Context. LS 5039 is a powerful high-mass gamma-ray binary that probably hosts a young non-accreting pulsar. However, despite the wealth of data available, the means by which the non-thermal emitter is powered are still unknown.
Aims. We use a dynamical-radiative numerical model, and multiwavelength data, to constrain the properties of a hypothetical pulsar wind that would power the non-thermal emitter in LS 5039.
Methods. We ran simulations of an ultrarelativistic (weakly magnetized) cold e±-wind that Compton scatters stellar photons and that dynamically interacts with the stellar wind. The effects of energy losses on the unshocked e±-wind dynamics, and the geometry of the two-wind contact discontinuity, are computed for different wind models. The predicted unshocked e±-wind radiation at periastron, when expected to be the highest, is compared to LS 5039 data.
Results. The minimum possible radiation from an isotropic cold e±-wind overpredicts the X-ray to gamma-ray fluxes at periastron by a factor of ∼3. In the anisotropic (axisymmetric) wind case X-ray and ≳100 MeV data are not violated by wind radiation if the wind axis is at ≲20−40° from the line of sight (chance probability of ≲6−24%), depending on the anisotropic wind model, or if the wind Lorentz factor ∈102 − 103, in which case the wind power can be higher, but it requires e±-multiplicities of ∼106 and 109 for a 10−2 s and 10 s pulsar period, respectively.
Conclusions. The studied model predicts that a weakly magnetized cold pulsar e±-wind in LS 5039 should be strongly anisotropic, with either a wind Lorentz factor ∈102 − 103 and very high multiplicities or with a fine-tuned wind orientation. A weakly magnetized, cold baryon-dominated wind would be a possible alternative, but then the multiplicities should be rather low, while the baryon-to-e± energy transfer should be very efficient at wind termination. A strongly magnetized cold wind seems to be the most favorable option as it is consistent with recent research on pulsar winds and does not require fine-tuning of the pulsar wind orientation, and the wind multiplicity and Lorentz factor are less constrained.