摘要： Laser-based lighting systems are an emerging technology, the next step in solid state lighting that revolutionized the way artificial light is generated. The configuration of interest here is the laser-excited remote phosphor (LRP) scheme that consists of a laser diode as the excitation source of an appropriately chosen phosphor sheet. The phosphor is employed for the down-conversion of the incident laser light and broadening of the output spectrum. Although some commercial applications have already been developed, the optimization of LRP systems has yet to be achieved. A bottleneck in their performance is the thermal dependency of the phosphor’s emission characteristics, a phenomenon also known as thermal quenching. As a result, the need for an opto-thermal simulation strategy arises that will enable the study and optimization of LRP systems [1]. The opto-thermal simulation model discussed here is based on Monte Carlo simulations for the optical part, where the absorbed radiant flux is calculated. These optical thermal losses are subsequently used as a volume heat source to solve the transient heat equation by applying the finite element method (FEM) [2]. As thermal quenching is a time-dependent phenomenon in nature, this is an iterative procedure, where the absorbed flux must be calculated for most of the time steps. Typical Monte Carlo ray tracing algorithms use voxel-based meshes to store any calculated properties / attributes. The problem that arises here is that the computational cost for converting the voxel-based mesh to an FEM mesh would be prohibiting for time-dependent analysis. The solution is to directly store the absorbed flux to the FEM mesh. However, the issue that now emerges is locating the interpolating point, namely the point in space where the absorption occurs, within the unstructured FEM mesh. The processing time of a brute force search would be too long, so more sophisticated solutions must be found. SAM algorithms, which were first introduced in [3], are part of a class of algorithms known as geometric search algorithms [4, 5] that deal with point location in unstructured meshes. We propose here a modified SAM algorithm that uses the optical voxel-based mesh as the auxiliary structured mesh for geometric searching. The two meshes, optical and FEM, are superimposed. As point location in voxel-based meshes is trivial, by mapping which elements of the FEM mesh belong to each voxel, we can easily narrow down the number of searches required. To this end, two maps, implemented as binary search trees, are implemented. The first map, maps the voxel number to the nodes of the FEM mesh that lie within it, while the second map, maps the elements of the unstructured mesh that these nodes belong to. The set-up times of these maps heavily depend on the density of the FEM mesh and the order of the elements used. The use of higher-order elements results in considerable set-up times. As higher order elements are not necessary for thermal analysis, this is not a critical issue here. On the other hand, the denser the optical mesh, the fewer FEM elements are mapped to each voxel. However, the size of voxels should be appropriately chosen, since too small voxels may lead to degenerative cases where there are voxels without any nodes lying in them. A distinct advantage of this method is that once the maps are assembled, the search time of elements is O(1). Simulation plays an increasingly crucial role in the study and optimization of optical systems. Due to the increase in computational capabilities, modelling of more complex phenomena can be included and the need of multi-physics approaches rises. The optical properties of materials often shift to temperature above tolerance levels that may render a particular optical design ineffective. In other cases, structural loads may be the critical issue as they can lead to misalignment of optical elements. The proposed SAM algorithm that enables a more efficient coupling of optical and FEM analysis is a valuable tool to such approaches of optical problems.