High-energy laser irradiation with large-area uniform intensity distribution plays a crucial role in material initiation and related applications such as propulsion testing， industrial detonation， and advanced material processing. However， achieving both high energy density and uniform spatial distribution remains a significant challenge due to the inherent Gaussian profile of laser sources and the limitations of conventional beam shaping techniques under high-power conditions， such as low damage thresholds or complex fabrication requirements. This study aims to develop a high-energy laser beam expansion and homogenization system based on a tapered optical waveguide， and to systematically investigate the influence of optical surface roughness on irradiation uniformity from a microstructural perspective， providing a new approach for passive beam homogenization through surface engineering.A 2×2 arrayed laser homogenization system is designed using multiple high-power fiber lasers （1064 nm） combined with a tapered solid fused-silica waveguide and a beam-expanding objective. To quantitatively describe surface roughness at the nanometer scale， the fractal dimension （D） is introduced as a key parameter. Three-dimensional rough surfaces with different fractal characteristics are generated using the Ausloos–Berman function in MATLAB， enabling accurate representation of multi-scale surface features. These surfaces are then imported into the COMSOL Multiphysics geometric optics module to simulate the propagation behavior of Gaussian beams under multiple total internal reflections within the waveguide. The output beam profi

les are processed using MATLAB， and the homogenization efficiency （

η

） is evaluated based on one-dimensional intensity distribution analysis. Additionally， spatial frequency analysis via Fourier transform and phase modulation modeling are conducted to reveal the underlying physical mechanisms of roughness-induced beam modulation and energy redistribution.The results indicate that surface roughness significantly affects beam homogenization through phase modulation and scattering mechanisms. The homogenization efficiency shows a non-monotonic dependence on the fractal dimension. As D increases from 2.1 to 2.6， the uniformity improves progressively due to enhanced multi-path mixing and moderate phase perturbation. The optimal performance is achieved at D=2.6， where the homogenization efficiency increases from 83.79% to 95.56%， approaching the theoretical limit of 98.73%. At this condition， the surface exhibits a balanced spatial frequency distribution， enabling effective lateral energy redistribution without introducing excessive random scattering. When D exceeds 2.6， excessive high-frequency components lead to over-randomization of light propagation and reduced uniformity. System-level simulations further demonstrate that the optimized rough surface maintains high performance （

η

=92.05%） in the full optical system， satisfying practical application requirements for large-area irradiation （

>

90%）. Meanwhile， the corresponding surface roughness （0.1558 nm） remains within the achievable range of ultra-precision polishing processes.This study demonstrates that optical surface roughness is not merely a detrimental factor but can be transformed into an effective design parameter for improving beam homogenization when properly controlled. By introducing fractal dimension as a quantitative descriptor， a clear relationship between surface morphology and beam uniformity is established. The optimal fractal dimension （D = 2.6） provides a balance between phase modulation and scattering effects，

leading to near-ideal homogenization performance while maintaining system stability. These findings provide theoretical guidance and practical insights for integrating surface microstructure optimization into high-power laser systems， enabling passive and efficient beam homogenization without increasing system complexity. Future work will focus on experimental validation and further investigation of laser-induced damage behavior under different roughness conditions to support real engineering applications.