Surface Photon Monte Carlo (SPMC) Radiation

The Surface Photon Monte Carlo (SPMC) radiation model applies Monte Carlo statistical methods to randomized bursts of photons to calculate close approximate solutions for radiative energy exchange between surfaces.

In the SPMC method, the key processes of radiation (such as emission, absorption, scattering, and boundary interaction) are modeled explicitly by stochastic methods, as opposed to obtaining a numerical solution of the governing equations. The method entails simulating the radiative processes by a ray trace procedure for a large number of photon bundles (representative samples of radiative energy). The generation of these photon bundles mimics the emission process, and the interaction of these photon bundles with the surfaces as they travel in the computational domain represent the radiation boundary treatment: absorption, reflection, transmission, or refraction. During the ray trace procedure, statistics are collected for the mesh faces to obtain the radiative quantities of interest.

A surface-to-surface radiation simulation with SPMC has the following modeling components:

  • Emission modeling
  • Movement of photon bundles
  • Boundary treatment modeling
Emission Modeling
Emission of radiative energy from surfaces in a SPMC simulation is represented by the generation of photon bundles. This requires decisions on the number of photon bundles to be emitted at a given patch, the direction vector of the bundles, and their starting location on the patch.

In Simcenter STAR-CCM+, boundary surfaces are discretized into smaller elements called patches. These patches are sets of contiguous, non-overlapping boundary cell faces. The SPMC model solves for radiative exchange between the patches. The emissive power and radiation properties are assumed to be uniform over the surface of each patch, and are computed by surface averaging, as shown in Eqn. (1707).

Number of photon bundles from a given patch: For a SPMC simulation, the total number of photon bundles in a problem (i.e., user specified average number of rays per patch times the total patch count in the simulation) is distributed among the patches so that patches with higher emissive power emit more photon bundles. It has been shown in the literature that such sampling of photon bundles leads to lower statistical noise in the results for a given number of the photon bundles.

Emission direction: For diffuse emission from a patch, random direction is sampled according to this equation: [412]:
Figure 1. EQUATION_DISPLAY
ψ = 2 π R ψ ; θ = sin -1 R θ
(1714)
where:
  • ψ and θ the azimuthal and polar angles, respectively, for the emission direction.
  • R ψ and R θ are uniform random numbers ( [ 0 , 1 ] ) used in the sampling.
Emission location on a patch: The entire patch surface area emits energy, and therefore the starting location for a photon bundle is randomly chosen on the patch. It is important for the emission locations to be randomly sampled from a patch surface, for an accurate representation of radiative energy exchange between surfaces in a simulation. For an arbitrary patch, the emission location can be sampled according to these equations:
Figure 2. EQUATION_DISPLAY
R x = 1 E p A p x < x ϵ σ T 4 d A R y = 1 E p A p y < y ϵ σ T 4 d A R z = 1 E p A p z < z ϵ σ T 4 d A
(1715)
where:
  • R x , R y , and R z are the uniform random numbers between 0 and 1.
  • A p is the patch area.
  • E p is the emissive power of the patch. ( E p = A p ϵ σ T 4 d A )
  • ϵ is the emissivity.
  • σ is the Stefan-Boltzmann constant.
  • T is the temperature.
The above expressions are inverted to get the (x,y,z) emission location.

Multi-band modeling: For multi-band simulations, spectral energy content is not sampled randomly on photon bundles and they carry emissive powers for all bands.

External collimated and diffuse sources: External directional or diffuse sources (such as solar loads or external diffuse flux at boundaries), if present, are included in SPMC. The incident fluxes on patches due to the specular component of the solar and environment loads are computed separately, and are used in SPMC to provide the effective diffuse emissive power from patches, using the following equation:
Figure 3. EQUATION_DISPLAY
E e f f , p = E p + q p,src + ρ d ( q d i r , e x t + q d i f , e x t )
(1716)
where:
  • E e f f , p is the effective diffuse emissive power from a patch.
  • E p is the patch local emissive power due to temperature-based emission.
  • ρ d is the diffuse reflectivity for the patch.
  • q p,src is the specified external diffuse energy source from the patch.
  • q d i r , e x t and q d i f , e x t are the incident fluxes on the patch due to the specular components of the directional and diffuse external sources, respectively.
The effective diffuse emissive power is used in the SPMC formulation for photon bundle emission modeling.
Movement of Photon Bundles
The photon bundles are traced in the direction of their travel till they encounter a boundary, and then the boundary treatment is performed.
Boundary Treatment Modeling
The radiative processes of absorption, reflection and transmission are modeled by a stochastic interaction scheme of photon bundles with the boundary mesh faces. In this scheme, photon bundles deposit some energy on the boundary proportional to its absorptivity, and rest of the radiative processes are modeled in a statistical sense, where a process is randomly chosen (weighted by their energy importance) for the boundary interaction of the incident photon bundle. The photon bundle goes through the randomly chosen radiative process (transmission or diffuse reflection or specular reflection) at the boundary.

The key idea is that for any boundary surface, the radiative interactions for a large number of photon bundles are selected randomly in such a way that they provide the desired statistics as dictated by the boundary properties. Instead of a stochastic approach, a deterministic approach (in which a photon bundle is split at the boundary to account for all possible interactions) can be used for the boundary treatment, and that might provide lower statistical noise in the solution than does the stochastic approach for the same number of the initial photon bundles. However, the subsequent generation of split photon bundles during deterministic boundary treatment can lead to a very large number of photon bundles in the simulation, several orders of magnitude larger than the original number of bundles. Generally, it is far more computationally efficient to employ the stochastic approach, albeit with a relatively higher number of initial/primary photon bundles to have the same level of statistical noise as the deterministic approach.

In this approach, for an arbitrary patch with diffuse reflectivity ρ d , specular reflectivity ρ s , and transmissivity τ , the boundary interaction for an incident photon bundle can be randomly sampled by drawing a uniform random number ( R ρ d , ρ s , τ ) between 0 and 1, using the following equations:

Figure 4. EQUATION_DISPLAY
Diffuse reflection: 0 < R ρ d , ρ s , τ ρ d ρ d + ρ s + τ Specular reflection: ρ d ρ d + ρ s + τ < R ρ d , ρ s , τ ρ d + ρ s ρ d + ρ s + τ Transmission: ρ d + ρ s ρ d + ρ s + τ < R ρ d , ρ s , τ 1
(1717)
The direction vector after the boundary interaction for the incident photon bundle is determined according to the boundary treatment. For diffuse reflection, expressions given in Eqn. (1714) are used whereas for specular reflection the outgoing direction can be determined from the law of optics. For transmission, refractive effects are considered while determining the direction.

The emission and boundary treatment for SPMC provides the solution for the radiative energy exchange between boundaries. The incident radiative heat fluxes are recorded on the patches during the SPMC ray trace procedure, and these fluxes are treated to account for the statistical nature of SPMC solution. This treatment is performed according to the statistical sampling factor specification in the user-interface for the SPMC Solver, and is discussed below.

The incident radiative heat fluxes, that is, irradiation, on patches after statistical treatment are then used to get the irradiation on faces, which is subsequently used in the computation of the radiation source term for the emission equation. The methodology followed here for the SPMC model is identical to the S2S model. See Eqn. (1708)Eqn. (1712) for more details.

Statistical Sampling Factor

The sampling factor controls the update of the boundary irradiation during the SPMC computations. The irradiation obtained from a single SPMC ray tracing at a given iteration contains statistical noise due to the use of a finite number of photon bundles, therefore only a fraction of this SPMC solution is used to update the boundary irradiation. The sampling factor is the fraction used in this update procedure. At every iteration, the boundary irradiation is updated as follows:

Figure 5. EQUATION_DISPLAY
I n + 1 = f ssf I PMC + ( 1 f ssf ) I n
(1718)
where:
  • I n + 1 is the boundary irradiation in the current iteration.
  • I n is the boundary irradiation in the previous iteration.
  • f ssf is the sampling factor.
  • I PMC is the irradiation computed from the latest SPMC compute call.
The I field gets used in computing the radiative sources for the energy equation, as in Eqn. (1708)Eqn. (1712).

The sampling factor has two purposes:

  • I PMC contains statistical noise, and can vary significantly from one SPMC compute call to another (due to statistical noise or flow evolution). This possibly large variation in the I PMC field can cause numerical instabilities in the energy equation, so the newly obtained SPMC solution ( I PMC ) is slowly phased into the I field using Eqn. (1718). To give an example, for a steady simulation, if the SPMC solver update frequency is 10, the SPMC ray tracing is performed before iteration 1, and then the I PMC field is used to continually update I till iteration 10, as shown below.
    Figure 6. EQUATION_DISPLAY
    Iteration 1: I 1 = f ssf I PMC + ( 1 f ssf ) I 0 Iteration 2: I 2 = f ssf I PMC + ( 1 f ssf ) I 1 Iteration 9: I 9 = f ssf I PMC + ( 1 f ssf ) I 8
    (1719)

    where I i is the irradiation field at iteration i .

    The call to SPMC ray tracing is made again at iteration 10, leading to a new SPMC solution (call it I PMC , new ), and that new SPMC solution can be gradually phased into into the I field.
    Figure 7. EQUATION_DISPLAY
    Iteration 10:    I 10 = f ssf I PMC , n e w + ( 1 f ssf ) I 9
    (1720)

    This procedure slowly includes the SPMC computed irradiation solution while computing radiative feedback to the energy equation.

  • The second purpose of the sampling factor is to be able to obtain the statistics once the solution has reached a statistically stationary stage. At that point, the sampling factor can be set or gradually reduced to a low value (for example, 0.02) and then the statistics can be collected over a number of iterations (depending on the problem and the photon bundles used per SPMC computation). The collection of statistics over a small number of iterations may suffice if the photon bundles used per SPMC computation is high enough for the problem. Monitoring of averaged, minimized, maximized, and integrated quantities on boundaries of interest can be used in determining whether a statistically stationary stage has been achieved and enough statistics have been collected.