Description: Highlights: • Abyssal AUV-based microstructure measurements over rough topography. • Indications for hydraulic control downstream of a sill in a Mid-Atlantic Ridge valley. • Asymmetric distribution of dissipation rate and elevated density variability. Abstract: Diapycnal mixing in the deep ocean is known to be much stronger in the vicinity of rough topography of mid-ocean ridges than above abyssal plains. In this study a horizontally profiling microstructure probe attached to an autonomous underwater vehicle (AUV) is used to infer the spatial distribution of the dissipation rate of turbulent kinetic energy (ε ) in the central valley of the Mid-Atlantic Ridge. To the authors’ knowledge, this is the first successful realization of a horizontal, deep-ocean microstructure survey. More than 22 h of horizontal, near-bottom microstructure data from the Lucky Strike segment (37 °N) are presented. The study focuses on a channel with unidirectional sill overflow. Density was found to decrease along the channel following the mean northward flow of 3 to 8 cm/s. The magnitude of the rate of turbulent kinetic energy dissipation was distributed asymmetrically relative to the position of the sill. Elevated dissipation rates were present in a segment 1-4 km downstream (north) of the sill with peak values of 1⋅10−71⋅10−7 W/kg. Large flow speeds and elevated density finestructure were observed within this segment. Lowered hydrographic measurements indicated unstable stratification in the same region. The data indicate that hydraulic control is established at least temporarily. Inside the channel at wavelengths between 1 m and 250 m the slopes of AUV-inferred horizontal temperature gradient spectra were found to be consistent with turbulence in the inertial-convective subrange. Integrated temperature gradient variance in this wavelength interval is consistent with an ε2/3 dependence. The results illustrate that deep-reaching AUVs are a useful tool to study deep ocean turbulence over complex terrain where free-falling and lowered turbulence measurements are inefficient and time-consuming.