On-demand holographic VCSELs with integrated nanoprinted diffractive neural networks

Vertical-cavity surface-emitting lasers (VCSELs) are promising light sources for integration with phase structures to realize active, chip-scale holographic devices. However, conventional hologram design methods neglect the actual VCSEL emission profile, leading to reduced reconstruction accuracy, and they also exhibit low generality and scalability. These limitations impede the development of more complex and system-level holographic applications. Here, we demonstrate on-demand holographic VCSELs enabled by a diffractive neural network (DNN) design framework combined with two-photon laser nanoprinting. By experimentally characterizing the transverse modes of oxide-confined VCSELs and fitting them with a weak-waveguide fiber model, accurate near-field emission profiles are incorporated directly into the DNN-based inverse-design process. This yields a unified and high-precision approach for engineering holographic diffractive layers, which are subsequently nanoprinted onto the VCSEL facet for monolithic integration. We experimentally realize two representative holographic functions—image projection and multi-plane beam focusing. The reconstructed results closely match the target designs, as confirmed by the peak signal-to-noise ratios, demonstrating their potential for on-chip display and optical-communication applications. The laser nanoprinting process offers high design flexibility while preserving intrinsic VCSEL performance. These results establish a scalable and programmable platform for next-generation holographic VCSELs, advancing compact displays, parallel optical links, and integrated photonic computing.