Context : Par-Lup3-4 is a very low-mass star ( spectral type M5 ) in the Lupus III star-forming region .
It shows spectroscopic evidence of accretion and mass-loss .
In the optical and near-infrared , the object is underluminous by \approx 4 mag when compared to objects of similar mass in the same association .

Aims : The aim of this work is to characterize the circumstellar environment of Par-Lup3-4 to better understand the origin of its underluminosity .

Methods : We have analyzed high angular resolution near-IR observations and searched for extended emission from a disk and/or an envelope .
We have studied the spectral energy distribution ( SED ) of the target from the optical to the sub-millimeter regime , and compared it to a grid of radiative transfer models of circumstellar disks .
Since the target is strongly variable , we modeled two different near-infrared datasets .

Results : The SED of Par-Lup3-4 resembles that of objects with edge-on disks seen in scattered light , that is , a double peaked-SED and a dip at \sim 10 \mu m. The diffraction-limited infrared observations do not show obvious extended emission , allowing us to put an upper limit of \sim 20 AU to the disk outer radius .
Par-Lup3-4 is probably in a Class II ( rather than a Class I ) evolutionary stage , which is indicated by the lack of extended emission together with the non detection of a strong 9.8 \mu m silicate in absorption .
This last feature is indeed seen in emission .
We fitted the whole SED of Par-Lup3-4 with a single disk model .
Our modeling predicts a disk inclination of 81° \pm 6° , which agrees well with previous estimates , and provides a natural explanation for the under-luminosity of the target .
The detection of the silicate feature in emission at such a high inclination might be related to a more complex disk structure ( e.g .
asymmetries , inhomogeneities ) than the one assumed here .
Our analysis allows us to put constraints on the disk inner radius , R _ { \mathrm { in } } \leq 0.05 AU , which is very close to the dust sublimation radius , and the maximum size of the dust grains , a _ { \mathrm { max } } \geq 10 \mu m , which indicates that dust processing has already taken place in Par-Lup3-4 .
Some of the derived disk parameters vary depending on the modeled near-infrared data-set , which emphasizes the importance of taking variability into account when modeling the SED of young stellar objects .

Conclusions :