Dropping a chunk of sodium into water and watching it explode is a classic high-school chemistry demonstration. The violent reaction is caused by the alkali metal’s dissociation into Na+ ions and electrons when it enters H2O. The electrons react with the water to liberate hydrogen atoms, and those quickly pair to form H2 gas that is ignited by the exothermic reaction.
The same demonstration becomes less incendiary if H2O is replaced with liquid ammonia, because NH3 is harder to break apart. Whereas about 1 in every 109 H2O molecules self-dissociates (roughly 10−7 moles per liter), only about 1 in 1017 NH3 molecules does. Free electrons’ brief lifetimes in water—less than a microsecond—make them, and their effects on the molecules around them, difficult to observe. But in ammonia, the formation of H2 gas is slow enough to form metastable alkali metal–NH3 solutions. Ammonia solutions can therefore be used to study solvated electrons, which are highly reactive participants in many chemical reactions, including the Birch reduction that was critical for developing synthetic steroids and the first oral contraceptives.
Nuclear magnetic resonance and electron spin resonance have long been used to study alkali metal–NH3 solutions,1 but to directly probe their energetics, photoelectron spectroscopy (PES) is the favored technique. It entails bombarding a material with x rays or UV light to eject electrons; the electrons’ kinetic energies are then measured to determine their binding energies. Unfortunately, applying PES to alkali metal–NH3 solutions is difficult. The ultrahigh vacuum needed to give photoelectrons an unimpeded path to the spectrometer causes the volatile liquid ammonia to evaporate almost immediately. Samples must be kept below −33 °C, the temperature at which ammonia boils, and in a meticulously clean environment to avoid unwanted auxiliary reactions.
Now Tillmann Buttersack (then at IOCB Prague, Czech Academy of Sciences, now at the University of Southern California), Ryan McMullen (USC), Phil Mason (IOCB Prague, CAS), Christian Schewe (Fritz Haber Institute of the Max Planck Society), and coworkers have overcome those challenges by performing PES on alkali metal–NH3 microjets using lithium, potassium, and sodium.2
Their measurements span a range of concentrations over which the solution’s behavior transitions—from electrolytic, in which localized ions carry currents as in salt solutions, to metallic, with extended electron states that facilitate charge transport. The spectrum’s evolution points to a gradual change in electronic structure as the dissolved metal concentration increases. Along with simulations, the results suggest that localized electrons coalesce into a system-spanning electron network to cause the metallic transition.
A fine blue color
Alkali metal–NH3 solutions have been studied for more than 200 years, and much is already known.3 Humphry Davy first reported on them in 1807 while trying to prove that potassium was an element rather than a hydride of potash; he observed that the metal was dissolved by gaseous ammonia and produced a blue film. Ammonia having been liquefied in 1823 enabled further experiments, including the discovery of the solution’s metal transition more than 50 years later.
Subsequent studies have filled in details. At low concentrations—below a mole fraction of about 10−5, or 10−3 mol percent metal (MPM)—electrons and cations from the dissolved metal are isolated in localized cavities surrounded by NH3 molecules. A conductive electrolytic solution results. The solvated electrons absorb IR light, and in ammonia that absorption peak bleeds into the visible spectrum and produces the deep blue color shown in figure 1.
Alkali metal–ammonia solutions are a deep blue color at concentrations below about 4 mol percent metal. At higher concentrations their electrons transition from localized to extended states, which causes the solutions to turn bronze and exhibit metallic behavior. (Courtesy of Philip Mason.)
Alkali metal–ammonia solutions are a deep blue color at concentrations below about 4 mol percent metal. At higher concentrations their electrons transition from localized to extended states, which causes the solutions to turn bronze and exhibit metallic behavior. (Courtesy of Philip Mason.)
Increasing the metal concentration beyond 10−3 MPM results in pairing—first between electrons and cations, and then between opposite-spin electrons—which reduces the solution’s conductance. By about 0.5 MPM the conductivity reaches a minimum, a feature also seen in simple salt solutions.
Above 1 MPM, however, the ammonia solution’s conductivity increases, and around 4 MPM the solution turns bronze (see figure 1). Ammonia is unusual in that it can support high enough solvated electron concentrations to reach that transition. The metallic sheen reflects underlying behavior: Previously localized electrons inhabit extended states like those responsible for conduction in familiar metals; they screen the electric field of incoming visible light to produce a telltale reflectivity. When the solution reaches saturation around 20 MPM its conductance is about half that of mercury; that’s noteworthy considering there are four nonconducting ammonia molecules for every solvated electron in the solution.
The research groups of long-standing collaborators Bernd Winter (Fritz Haber Institute of the Max Planck Society), Stephen Bradforth (USC), and Pavel Jungwirth (CAS), all authors on the paper, teamed up to study and help explain the mechanism behind the metallic transition. They collaborated remotely to design and perform PES measurements on metal–NH3 solutions, with their experimental prototype being assembled in Jungwirth’s lab in Prague. But before they could tackle the question at hand, they had to overcome a critical experimental hurdle: keeping the ammonia solutions liquid under experimental conditions for long enough to make measurements.
Jet setters
The collaborators turned to a microjet technique that has been used to apply PES to water, simple alcohols, and even liquid nitrogen and argon. Fast-flowing jets refresh the liquid in the observation window quickly enough to compensate for evaporation. Ammonia is liquid between −77 °C and −33 °C at atmospheric pressure, so they cooled their samples and the surrounding equipment to −60 °C. But that’s not so cold compared to argon, whose boiling point is −186 °C. The bigger challenge was making sure the apparatus was impeccably clean: Ammonia is reactive, and any insoluble products could clog the micronozzle. With conditions just right, the ammonia formed a stable liquid jet a few centimeters long.4
By the time the researchers reported on pure ammonia jets in 2019, they were already applying the technique to solutions with solvated electrons. As in their previous experiments, they bombarded microjets of the solutions with high-brilliance x rays from the BESSY II synchrotron radiation source.
Figure 2 shows the photoelectron spectra for lithium–ammonia samples at concentrations from 0.08 to 9.7 MPM, spanning the metallic transition. For the lowest concentrations, 0.08 and 0.35 MPM, a single peak corresponds to the energy needed to kick a solvated electron out of its ammonia cage. Then, around 1 MPM—the concentration at which conductivity starts increasing—the spectrum develops a more complicated form. The researchers showed that contributions from two electron populations, those in localized and extended states, explain the change. A Gaussian peak accounts for isolated electrons, and a free-electron gas model accounts for metallic ones. The free-electron gas model, commonly used to describe metals, contributes a conduction band and plasmon resonances.
Photoelectron spectra of lithium–ammonia solutions show an evolution from electrolytic to metallic behavior. At low mol percent metal (MPM), dissociated electrons sit in localized cavities surrounded by NH3 molecules. A single spectrum peak reflects the binding energy associated with those cavities. As the MPM increases, the electrons form extended wavefunctions like those found in metals. Models (solid lines) show that the increasing prevalence of a free-electron conduction band and collective oscillations, or plasmons, account for the spectrum’s peaks. (Reprinted from ref. 2, with permission.)
Photoelectron spectra of lithium–ammonia solutions show an evolution from electrolytic to metallic behavior. At low mol percent metal (MPM), dissociated electrons sit in localized cavities surrounded by NH3 molecules. A single spectrum peak reflects the binding energy associated with those cavities. As the MPM increases, the electrons form extended wavefunctions like those found in metals. Models (solid lines) show that the increasing prevalence of a free-electron conduction band and collective oscillations, or plasmons, account for the spectrum’s peaks. (Reprinted from ref. 2, with permission.)
The fraction of isolated electrons needed to account for the observed spectra decreases with increasing metal concentration. At the highest concentration, 9.7 MPM, the spectrum is entirely described by the free-electron gas model, and the conduction-band peak has the predicted sharp edge at the Fermi energy. A plasmon peak in the visible range accounts for the solution’s bronze, rather than silver, appearance. The 9.7 MPM spectrum’s shape closely mirrors that of 50–50 sodium–potassium, a more standard liquid metal that the researchers tested for comparison.
Structural shifts
Ab initio molecular dynamics simulations performed by coauthor Ondřej Maršálek, an assistant professor at Charles University in Prague, paint a more detailed molecular picture. He found that both isolated and spin-paired electrons in solution are surrounded by a diffuse layer of approximately 12 ammonia molecules (see figure 3). The cavities confining solvated electrons in ammonia were about 3.9 Å in diameter for a single electron and 4.4 Å for paired electrons.
Molecular dynamics simulations provide details about the size and structure of the cavities that house electrons solvated in ammonia. Spin-paired electrons, shown in green, are surrounded by approximately 12 ammonia molecules (blue and white). Those molecules form cavities about 4.4 Å across; the cavities around unpaired electrons are approximately 3.9 Å. (Courtesy of Tomáš Martinek and Ondřej Maršálek.)
Molecular dynamics simulations provide details about the size and structure of the cavities that house electrons solvated in ammonia. Spin-paired electrons, shown in green, are surrounded by approximately 12 ammonia molecules (blue and white). Those molecules form cavities about 4.4 Å across; the cavities around unpaired electrons are approximately 3.9 Å. (Courtesy of Tomáš Martinek and Ondřej Maršálek.)
One simple explanation for the onset of metallic behavior is the Mott criterion, which says that metallic conduction should arise when the average distance between the diffuse solvated electrons is less than approximately four times their extent. At 1 MPM, that distance would be about 4 Å. But that picture is too simplistic—such a transition would be abrupt, whereas the Li–NH3 spectrum evolves gradually. Unfortunately, there isn’t another system with an analogous transition to guide the researchers’ thinking. Based on their data, they suspect that coalescence of electron cavities into a continuous network underlies the transition to a metallic state. But at the moment that’s just a guess.
Pinning down what’s really happening in the transition will require improved simulations. Maršálek had already upgraded to a more accurate electron density functional and a more extensive set of basis functions compared with his earlier studies of water.5 Both were necessary to reproduce the known bound state, which is more diffuse in ammonia than water, but they increased the simulation time by up to three orders of magnitude.
Extending simulations to higher metal concentrations will require Maršálek to include cations, which he reasonably neglected at low concentrations. The main challenge is that ammonia solutions are dynamic and disordered. Fully simulating the diffuse electrons, metal cations, and solvent molecules in all their possible configurations requires a lot of sampling, which makes the already expensive computation even more daunting.
Despite its challenges, the inclusion of cations is an important next step. At increasing metal concentrations, they’re likely to be important to the formation, structure, and dynamics of an electron network. The researchers suspect that, like the atomic lattice in crystalline conductors, the cations provide the background potential for extended electron states.
Acquiring complementary experimental information on electronic structures throughout the metallic transition would require neutron or x-ray scattering, and the collaboration’s experimentalists hope to attempt those measurements. But that doesn’t mean they’re done with PES; studies on highly concentrated electrons in water are already in the pipeline.
