The Cold and Ultracold Molecules Group

Our Mission: Develop aluminum monofluoride (AlF) as a new platform for ultracold molecular science

The Cold and Ultracold Molecules Group focusses on using lasers to cool, trap and spectroscopically investigate small molecules for the purpose of fundamental science. Since 2017, the group has been primarily focussed on laser cooling and trapping the diatomic molecule aluminum monofluoride (AlF).

Why cool molecules with lasers?

The magneto-optical trap (MOT) is the starting point for a wide range of experiments in the cold and ultracold regime.

The magneto-optical trap (MOT) is the starting point for a wide range of experiments in the cold and ultracold regime.

Laser cooling of atoms was established over 40 years ago, and is a powerful technique to study microscopic and mesoscopic physics with unprecedented resolution and control. The radiation pressure exerted by light can, over thousands of absorption and emission cycles, confine and cool atoms to below one thousandth of a degree above absolute zero temperature. This has enabled the generation of degenerate quantum gases, the observation of phase transitions in atoms confined to artificial crystals made of light, and the realisation of optical atomic clocks. The so-called “Magneto-optical trap”, a subject of the 1997 Nobel Prize in physics, now underpins a vast array of experiments and devices in quantum information, simulation and metrology.

Applying the technique of laser cooling to a diatomic molecule - the simplest chemical compound possible - took around 30 years longer than for atoms, because of the complexities of molecular structure. However, an intense field of research has now been established, motivated by key scientific questions of our time. Polar molecules at ultralow temperatures can be used to study strongly interacting quantum matter, they can serve as measurement tools to test the symmetries of fundamental physics, and they are a platform in which quantum information can be encoded and manipulated. The rich internal energy structure of molecules also enables studying chemistry and collisions at the point where quantum behaviour emerges. Together, this potential is driving experimentalists to bring new molecular species into the ultracold regime.

Aluminum monofluoride: a deeply bound, stable molecule for direct laser cooling

In the figure below, a large selection of diatomic molecules are organised by the symmetry of their electronic ground states, and plotted according to their molecular mass and binding energy. As of early 2025, only species with ²Σ⁺ ground states have been directly laser-cooled and loaded into a MOT (red points in the lower left plot). These molecules tend to be chemically reactive, difficult to produce efficiently in experiments, and laser cooling is most effective only from the first excited (N=1) rotational level.

A large selection of diatomic molecules organised into four plots by the classification of their electronic ground state. Species highlighted in black are of interest or under investigation for laser cooling and trapping. Those in red were loaded into a magneto-optical trap by the year 2025. Source data: the Diatomic Molecular Spectroscopy Database (https://rios.mp.fhi.mpg.de/).

A large selection of diatomic molecules organised into four plots by the classification of their electronic ground state. Species highlighted in black are of interest or under investigation for laser cooling and trapping. Those in red were loaded into a magneto-optical trap by the year 2025. Source data: the Diatomic Molecular Spectroscopy Database (https://rios.mp.fhi.mpg.de/).

AlF is fundamentally different to the species that have been laser-cooled thus far:

Its ground state is a spin-singlet (¹Σ⁺, upper left plot) with a very large binding energy of 6.9 eV. It is an exceptionally stable molecule with many similarities to carbon monoxide; all bimolecular AlF-AlF reactions are more than 1 eV endothermic and therefore forbidden at low temperature.
Any excited rotational level in AlF can be laser-cooled in a simple scheme. Under electric dipole transition selection rules, all Q(J) lines of the A¹Π ← X¹Σ⁺ transition support optical cycling.
AlF possesses a narrow, spin-forbidden a³Π ← X¹Σ⁺ transition, and molecules in the metastable a³Π state phosphoresce to the ground state on millisecond timescales. This transition is analogous to the intercombination lines in the alkaline earth atoms, which are utilised in optical atomic clocks.

Breaking new ground with deep ultraviolet laser light

Russell Thomas, laser engineer in the molecular physics department.

Russell Thomas, laser engineer in the molecular physics department.

The principal A¹Π, v’=0 ← X¹Σ⁺, v’’=0 laser cooling wavelength for AlF is at λ=227.5 nm, deep in the ultraviolet (uv) and just beyond the shortest ever used to trap atoms in a MOT. This is a challenging wavelength range for laser cooling experiments, for which stable, continuous optical power is essential. At present, the only effective means of generating the light we need is via frequency conversion of high-power infrared or visible laser light in external resonant cavities. This results in complex laser systems that are challenging to operate: four deep-uv laser systems are needed for magneto-optical trapping experiments, compounding the experimental challenges.

Working with industry has been key pushing laser technology to new limits and making our research possible. A key part of this strategy is the industry-academic partnership UVQuanT, a Horizon Europe-funded project bringing together three laser companies and four academic partners to develop new technology in the deep uv. In addition we have directly brought laser expertise from industry into the research labs of the molecular physics department, with laser engineer Russell Thomas having a decade of experience with deep uv laser light.

The pathway to a trap: spectroscopy, sources, lasers, hardware

Deep ultraviolet laser slowing of cadmium atoms (upper) and AlF molecules (lower). Velocity-time plots are shown for the atomic/molecular sources, and when chirped-frequency laser slowing is applied.

Deep ultraviolet laser slowing of cadmium atoms (upper) and AlF molecules (lower). Velocity-time plots are shown for the atomic/molecular sources, and when chirped-frequency laser slowing is applied.

The work of the group began in 2017 under group leader Stefan Truppe with a detailed spectroscopic investigation of AlF in supersonic beam experiments. This was essential to verify that electronic, vibrational and hyperfine leaks in the laser cooling cycle were below a level which would be problematic, and that no other large losses were present. In particular, the two lowest energy levels in the spin triplet manifold, the a³Π, v’=0 and b³Σ⁺ states, were characterised and their interaction with the A¹Π state was investigated. In parallel, the first deep uv laser systems were installed and tested, and a cryogenic buffer gas source for slow AlF molecules was constructed.

In 2021, the first detailed optical cycling measurements on the A¹Π ← X¹Σ⁺ transition were completed and it was demonstrated that laser deflection of an AlF molecular beam was directly observable, verifying the large radiation pressure achievable in the deep uv. At the same time, much of the hardware and apparatus was tested on atomic cadmium, an analogous but simpler atomic system to AlF with a ¹S₀ → ¹P₁ cooling transition at the nearby wavelength of 228.9 nm. In mid-2022, Stefan Truppe joined Imperial College London and Sid Wright took over as group leader. As the final deep uv systems were installed, it became possible to laser-slow the AlF molecular beam and directly verify the expectations from prior spectroscopic investigations. As can be seen from the velocity-time plots on the right, chirped-frequency laser slowing for AlF qualitatively matches results obtained with an atomic cadmium beam.

A deep ultraviolet magneto-optical trap of chemically stable molecules

Schematic view of the AlF MOT experiment. Molecules from a cryogenic buffer gas source are laser slowed and captured into a magneto-optical trap, and their fluorescence is imaged onto a camera. The trap can be loaded with molecules in three different rotational levels.

Schematic view of the AlF MOT experiment. Molecules from a cryogenic buffer gas source are laser slowed and captured into a magneto-optical trap, and their fluorescence is imaged onto a camera. The trap can be loaded with molecules in three different rotational levels.

In early 2025, we achieved a magneto-optical trap of AlF at the FHI, a significant landmark within the field of laser cooling. As shown in the schematic view of the experiment, AlF molecules are produced in a buffer gas cell and cooled with cryogenic helium to about 3 K. Molecules leaving the cell are slowed by two lasers allow capture into a magneto-optical trap. Fluorescence of the trapped molecules is imaged onto a camera, which results in a dense cloud shown in images in the inset. The temperature of the cloud is about 15 mK and molecules remain trapped there over tens of milliseconds.

This is the first time that laser-cooled molecules have been trapped in anything other than the first excited rotational level, and is possible because of AlF’s distinct electronic structure. We estimate that around 60,000 molecules in the J=1 level are captured into a MOT using the Q(1) line, and at least 10,000 are captured into MOTs targeting the J=2 and 3 levels.

The next steps: alternative molecular sources, metastable states, optical dipole trapping

Following our demonstration that an AlF magneto-optical trap is possible, a number of avenues become appealing to explore:

First, several properties of AlF make it possible to trap molecules from molecular beam sources that are much cheaper and more compact than the cryogenic sources used at present. The ability to efficiently generate AlF by a thermochemical reaction at moderate temperatures (~900 K) has been known and used for decades to study this molecule. In the cold and ultracold molecules group, we have developed a high-brightness, continuous molecular beam oven, a molecular “dispenser” source (similar to those used in experiments with laser-cooled atoms), and have observed that AlF can thermalize to room-temperature surfaces without immediately freezing or reacting. Showing that any of these alternative sources can be used to load a molecular MOT would represent a step-change in our field of research, significantly reducing complexity, cost, and experimental downtime.

A laser-induced fluorescence spectrum of our AlF thermochemical molecular beam, taken by exciting the A1Π ← X1Σ+ transition with a cw deep uv laser. The molecular beam is too hot for direct trap loading, but in future we believe this thermochemical reaction could serve as a source for continuous molecular MOT. — A laser-induced fluorescence spectrum of our AlF thermochemical molecular beam, taken by exciting the A¹Π ← X¹Σ⁺transition with a cw deep uv laser. The molecular beam is too hot for direct trap loading, but in future we believe this thermochemical reaction could serve as a source for continuous molecular MOT.

A laser-induced fluorescence spectrum of our AlF thermochemical molecular beam, taken by exciting the A¹Π ← X¹Σ⁺transition with a cw deep uv laser. The molecular beam is too hot for direct trap loading, but in future we believe this thermochemical reaction could serve as a source for continuous molecular MOT.

Second, the most exciting scientific possibilities require loading ultracold AlF into a conservative (i.e. non-dissipative) trap. Driving the a³Π ← X¹Σ⁺ transition for molecules in the MOT transfers molecules into a state with a non-zero electronic magnetic moment, and hence magnetic trapping in (weak-field seeking) sub-levels of the metastable triplet state is possible. Radiative decay of the a³Π state would set the lifetime limit in such a trap, in the range of two to several hundred milliseconds. Long-term, we aim to load AlF molecules into a far off-resonant, focused optical trap using high-power visible or infrared light. In this trap, the intrinsic lifetime should be of the order of seconds and it will be possible to trap the absolute ground state. At present, we are theoretically investigating the dynamic polarizabilities to determine suitable “magic” trapping conditions for which motional dephasing of the a³Π ← X¹Σ⁺ transition is eliminated.