Theses 2019

Abstract: In high power thin disk lasers the achievable inversion is limited by ASE. In this thesis, a simulation model is presented in order to investigate the effect of ASE. Special attention is paid to spatial distribution of the upper laser level population. The developed simulation model is based on a Monte-Carlo ray tracing approach to solve the radiation transport within the thin disk. In order to get location depended and time resolved amplification and absorption of the radiation the disk is divided up in cells. In each cell the laser rate equations are solved. First, the evolution of the spatial distribution of the population inversion is examined with given initial distributions. It could be shown that ASE has a stronger influence on the edge of the exited area than in the center due to the longer average path lengths. This distorts the population inversion profile noticeably and grows with increasing inversion. The same effect could be also shown if the pump effect is taken into account. To benchmark the simulation, the results are compared to experimental data. Especially the maximum reachable single pass gain is in good agreement, but also the temporal evolution of the gain shows great similarities.

Within this thesis project, a simulation has been made of the measurement system for ocular wavefront aberrations that is used within the diagnostic device InnovEyes Sightmap, developed by WaveLight GmbH. The InnovEyes Sightmap can perform multiple measurements on the eye, including but not limited to, the tomography of the cornea and the wavefront aberrations of the eye. Measurements with InnovEyes Sightmap are made and compared to the simulations made with Zemax.

One measurement and simulation measures only astigmatic eyes from -6D up to +6D, while the other takes a combination of negative cylinder values with varying degrees of defocus. These measurements were made with a reference eye; no real eyes were used for any of the measurements.
The results that were sought after are spherical errors, cylinder errors, and SEQ values. To get these results from the simulation, the Zernike polynomials are calculated from which the sphere, cylinder, and SEQ values can be derived. These are then compared to the theoretical (expected) values and should have a maximum deviation of 0.16D as required by WaveLight GmbH.

The results that InnovEyes Sightmap gives for these measurements show that while it performs within the requirements for spherical errors (defocus), the cylinder values are not within the requirements. The results also show that the simulation made in Zemax needs to be improved as the cylinder values gained from the simulations are also not within the requirements of WaveLight GmbH.

Brain shift is one of the critical challenges that faces neurosurgeons during complex operation. This phenomenon caused by many factors such as gravity, tissue manipulation and tumor size and it affects the accuracy of neuronavigation during tumor resection. Stereo vision is one of Image-Guided Neurosurgical Systems that used in calculating the deformation of the cortical brain surface. In this thesis, two stereo vision modalities, RGB and infrared, are directly connected to a neurosurgical microscope. The calibration of these modalities is challenging due to multifocal lens of the microscope. Therefore, single and stereo camera calibration are performed as well as dynamic zoom calibration. Each calibrated modality can estimate the depth of an object locates within the microscope field of view. These depth measurements are validated for different calibration methods by using an object that has a pre-defined 3D structure. The maximum error in relative depth measurements of that 3D structure is 4 mm while the minimum error is below 1 mm. Additionally, a depth map of brain phantom is calculated before and after deformation. Both modalities were able of showing and estimating the direction of that deformation.

This thesis proved for the first time the feasibility of using ultra-fast avalanche diode array (APD) camera for remote photoacoustic detection using speckle sensing. For the proof of concept, a photoacoustic generation and detection system were built. PVCP phantoms and porcine fat ex vivo samples were produced and used for the photoacoustic measurements. All the photoacoustic measurements were conducted in transmission mode where the photoacoustic signal generation is achieved on one surface of the sample and the detection is done on the other surface.

Experimentally, the photoacoustic signal was generated using a single short laser pulse that hits the sample absorber and travel to reach the second surface causing a shift in the speckle pattern. The speckle pattern was produced on the other sample surface with a CW laser and the shift was then detected using the APD camera.

Based on the phantoms geometry and the speed of sound in its material, theoretical times of travel for photoacoustic signals were calculated and compared to the experimental values extracted from speckle pattern shift analysis. Verification measurements were also conducted using a piezoelectric contact transducer to further prove the analysis results.

In conclusion, this work proved the suitability of using ultra-fast diode-array detector for photoacoustic signals detection with an accepted precision and with high sampling frequency up to 4 MHz. It proved also feasible to work on ex vivo tissue with sampling frequency up to 1 MHz.

Since the laser was invented, it has been found in many applications. The laser is a very useful tool for a wide variety of clinical diagnostic and therapeutic procedures. In ophthalmology, the laser is involved into solving problems with both anterior segments and posterior segments of the eye.

Currently, a laser procedure called LASIK (Laser – Assisted In Situ Keratomileusis) is the most popular refractive surgery in the laser vision correction. Like other types of refractive surgery, the LASIK procedure reshapes the cornea to enable light to be properly focused onto the retina. Before changing the cornea curvature, a cut of the first corneal layer should be performed. Creation of the, so called, Flap, which is lifted after this cut enables the Excimer laser radiation applied directly onto the stroma. In LASIK, originally, the Flap is made using a microkeratome.

Rapid development of new laser technologies enabled the application of ultra-short lasers in refractive surgery. Focused ultra-short laser pulses in near-infrared spectral range can generate laser induced breakdown (LIB) in the cornea, which will disrupt the tissue. Refractive surgery which allows reshaping of the cornea without the mechanical impact but by using femtosecond (FS) pulses is called FS-LASIK. The cutting depth and position can be established by varying the laser focus position.

The cutting process remains an area for development. Improving the efficiency, precision, and safety continues to present a challenge and requires better knowledge of the cutting mechanisms.

This thesis describes a way of improvement of the Flap performed by Wavelight® FS200 by reshaping of the beam profile from a simple Gaussian to a donut-shaped beam.

In the present thesis an experimental setup for two-photon polymerization consisting of an optical and a mechanical part has been developed, constructed and programmed. Furthermore, first specimens have been generated in order to estimate its capabilities regarding the fabrication of intraocular lenses. The superior aim is to enable the fabrication of intraocular lenses (IOLs) by means of 2PP. This would allow the entirely free customization of IOLs which are globally used for the treatment of cataract.

The constructed optical setup provides sufficiently high photon densities for two-photon polymerization within the sample at an appropriate wavelength. At this, the behavior of power values at certain positions which are important for the qualitative as well as quantitative understanding and description of the optical setup has been studied.

The investigation of the mechanical setup has shown that for structure sizes up to 100 μm in x- and y- direction the deviations in z-direction introduced by the setup itself lie below 2 μm. Depending on the requirements of the requested specimens, these deviations can be neglected. However, since state-of-the-art IOLs are fabricated with surface roughness values in the range of 10 nm, the deviations in z-direction introduced by the mechanical setup itself must be considered in a future setup. This can either be done by implementing a correction algorithm within the control software or adapting the mechanical components. Most probably, the required roughness values can only be reached by properly applying both solutions. However, when adjusting deviations in z-direction it is important to keep the derived correction factor CDOF in mind.

Furthermore, the adaption of the control software should also consider the acceleration and deceleration distances of the axes. It has shown that they may lead to agglomerations of polymerized structures at positions of reduced velocity due to acceleration or deceleration. Although the subsequent development with a solvent consisting of isopropyl and methyl isobutyl ketone in most cases detaches these agglomerates, they still might parasitically influence the fabrication of multi- layer specimens. An easy way to consider the acceleration and deceleration distances is to implement delay-times within the control software. Therefore, the kinematics of the axes should be investigated. Moreover, in a future control software the processing of three-dimensional data of the desired structure should be implemented.

The fabricated specimens qualitatively feature a sufficiently high cross-linking level by applying a target overlap between the pulses of TOL_pulse  = 0.75. However, due to the fact that the cross-linking level is a crucial value for the fabrication of IOLs by means of 2PP, a quantitative analysis at a later stage of the research project should be conducted. Therefore, μ-Raman spectroscopy is a suitable measurement method.

The speed of the generation of structures by 2PP is scalable. Since the applied NKT Origami-10 XP femtosecond laser can provide pulse repetition rates up to 1000 kHz, axis velocities of more than 100 mm/s are possible with the utilized objective. This allows a substantially faster fabrication process and therefore might be interesting for the fast generation of personalized IOLs in order to improve the patient care. A further requirement for this superior aim is the usage of a specially synthesized biocompatible monomer solution which is applicable for 2PP. In fact, such a monomer solution could be tested in the final stage of the present thesis. It has shown that the knowledge gained from the experimental work with the current setup applying the commercially available Ormocomp can be transferred to the specially synthesized solution. Concluding, it can be deduced that the experimental setup which has been developed, constructed and programmed within the present thesis, generates surface roughness values in the range of comparable state-of-the-art manufacturing procedures. Moreover, the setup possesses the capability of being further optimized as explained above finally resulting in a new method for the fabrication of intraocular lenses by two-photon polymerization.

Abstract: Quantum mechanics holds the promise of a second technological revolution, one driven by a special type of quantum correlations not present in the classical framework, denoted by entanglement. Optical parametric oscillators are a type of nonlinear optical devices that constitute the most reliable source of entangled light. As such, they are of great importance for the fields of quantum information, quantum optics, and quantum technologies. This thesis focuses on the quantum theoretical description of a special type of optical parametric oscillators, dubbed actively-phase-locked, which have the distinctive property of undergoing limit-cycle motion in a certain parameter regime. In regions where they reach stationary behavior, the quantum properties of these devices were studied through the usual linearization technique, which assumes that quantum fluctuations around the classical states are small, a generally good approximation for nonlinear optical cavities. In contrast, the nontrivial periodic time dynamics associated to limit-cycle motion offers a big challenge, because the fluctuations cannot be considered small along the closed trajectory drawn by the cycle in phase space. In this thesis we apply a recently developed technique that extends the linearized approach to such situations. The technique allows us to approximate the asymptotic quantum state by a mixture of Gaussian states centered at the points of the cycle’s trajectory. Analyzing in detail the quantum correlations of these Gaussian states, we show that actively-phase-locked optical parametric oscillators produce light within the cavity with large entanglement levels, even in the regions where they undergo limit-cycle motion. Therefore, the results found in this thesis pave the way for an experimental observation of entanglement in such a dynamically nontrivial scenario.“

Abstract: Laser-based production processes have become established in a wide variety of industrial sectors. In many applications, the targeted adjustment of the intensity distribution in the working plane can increase the productivity and/or the machining quality. While static, homogeneous distributions have already been used successfully employed in many areas, more complex, application adapted and possibly dynamically adaptable intensity distributions are often required.

However, with the traditional approach to dynamic beam shaping, the number of degrees of freedom required for generation of complex intensity distributions (approximately 105) conflicting with the laser powers typically required for material processing (100Wto several kW): High-resolution technologies such as Liquid Crystal on Silicon (LCoS) and Digital Micromirror Devices (DMD) have several million degrees of freedom, but their damage thresholds are only in the one to two-digit W~cm2 range.

A new approach to high-performance dynamic beam forming is to be investigated: a laser beam is first pre-formed at low power (1 W) with an LCoS and then amplified. The influences of the amplification are to be simulated and taken into account during the beam shaping so that the desired distribution results are obtained after the amplifier. In preliminary work, the beam shaping process was simulated by near field beam propagation
method and validated with LCoS in the experimental setup. Based on these results, the influence of the gain on the beam forming process will be examined in more detail in this master thesis. An approach is to be identified with the aid of OPT, to compensate the influence of the amplification during the beam shaping.

Conclusions and outlook: In this work, experiments were carried out with silica beads of different sizes (radius a =25…71.5 nm) using a radially polarized optical tweezer implemented by a deep parabolic mirror (PM). By controlling the pressure, one gained control over how strong the motion of the trapped particle is coupled to the thermal gas in the chamber. Power spectral density (PSD) signals were calculated from the intensity variation of the trapping light and light which was scattered of the trapped particle. Corresponding functions from the Langevin theory were fitted to measured PSD. Sizes of the trapped particles were estimated from the measured damping rates using the kinetic theory. Trap stiffness and the ratio between the radial and axial stiffness parts were determined more reliably via the motion of the particle using various methods e.g. when the trapping power was varied at low pressure. All the trapped particles presented in this work were lost, when the vacuum chamber was evacuated on the order of 1 mbar. Heating was suspected to be the cause of this as the minimum pressure that can be reached was lower when the trapping power was decreased (cf. [1]). But on rare occasions, we were able to trap particles down to 20*10^(-3) mbar, limited by the pump that was in use.

It was shown using two different detection techniques that one can decompose the motion of the particle in harmonic trap. It was discovered that the individual eigenaxes of motion in radial direction were not strictly ortogonal to each other. We were able to track full translational dynamics of the trapped particle in 3D using the calculated calibration coefficients for each detector.

There were also preliminary measurements made with the same samples with varying pressure and trapping power as well, when the radially polarized trapping beam was switched to azimuthal polarization, for the first time to our knowledge. These need to be processed further and analyzed within a complete theoretical framework. Moreover, at this current stage of the experiment, it is possible to determine diffusion coefficients. One of the possible future goals of the experiment would be implementing feedback cooling based via the three detection setups built during this thesis which should allow realizing experiments in the quantum regime. This would most likely require also upgrading the vacuum pump to a high vacuum one involving a turbo pump.

The end goal remains to cool a nanoparticle to its ground state (which was not shown up to this date) for which one can surely benefit from the properties of a tweezer with a „optimal mode converter“-PM mirror.

Conclusion and summary: We presented two different methods for automatic segmentation of the glottis from LHSVs. Both methods, especially the UNet, show promising results that are comparable with the results of clinical software (GAT). The first method uses the combination of the contrast stretching and denoising in order to address some famous problems of vocal fold endoscopy recordings. By using the OTSU thresholding, the glottis area is separated from the rest of the video frames. Then the algorithm labels the glottis using the connected components method . Finally, we use the nearest centroid classifier in order to segment glottis. We evaluated the results of this method by conducting 4 experiments using different images and the algorithm could get the Dice score of 69.30%.

Although this method shows good results, there is space for improvement. We used the GAT software for finding the ROI and bounding box. But it can be achieved using the proper filter to separate the edges of the vocal folds from the other part of the image. Another improvement could be using the motion compensation method such as tracking of the glottis throughout the frames using the first frame.

The UNet method has been chosen because it proves its efficiency in medical image segmentation applications. Although we used a very small dataset, the results show a high-level efficiency of the UNet, especially in comparison to the first method. Our dataset consists of totally 480 images from 10 different LHSVs recordings. The UNet model was created using the Keras API that running on top of TensorFlow. Because of the small dataset, we used real-time data augmentation during training for both testing and validation step which adds many images to the dataset. For evaluation, we conducted 2 different test scenarios, one dependent and one independent to the videos of the training data. We could get the Dice score of 87.23 which shows a significant improvement over the previous method and the related work in [Lav19] as well.

The next step is definitely to train the UNet using the large dataset to improve the prediction result of the UNet. The next step is definitely to train the UNet using the large dataset to improve the prediction result of the UNet. Also, to compare the results of both methods with the different segmentation results in order to achieve a better evaluation of the algorithms especially the UNet.