Solutions of Alkali Metals in Liquid Ammonia

Metal–ammonia solutions remain a working reagent, not merely a historical curiosity. The Birch reduction — reduction of aromatic rings to non-conjugated 1,4-cyclohexadienes using an alkali metal (typically Na or Li) in liquid ammonia with an alcohol proton source — is a textbook and industrial method for accessing partially saturated ring systems used throughout steroid and terpenoid synthesis, and it proceeds through exactly the radical-anion intermediates generated by single-electron transfer from the solvated electron to the aromatic π system. The same solvated-electron chemistry underlies dissolving-metal reductions of alkynes to trans-alkenes (a stereochemical outcome complementary to catalytic syn-hydrogenation), reductive cleavages of benzylic C–O and C–N bonds in protecting-group chemistry, and acyloin condensations of esters. Beyond synthesis, concentrated alkali-metal–ammonia and related metal–amine solutions are the classic route to alkalides and electrides — extraordinary compounds in which an alkali-metal cation is sequestered by a crown ether or cryptand and the counter-anion is either another alkali-metal anion (M⁻, an "alkalide") or a bare, trapped electron occupying a crystallographically defined cavity (an "electride," with no anion at all in the conventional sense). These compounds, first isolated by extending exactly the electron-solvation chemistry developed for metal–ammonia solutions into complexing agents that could stabilise the free electron or metal anion at room temperature in the solid state, are studied as candidate low-work-function materials, exotic magnetic solids, and extreme reducing agents for materials synthesis. The solvated electron itself, as first characterised quantitatively in metal–ammonia solutions, is also the central reactive species in radiation chemistry: pulse radiolysis of pure water or ammonia generates the same species (with an analogous, though blue-shifted, absorption band in water), and its reactions with radicals, ions, and biomolecules underpin radiation biology, dosimetry, and the design of radioprotective agents. Finally, the physics of the concentration-driven insulator-to-metal transition documented so cleanly in Na–NH₃ solutions is a benchmark system for the general Mott-transition problem in condensed-matter physics, informing theoretical treatments of doped semiconductors, expanded liquid metals, and other systems where carrier density alone drives a qualitative change between localized and itinerant electronic behaviour.

Ask the tutor: How does the solvated-electron mechanism in Na/NH₃ actually deliver two electrons, one at a time, to reduce an aromatic ring in the Birch reduction?

Liquid ammonia is an unusually good medium for stabilising both cations and a novel anionic species — the free electron — because it combines several favourable properties at once: a reasonably high dielectric constant (ε ≈ 22 at the boiling point, versus 78 for water), strong Lewis basicity at nitrogen enabling efficient cation solvation (Na⁺ is comparably or better solvated by NH₃ than by H₂O), and, crucially, a much weaker tendency to protonate a free electron than water has. Water's O–H bond is far more acidic in the presence of an electron pair than ammonia's N–H bond; the reaction

is thermodynamically downhill but has a substantial activation barrier in pure, metal-ion-free ammonia, so that in the absence of catalytic surfaces or paramagnetic impurities the blue solution is metastable for hours to days. This kinetic protection is the single most important practical fact about the system: it is what makes Na/NH₃ solutions usable as reagents at all, rather than merely a beautiful but instantaneous transient. When an alkali metal atom is exposed to the solvent, its lone valence s electron is not transferred to a neighbouring molecule (there is no good acceptor orbital at the modest electron affinity ammonia offers) but instead is "released" into the liquid, where it polarises the surrounding NH₃ dipoles to form its own solvation shell — precisely as a cation or anion would. The picture is often drawn as a spherical cavity of radius on the order of 3–4 Å, from which the electron density leaks only slightly into the surrounding solvent, with the ammonia molecules arranged so that their partially positive N–H hydrogens point toward the cavity centre (analogous to how water orients around a real anion). This is the origin of the term solvated electron, written e⁻(am) or e⁻(NH₃)ₙ. Because the trapped electron behaves like a particle confined to a roughly spherical potential well, its allowed energies are quantised in a manner reminiscent of the particle-in-a-box or hydrogenic problem, with a ground state of predominantly s character and low-lying excited states of p-like character. The energy gap between these levels falls in the visible/near-infrared, and the resulting broad absorption band (peaking near 1500 nm in the near-IR, with an extended tail into the visible) is precisely what makes light of most visible wavelengths pass through while a portion is absorbed, imparting the characteristic intense blue colour — deep blue in transmitted light because the strong absorption is centered in the red/near-IR, leaving blue light relatively unattenuated.

As metal concentration rises, the picture changes qualitatively. At low concentration (below roughly 0.05 mol L⁻¹) the solutions behave as a dilute "electron gas" dissolved among ordinary ion pairs — paramagnetic, with magnetic susceptibility corresponding closely to one unpaired electron per formula unit, and only modestly conducting. As concentration increases past about 0.5–1 M, the average separation between solvated-electron cavities and metal cations shrinks to the point that electron wavefunctions on neighbouring sites begin to overlap, cavities cluster into diamagnetic electron pairs (which show a distinct, blue-shifted absorption sometimes called the "monomer/dimer" equilibrium), and eventually — above the Mott density around 4 mol % metal (roughly M/NH₃ molar ratio near 1:60) — the electronic wavefunctions merge into a continuous, singly-occupied conduction band. The solution then acquires metallic lustre (an actual bronze-coloured film separates from the blue liquid, sometimes as a distinct liquid layer above ~20 mol % metal), metallic conductivity (10³–10⁴ Ω⁻¹cm⁻¹, comparable to liquid mercury), and near-temperature-independent Pauli paramagnetism rather than Curie-law paramagnetism.

The energetics of dissolution can be dissected using a Born–Haber-style cycle exactly parallel to the one used for ionic dissolution or gas-phase ionisation, but with the crucial addition of an electron-solvation term instead of an electron-affinity/lattice term:

where: $\Delta H_{\text{sub}}$ is the sublimation enthalpy of the solid metal (atomisation), $IE_1$ its first ionisation energy, $\Delta H_{\text{solv}}(M^+)$ the (negative, favourable) solvation enthalpy of the cation by ammonia, and $\Delta H_{\text{solv}}(e^-)$ the (negative) "solvation" enthalpy of the free electron into its cavity state. This cycle is the direct ammonia analogue of the familiar Born–Haber treatment of ionic solids, and it makes explicit why the process is possible at all: the two unfavourable gas-phase steps (breaking metallic bonding, ionising the atom) are overcompensated by two large solvation terms, one of which — electron solvation — has no counterpart in classical ionic solubility problems and is unique to a solvent that can stabilise a bare electron. The absorption energy of the trapped electron is well captured, to first approximation, by treating it as a particle in a finite (or infinite, as a limiting case) spherical well of radius $r$:

where: $n$ is a principal-like quantum number labelling the s, p, … cavity levels, $\hbar$ is the reduced Planck constant, $m_e$ the electron mass, and $r$ the effective cavity radius. This reading immediately explains two qualitative trends: the transition energy (and hence colour) is sensitive to cavity size, which is why the absorption maximum shifts somewhat with temperature and solvent (larger, "looser" cavities at higher temperature red-shift the band, exactly as a bigger box lowers the level spacing); and because $E \propto 1/r^2$, modest changes in solvent structure around the electron produce large, easily observed spectral shifts — the basis of using solvated-electron spectroscopy as a sensitive local structural probe. The magnetic behaviour is governed by a smooth interpolation between localized-spin (Curie) paramagnetism and delocalized (Pauli) paramagnetism as concentration $c$ increases. The molar magnetic susceptibility of the dilute solution follows the Curie law expected for a single unpaired spin per formula unit,

where: $N_A$ is Avogadro's number, $g \approx 2.00$ the electron g-factor, $\mu_B$ the Bohr magneton, $S = 1/2$ the electron spin, $k_B$ Boltzmann's constant, and $T$ the temperature. This reading is that the dilute limit behaves like a paramagnetic gas of independent, uncoupled spin-½ centres — exactly the "free radical" behaviour expected if each dissolved metal atom truly deposits one unpaired electron into its own cavity, uncorrelated with any other. As concentration rises, increasing overlap between neighbouring electron wavefunctions produces spin-pairing (singlet dimers, or delocalisation into a spin-compensated band), so that the effective number of unpaired spins per mole of metal falls well below 1, and in the metallic limit the susceptibility becomes small, positive, and only weakly temperature-dependent — the hallmark of Pauli paramagnetism of a degenerate electron gas.

Example 1 (foundational) — locating the concentration regimes for a laboratory-scale Na/NH₃ solution. A student dissolves 0.046 g of sodium metal (atomic mass 22.99 g mol⁻¹) in 50.0 mL of liquid ammonia (density of liquid NH₃ at −33 °C is 0.6819 g mL⁻¹, molar mass 17.03 g mol⁻¹). Determine the molar concentration and decide, using the ~1–4 mol L⁻¹ Mott-transition benchmark developed above, whether this solution should be in the dilute (blue, paramagnetic) or concentrated (bronze, metallic) regime. Moles of Na:

Volume of solvent (treating the small metal volume as negligible, a standard approximation at this scale):

Molar concentration:

This value, 0.040 M, lies comfortably below the ~1 mol L⁻¹ threshold estimated for the Mott/overlap transition and even below the conductivity-minimum region (~0.5 M), so the solution should show the dilute-regime signatures: an intense blue colour, Curie-law paramagnetism close to one unpaired electron per Na, and modest (non-metallic) molar conductivity. As a limiting-case sanity check: if the student had instead dissolved 500-fold more sodium (23 g, essentially one mole, in the same 50 mL), the concentration would be roughly 20 mol L⁻¹ — deep into the saturated, metallic-bronze regime — confirming that the formula correctly predicts a qualitative change in physical behaviour purely from stoichiometric input.

Example 2 (applied) — energetics of the Born–Haber-style dissolution cycle for potassium. Estimate whether dissolution of potassium metal in ammonia should be thermodynamically comparable to, or more favourable than, sodium, using approximate cycle components: $\Delta H_{\text{sub}}(K) = +89$ kJ mol⁻¹ (versus $+108$ kJ mol⁻¹ for Na), $IE_1(K) = +419$ kJ mol⁻¹ (versus $+496$ kJ mol⁻¹ for Na), and assume the sum of the two solvation terms $\Delta H_{\text{solv}}(M^+) + \Delta H_{\text{solv}}(e^-)$ is roughly $-590$ kJ mol⁻¹ for K⁺/e⁻ pairs and $-640$ kJ mol⁻¹ for Na⁺/e⁻ pairs (K⁺ is a larger, less strongly solvated cation, partially offset by a comparable electron-solvation term that is roughly metal-independent since the cavity electron does not "know" which cation nearby created it). For potassium:

For sodium:

Both are exothermic (favourable), consistent with both metals dissolving readily in liquid ammonia to give deep blue solutions, but the calculation predicts potassium's dissolution to be more exothermic than sodium's by roughly 46 kJ mol⁻¹ — driven mainly by potassium's substantially lower first ionisation energy outweighing its smaller cation-solvation stabilisation. This tracks the experimental observation that potassium (along with the heavier alkali metals) dissolves somewhat more readily and to a higher saturation concentration than sodium. As a limiting-case check, note that if ionisation energy alone controlled the trend (ignoring solvation entirely), one would predict the same qualitative ordering (K more favourable than Na) since $IE_1(K) < IE_1(Na)$ by a wide margin (77 kJ mol⁻¹) that dominates the smaller sublimation-enthalpy difference (19 kJ mol⁻¹) — the full cycle simply refines, rather than reverses, this basic periodic-trend expectation.

Example 3 (exam/competition level) — using the spherical-well model to rationalise a spectral shift. The absorption maximum of the solvated electron in liquid ammonia occurs near $\lambda_{\max} \approx 1500$ nm at −65 °C. On warming toward −40 °C, the band is experimentally observed to red-shift (move to longer wavelength, i.e., lower transition energy). Using the particle-in-a-sphere model, determine what this implies about the change in effective cavity radius, and estimate the fractional change in radius consistent with a red-shift to $\lambda_{\max}' \approx 1650$ nm, assuming the transition is dominated by the $n=1 \to n=2$ (1s → 1p-like) spacing so that $E \propto 1/r^2$ holds with the same proportionality constant at both temperatures. Since transition energy $E = hc/\lambda$ and the model gives $E \propto 1/r^2$, we have $1/\lambda \propto 1/r^2$, i.e.

Taking the ratio between the warmer and colder states:

Substituting the observed wavelengths:

So the model predicts the effective cavity radius must expand by about 4.9% on warming from −65 °C to −40 °C to account for the observed red-shift. This is physically sensible: raising the temperature increases thermal disorder and reduces the degree of ordered dipole packing around the trapped electron, allowing the cavity to expand slightly, which — because $E \propto 1/r^2$ makes the transition energy very sensitive to small radius changes — is enough to produce a clearly measurable shift in $\lambda_{\max}$ even though the geometric change is under 5%. As a limiting-case check, note that if temperature had no effect on cavity size ($r' = r$), the model would predict no shift in $\lambda_{\max}$ at all, contradicting the observed red-shift — confirming that the direction of the effect (expansion, not contraction, on warming) is the physically required one, since a contracting cavity would blue-shift the band instead.

The modern understanding of metal–ammonia solutions rests on foundational spectroscopic and structural work from the mid-twentieth century, but the phenomenon itself was first reported far earlier: Humphry Davy noted the striking blue colour on exposing potassium to ammonia gas in 1808, shortly after his celebrated isolation of the alkali metals themselves by electrolysis. It took over a century of subsequent work — spanning conductivity measurements, magnetic-susceptibility studies, and eventually optical and photoelectron spectroscopy — to establish the cavity/solvated-electron picture now taught as standard.