Choose a transport model ======================== MomentRT provides monochromatic moment approximations. Select the model from the physical domain and desired result. Start with your question ------------------------ * **What is the fluence around an embedded point source?** Use ``InfiniteP1`` for an infinite homogeneous medium. Its Cartesian mesh is an observation window, and the analytical solution is singular at the source. It cannot predict escape through a physical surface. * **What are the fluence and escaping power in a finite sample?** Start with ``P1`` when a P1 approximation is appropriate. Define the sample, coupled diffuse surface source, mesh, and all six face conditions explicitly. Consider ``SP3`` or ``P3`` when their additional angular moments and supported boundary conditions are relevant to the question. * **How much power reaches a contact aperture?** Add a ``SurfaceDetector`` to a finite simulation and run a :doc:`detector convergence study `. Detector power uses a Lambertian angular approximation; per-face escaping power is not an aperture measurement. The model does not include external detector optics or specular entry losses. * **How does a diffuse pulse evolve over time?** Use ``TransientP1`` with an explicit time grid, source profile, and energy speed. See :doc:`time_dependent` for initial conditions, boundary assumptions, and validation limits. For any finite solver, refine the mesh and assess source footprint sampling; for detector power, assess aperture sampling and power stability as well. Area agreement and energy balance alone do not establish convergence or model accuracy. Compare the observable against a suitable transport reference for the intended geometry, particularly near a source or in weakly scattering media. See :doc:`meshes`, :doc:`detectors`, and :doc:`convergence`. Check the configuration and its assumptions before solving: .. code-block:: python simulation = MomentRT.Simulation(...) report = simulation.validate() print(report.summary()) The report lists errors that prevent a solve, mesh and footprint cautions, and ``model_notes`` containing the configured solver's assumptions and known limits. Validation does not solve the equation, establish mesh convergence, or determine whether an approximation is accurate for a particular sample. .. list-table:: Solver choice and principal limits :header-rows: 1 :widths: 18 34 32 16 * - Solver - Use it for - Key assumptions and limits - Example * - ``InfiniteP1`` - A homogeneous, unbounded medium with an embedded isotropic point source. - Analytical P1 Green function; no physical boundaries or finite source. Singular at the source; diffusion is least reliable nearby. No FEM solve. - :doc:`Infinite medium ` * - ``P1`` - A finite sample illuminated by diffuse light on a face. - P1 diffusion with continuous fluence, centroid-sampled properties, and explicit six-face conditions. Internal refractive-index jumps are rejected. - :doc:`Finite sample ` * - ``SP3`` - A finite sample when the P1 closure is too restrictive and a higher angular moment model is useful. - Simplified P3; requires phase-function moments, zero-reflectance Robin conditions on nonreflecting faces, and equal internal indices. Unilluminated faces may be reflecting. Two coupled FEM fields; it is not full P3 or the radiative transfer equation. - :doc:`P1 and SP3 ` * - ``P3`` - A finite sample when retaining all three-dimensional harmonics through degree three is useful. - Full P3 moment approximation; requires phase-function moments and matched-index zero-reflectance Robin faces. Six coupled FEM fields can cost substantially more than SP3. More angular detail does not guarantee accuracy for a given geometry. - :doc:`P1, SP3, and P3 ` P1 diffusion is generally appropriate when multiple scattering supports a diffusive description and the observable is not dominated by the source boundary layer. Higher-order moment models retain more angular information, but remain approximations and have their own boundary and phase-function limits. For optically thin, weakly scattering, strongly absorbing, or very short source-detector configurations, compare against a suitable radiative transfer or Monte Carlo reference rather than assuming a higher order is accurate. Model notes are available programmatically as ``report.model_notes``. The complete report also identifies unsupported internal refractive-index jumps with region names and resolved indices. Fixing configuration errors does not remove numerical or physical model error. Independent references ---------------------- The :doc:`planar slab example ` compares P1 and SP3 finite-element fields with independently calculated one-dimensional analytical solutions. This checks the FEM implementation and spatial discretization for a planar homogeneous case. It does not validate the approximations against the full radiative transfer equation. The :doc:`sensor validation benchmark ` compares measured aperture power from P1, SP3, and P3 against an independent layered-slab photon transport reference across source-detector separations and mesh refinements. For a broader view of model breakdown in scattering slabs, see Akdemir et al., `Breakdown of light transport models in photonic scattering slabs with strong absorption and anisotropy `_. Its Monte Carlo comparisons show why P1 and P3 applicability depends on optical thickness, absorption, anisotropy, and refractive-index contrast. Those results are not a direct validation of MomentRT's finite-element implementation.