The existance of a photon subsection in the output section of the configuration file enables the photon output together with the possible photon producing scattering processes. If photons are enabled, the output file named Photons (followed by the appropriate suffix) is generated when SMASH is executed. It's format is identical to the collision output (see OSCAR Collisions Format), it does however only contain information about all particles participating in the photon producing interactions at each timestep.
Further, the block headers differ from the usual collision output:

# interaction in nin out nout rho density weight photon_weight partial part_weight type proc_type

where

density: Density at the interaction point
photon_weight: Weight of the photon process relative to the underlying hadonic interaction. Make sure to weigh each photon in your analysis with this value. Otherwise the photon production is highly overestimated.
part_weight: Always 0.0 for photon processes, as they are hardcoded.
proc_type: The type of the underlying process. See Process Types for possible types.

Note, that "interaction", "in", "out", "rho", "weight", "partial" and "type" are no variables, but words that are printed.
The photon output is available in binary, OSCAR1999, OSCAR2013 and OSCAR2013 extended format.

Photon production in SMASH

Photons are treated perturbatively and are produced from binary scattering processes. Their production follows the framework from Turbide et al. described in Turbide:2006. Following the perturbative treatment, the produced photons do not contribute to the evolution of the hadronic system. They are rather direcly printed to the photon output. The mechanism for photon production is the following:

Look for hadronic interactions of particles that are also incoming particles of a photon process. Currently, the latter include binary scatterings of \( \pi \) and \( \rho \) mesons.
Perform the photon action and write the results to the photon output. The final state particles are not of interest anymore as they are not propagated further in the evolution. To account for the probability that photon processes are significantly less likely than hadronic processes, the produced photons are weighted according to the ratio of the photon cross section to the hadronic cross section used to find the interaction, \( W = \frac{\sigma_\gamma}{\sigma_\mathrm{hadronic}}\). This weight can be found in the weight element of the photon output, denoted as photon_weight in the above.
Perform the original hadronic action based on which the photon action was found. Propagate all final states particles throughout the hadronic evolution as if no photon action had occured.

As photons are produced very rarely, a lot of statistics is necessery to yield useful results. Alternatively, it it possible to use fractional photons (see Content-specific output options on how to activate them). This means that for each produced photon, \( N_{\text{Frac}} \) photons are actually sampled with different kinematic properties so that more phase space is covered. In case fractional photons are used, the weight es redefined as \( W = \frac{\frac{\mathrm{d}\sigma_\gamma}{\mathrm{d}t} \ (t_2 - t_1)}{ N_\mathrm{frac} \ \sigma_{\mathrm{had}}} \).

Note: As photons are treated perturbatively, the produced photons are only written to the photon output, but neither to the usual collision output, nor to the particle lists.