High Density Loading and Collisional Loss of Laser-Cooled Molecules in an Optical Trap

We report optical trapping of laser-cooled molecules at sufficient density to observe molecule-molecule collisions for the first time in a bulk gas. SrF molecules from a red-detuned magneto-optical trap (MOT) are compressed and cooled in a blue-detuned MOT. Roughly 30% of these molecules are loaded into an optical dipole trap with peak number density $n_0\approx 3\times {10}^{10}{\mathrm{cm}}^{-3}$ and temperature $T\approx 40\mathrm{\mu }\mathrm{K}$. We observe two-body loss with rate coefficient $\beta ={2.7}_{-0.8}^{+1.2}\times {10}^{-10}{\mathrm{cm}}^3{\mathrm{s}}^{-1}$. Achieving this density and temperature opens a path to evaporative cooling towards quantum degeneracy of laser-cooled molecules.

Figure 1

Relevant SrF level structure and laser-driven transitions for different stages in the experiment, with hyperfine levels ($|J,F⟩$) and their magnetic $g$ factors ($g$) listed. Solid (dashed) lines indicate ${\sigma }^{+}({\sigma }^{-})$ laser polarization, and color indicates red or blue detuning. (a) Red MOT, which employs the dual frequency mechanism on $|3/2,1⟩$. (b) Blue MOT, where the laser addressing $|3/2,2⟩$ is now blue, but also provides the red detuning needed for the dual frequency mechanism on $|3/2,1⟩$ (purple arrow). (c) $\mathrm{\Lambda }$ cooling, where only two lasers address $|3/2,1⟩$ and $|1/2,1⟩$.

Figure 2

Fluorescence images showing capture into the blue MOT (2 ms exposure starting at $t$ after switching to the blue MOT). The loading is quick and efficient, with $\approx 80\%$ of molecules captured by $t=30\mathrm{ms}$.

Figure 3

Number of molecules in the trap as a function of hold time. Each point is an average of 15 images, and the error bars account for uncertainties as described in the main text. Data for $t_h<1\mathrm{s}$ are $\mathrm{\Lambda }$ images (blue circles) and the rest are MOT recapture images (red squares). The data show a clear deviation from an exponential decay, a classic signature of two-body loss. By fitting to a model where ${\sigma }_z$ is increasing linearly with time, we extract a two-body loss rate coefficient $\beta ={2.7}_{-0.8}^{+1.2}\times {10}^{-10}{\mathrm{cm}}^3{\mathrm{s}}^{-1}$, and a one-body loss time constant $\tau =1.3(1)\mathrm{s}$. The shaded area indicates the uncertainty range.

Figure 4

Short-time evolution of trap population for different starting conditions. Dashed lines are fits for the first 9 points to the two body loss rate model with fixed $\tau =1.3\mathrm{s}$ and the average $V_{\mathrm{eff}}$ for $t_h<250\mathrm{ms}$. Data with initial ODT number $N_0\approx 650$ (green triangles) have a slower initial loss than for $N_0\approx 4000$ (red circles), clearly demonstrating the density dependent loss. The presence of $\mathrm{\Lambda }$-cooling light leads to additional two-body loss (blue squares) due to light-assisted collisions. For all conditions, the one-body loss rate remains the same (as seen in longer-time data, not shown).