We present a study of seventeen LAEs at redshift 2$<z<$3 gravitationally lensed by massive early-type galaxies (ETGs) at a mean redshift of approximately 0.5. Using a fully Bayesian grid-based technique, we model the gravitational lens mass distributions with elliptical power-law profiles and reconstruct the UV-continuum surface brightness distributions of the background sources using pixellated source models. We find that the deflectors are close to, but not consistent with isothermal models in almost all cases, at the $2\sigma$-level. We take advantage of the lensing magnification (typically $\mu\simeq$ 20) to characterise the physical and morphological properties of these LAE galaxies. From reconstructing the ultra-violet continuum emission, we find that the star-formation rates range from 0.3 to 8.5 M$_{\odot}$ yr$^{-1}$ and that the galaxies are typically composed of several compact and diffuse components, separated by 0.4 to 4 kpc. Moreover, they have peak star-formation rate intensities that range from 2.1 to 54.1 M$_{\odot}$ yr$^{-1}$ kpc$^{-2}$. These galaxies tend to be extended with major axis ranging from 0.2 to 1.8 kpc (median 561 pc), and with a median ellipticity of 0.49. This morphology is consistent with disk-like structures of star-formation for more than half of the sample. However, for at least two sources, we also find off-axis components that may be associated with mergers. Resolved kinematical information will be needed to confirm the disk-like nature and possible merger scenario for the LAEs in the sample.

We use a sample of 17 strong gravitational lens systems from the BELLS GALLERY survey to quantify the amount of low-mass dark matter haloes within the lensing galaxies and along their lines of sight, and to constrain the properties of dark matter. Based on a detection criterion of 10$\sigma$, we report no significant detection in any of the lenses. Using the sensitivity function at the 10-$\sigma$ level, we have calculated the predicted number of detectable cold dark matter (CDM) line-of-sight haloes to be $\mu_{l} = 1.17\pm1.08$, in agreement with our null detection. Assuming a detection sensitivity that improved to the level implied by a 5-$\sigma$ threshold, the expected number of detectable line-of-sight haloes rises to $\mu_l = 9.0\pm3.0$. Whilst the current data find zero detections at this sensitivity level (which has a probability of P$^{{\rm5}\sigma}_{{\rm CDM}}(n_{\rm det}=0)$=0.0001 and would be in strong tension with the CDM framework), we find that such a low detection threshold leads to many spurious detections and non-detections and therefore the current lack of detections is unreliable and requires data with improved sensitivity. Combining this sample with a subsample of 11 SLACS lenses, we constrain the half-mode mass to be $\log$(M$_{\rm hm}) < 12.26$ at the 2-$\sigma$ level. The latter is consistent with resonantly produced sterile neutrino masses m$_{\rm s} < 0.8$ keV at any value of the lepton asymmetry at the 2-$\sigma$ level.