nonlinear interaction of the sampling pulse and the NIR pulse induces a polarization rotation of the sampling pulse that is sensitive to the strength and sign of the NIR pulse electric field. The interference of the sum-frequency signal and the sampling pulse is a delay-dependent signal, which is directly proportional to the polarization rotation of the sampling pulse. Therefore, by reading out the polarization rotation signal, the electric field of the NIR pulse can be reconstructed.

Domain Adaptation for Stereo Depth Estimatoin Using Deep Neural Network (DNN)

The work aims to improve the generalization ability of depth estimating neural networks on novel domains. Neural networks can be trained on real-world images but the ground truth depth labels are expensive to obtain. Alternatively, neural networks can also be trained on economically generated synthetic images and their ground truth labels. But real-world images differ from synthetic images in terms of details, texture, and colors etc.. Therefore, we observe a reduction in performance of neural networks trained on synthetic images and images accounts for this reduction in performance.

In this thesis, we present a novel way to increase the performance of a stereo depth estimating network over novel domains. The proposed technique use a carefully configured and trained cycleGAN to translate the style of real-world images to synthetic. This allows for the translation of any given domain to synthetic domain without requiring ground truth labels and consequentially increasing the performance of a given neural network. This work also performs an extensive ablation study to investigate the alternate approaches to address the problem.

Surface-Generated Polarization-Entangled Photons

In this work, the polarization-entangled photon pairs from ultrathin sources are studied for the first time. Moreover, although SPDC from ultrathin sources has been already reported, this is the first time an extremely thin source such as a surface is successfully used for the SPDC experiments.

First, the polarization dependence of SHG in a subwavelength layer of (100)GaAs is measured . The results cannot be explained by the bulk nonlinearity of GaAs but coincide with the predictions of SHG originating from the surface nonlinearity of the crystal. Thus, SHG observed in this work is caused by surface nonlinearity. Moreover, from the form of the surface χ⁽²⁾ it follows that type-II SHG is observed, when two pump photons with orthogonal polarizations merge into one photon.

Second, the process of SHG is inverted in the SPDC experiments. The polarization state of photon pairs generated in (100)-GaAs is reconstructed with the polarization tomography technique. From the reconstructed state, the conclusion follows that the detected photon pairs are mainly orthogonally polarized photons, which points at type-II SPDC. This is possible only if the process is brought about by surface nonlinearity. Thus, surface SPDC is reported in this work for the first time.

Finally, the polarization entanglement of the dctected photon pairs is shown. The degree of entanglement is high but not maximal due to the non-zero amplitudes C1 and C3 of the qutrit state. Nevertheless, the polarization dependence of the coincidence rate and the polarization Hong-Ou-Mandel effect are observed, with the high visibility of the two-photon interference. Polarization-entangled photon pairs from ultrathin sourccs have never been reported before.

Although in this work only degenerate SPDC is observed, in principle, the spectrum of the photon pairs is almost unrestricted due to the ultra-small thickness of the source. Moreover, photon pairs are expected tobe highly entangled in frequency, and momentum. The polarization entanglement adds one more degree of freedom, increasing the quantum information capacity. The polarization entanglement can be used for quantum key distribution or for establishing ternary logic in quantum computation. Decreasing the sizes of the sources of entangled photons makes it easier to implement them in integrated photonic circuits.

The efficiency of photon pair generation can be further increased by utilizing geometric resonances. Recently, the enhancement of the photon pair generation using Mie-like resonances in lithium-niobate metasurfaces was reported. Implementing a similar approach in metasurfaces made from (100)-GaAs can significantly increase the generation efficiency, making it an attractive platform for quantum technological applications.

Single-shot characterization of ultrashort near-infrared pulses using space-time duality

Single-shot optical waveform characterization based on space-time duality has opened a new gate for the characterization of transient and unrepeatable phenomena. However, the resolution of the monitoring system is determined by the bandwidth. In contrast to electro-optical systems, which are limited by the bandwidth of their electrical component, the all-optical system can provide high bandwidth in the range of THz. Silicon-based devices based on space-time duality has demonstrated a resolution of 220 fs over the range exceeding 100 ps [1].

In this project, a free space temporal single-lens structure has been theoretically studied and simulated. In this structure, the single-lens structure is decomposed into two consecutive parts, a 2f system, and a Fraunhofer dispersion. The 2f system maps the temporal profile of the pulse into the spectrum of its output, then the spectrum of the pulse is measured by a Free Space Angular Chirped Enhanced Delay (FACED) system. This study offers the potential of developing a device with a resolution of a few fs capable of characterization of optical pulse waveforms with a repetition rate in the range of MHz.