In 2015, the space-time distortion caused by GW150914 was found - a pivotal event that inaugurated the era of interferometric gravitational wave astronomy. As of today, gravitational wave observations are routinely made with proper sky localization by the world-wide operating detector network of the second generation. The implementation of cryogenic cooling can reduce the coating thermal noise in the next detector generation. In this case, optics made of fused silica are not suitable because of fused silica's large mechanical loss at low temperatures. Crystalline silicon is an alternative material but not transparent at 1064nm; therefore, other laser wavelengths, e.g. 1.5um, must be used. Single-frequency EYDFAs based on LMA fibers can deliver the required output power at 1.5um. A PM setup, however, has not been demonstrated on the desired ~100W power level so far; also, there has been no demonstration of any successful longterm operation (> hours) even of a non-PM setup. In this work, a prototype amplifier with PM fibers is presented on a laboratory- and advanced engineering-level. A numerical FEM analysis of the pump wavelength dependence of the Yb3+ ASE and non-linear SBS has been performed; off-peak pumping was found to suppress the unwanted Yb3+ ASE considerably. The achievable output power at 1.5um was limited by the Yb3+ ASE if the simulated amplifier was pumped from 880nm to 990nm; the onset of the Yb3+ ASE was linked to a deterioration of the Yb3+-to-Er3+ energy transfer. The simulated amplifier was limited by SBS if pumped at wavelengths shorter than 880nm or longer than 990nm. The power threshold was approximatable by adapting a well-known threshold approximation for passive fibers. Uncontrolled gain, e.g. resulting from a seed laser failure, must be prevented by interlocking the pumping process. In this work, the required reaction time has been studied with single-mode fibers by a combined experimental and numerical approach. It was found that a potential emergency-off system must switch-off the pumping process well below ~100us and/or ~300us to prevent catastrophic gain for the Yb3+ ASE and/or Er3+ ASE, respectively. An electronic circuit was designed; the board in PCB format was found capable to meet this requirement. The PCB prototype was installed as part of the engineering-level amplifier. A high-power single-frequency EYDFA made from 25/300 PM fibers is presented; the amplifier was implemented with low seed input power to match available GWD-compatible seed laser sources. A pump wavelength of 940nm was used. The pre-amplifier delivered 1.07W output power with low ASE power levels and operated free of SBS. The maximum output power of the high-power amplifier was 110W with 44.4+-0.3% optical-to-optical efficiency. The Er3+ ASE extinction ratio was 48.34dB at maximum output power; the Yb3+ ASE was negligible. SBS-free operation was confirmed by monitoring the amplifier noise at MHz frequencies. The PER ranged from 9.8dB to 12.6dB, probably owed to the used gain fiber. Further power scaling was limited by thermal fiber damage assumed to originate from photodarkening. Moreover, an advanced prototype with a revised cooling approach is presented. The performance of two suitable 25/xxx gain fibers was compared at the ~50W level over a 2-week period. The Nufern fiber showed a growing attenuation, i.e. 14.7+-2.2% per 13 days, that was tentatively attributed to the formation of P1 type color centers from POHCs; further research needs to be undertaken to confirm. The iXblue fiber seemed more heat resilient under operation. Furthermore, the PER from the iXblue fiber was in the range of 15.2dB to 20.7dB; the fundamental mode power was 95.7%. It was concluded that the iXblue fiber is suited to be used in GWD-compatible laser sources.