Chlorine Trifluoride: The Gas That Sets Sand on Fire

The molecule: three fluorines, two lone pairs, one T

Chlorine brings seven valence electrons. Three of them go into bonds with fluorine; the remaining four sit as two lone pairs. That is five electron domains around the central atom — VSEPR type AX₃E₂ — so the domains arrange as a trigonal bipyramid. Lone pairs are fatter than bonds and demand the roomier equatorial sites, and with two of the three equatorial positions occupied by lone pairs, what remains is a T: one equatorial fluorine flanked by two axial ones, point group C_2v.

Microwave spectroscopy pins it down. The equatorial Cl–F bond is 1.598 Å; the two axial bonds are noticeably longer at 1.698 Å; and the F(axial)–Cl–F(equatorial) angle is 87.5°, not 90°. The axial arms are bent 2.5° away from the lone pairs and toward the short equatorial bond — exactly the distortion VSEPR predicts, measured to a tenth of a degree.

The length difference is not decoration. The axial F–Cl–F unit is best described as a three-centre four-electron bond: one p orbital on chlorine shared across two fluorines, giving roughly half a bond order each. Half-order bonds are long and weak, and that weakness is the entire story of this molecule. Electronegativity does the rest: fluorine at 3.98 against chlorine at 3.16 leaves chlorine strongly δ+ and every fluorine a poised, loosely-held δ− leaving group.

Weak bonds break, very strong bonds form

The violence is arithmetic, not magic. Take the standard enthalpy of formation, ΔH_f°(ClF₃, g) = −163.2 kJ/mol, and atomise the molecule: ΔH = ΔH_f(Cl,g) + 3 ΔH_f(F,g) − ΔH_f(ClF₃) = 121.3 + 3(79.0) + 163.2 = 521.5 kJ/mol. Divide by three and the mean Cl–F bond is ~174 kJ/mol. For comparison: Cl–Cl is 243, O=O is 498. Among halogen–halogen bonds only F–F (159 kJ/mol) and I–I (151 kJ/mol) are weaker.

Now look at what those fluorines can build. Si–F is ~565 kJ/mol in the usual bond tables (atomising SiF₄ thermochemically puts it nearer 595 — sources differ by the usual 5%, and either figure makes the point). It is among the strongest single bonds known to chemistry. A reaction that trades 174 for 565, four times over, cannot fail to be exothermic — though against silica you must first pay to break Si–O bonds worth ~465 kJ/mol each, which is exactly what the Hess's-law sum in the next section accounts for.

So sand reacts:

• 3 SiO₂ + 4 ClF₃ → 3 SiF₄ + 2 Cl₂ + 3 O₂, ΔH = −1,460 kJ, or −487 kJ per mole of silica.

Watch the electron bookkeeping, because it defeats most people's intuition. Silicon is +4 before and +4 after — it is not oxidised at all, it merely swaps partners. Chlorine falls from +3 to 0, taking twelve electrons across the four ClF₃ molecules. Those twelve electrons come from oxygen, which climbs from −2 to 0 and leaves as O₂. In this fire the oxygen in the sand is the fuel, and the gas is the oxidiser. That inversion is why it feels impossible.

Working the numbers: enthalpy, entropy, and a 3,900 K flame

The −1,460 kJ comes straight out of Hess's law with tabulated formation enthalpies: SiO₂(quartz) −910.7, ClF₃(g) −163.2, SiF₄(g) −1615.0, with Cl₂ and O₂ at zero by definition.

ΔH = 3(−1615.0) − [3(−910.7) + 4(−163.2)] = −4845.0 + 3384.9 = −1460.1 kJ, i.e. −486.7 kJ per mole of SiO₂.

Per mole of silicon the accounting is transparent: you collect −1615 kJ forming SiF₄, you pay back +910.7 kJ to undo the silica, and the ClF₃ you consume returns +217.6 kJ of its own formation enthalpy. Net −487 kJ. Run the same sum on bare silicon and there is no 910.7 to pay: 3 Si + 4 ClF₃ → 3 SiF₄ + 2 Cl₂ gives −1,397 kJ per mole of Si, nearly three times as much. That single difference explains why fabs can use this gas at all — elemental silicon is enormously more reactive than its oxide, so ClF₃ strips silicon deposits at microns per minute while quartz chamber windows survive.

Entropy piles on. Three moles of solid plus four of gas become eight moles of gas: ΔS ≈ +660 J/K, so ΔG(298 K) ≈ −1,460 − 298(0.660) ≈ −1,657 kJ. And because every product is a gas, no slag or melt forms to smother the reaction — the sand does not melt, it leaves, exposing fresh surface continuously. Dump all 1,460 kJ into the eight moles of product gas (ΣC_p ≈ 374 J/K) and the adiabatic rise is ~3,900 K. Real flames run cooler, since the products begin dissociating at those temperatures, but the order of magnitude holds: this is a cutting torch carrying its own oxidiser.

Hypergolic with no measurable delay

Ordinary combustion must start by breaking an O=O bond worth 498 kJ/mol — that is what a spark or a glow plug buys you, and why sawdust sits safely in air. ClF₃ asks for nothing: its first bond-breaking step costs about 174 kJ/mol, and every step after is downhill. The barrier to attacking a C–H bond, an Si–O bond or a bare metal surface is low enough that room temperature clears it.

The propellant community calls this hypergolic — igniting on contact, no igniter. Hypergolicity is quantified as ignition delay, measured in the open-cup drop test: a droplet of oxidiser falls onto fuel under a high-speed camera, and you count frames until first light. Storable bipropellants such as nitrogen tetroxide with hydrazine or MMH land at 5–50 ms depending on propellant temperature, cup geometry and additives, and mission specifications typically demand under 50 ms so that unburnt propellant does not accumulate and detonate.

Chlorine trifluoride has never produced a number. At the few thousand frames per second these tests run at, the first frame after contact is already burning, so the delay is recorded as under 1 ms — never measured, meaning below instrument resolution rather than known. It is hypergolic not only with rocket fuels but with cloth, wood, leather, asbestos, glass wool and, as John D. Clark dryly noted, test engineers.

Holding it: nickel, Monel, and a film of its own fire

The trick to containing ClF₃ is to let it burn the container once, under control. Nickel 200, Monel 400, Inconel, copper and even carbon steel react on first exposure to grow a coherent metal-fluoride skin — NiF₂, CuF₂, FeF₃ — which is itself already fully fluorinated and therefore inert. New systems are deliberately passivated: purge with dry nitrogen, then admit ClF₃ or F₂ diluted in N₂ at a few percent, stepping the concentration up over hours. Nickel and Monel hold to a few hundred °C; above that the film loses to the reaction and the metal itself burns.

The film is thin and brittle, so ClF₃ practice is oxygen-service practice. Systems are cleaned to hydrocarbon-free standards (ASTM G93 cleanliness, ASTM G88 design guidance): a fingerprint left inside a valve is a fuel deposit. Velocities are limited, because particle impact strips passivation and starts a metal fire. Gaskets are metal, not elastomer. Nitrogen is the standard purge gas — argon and helium have no chemistry to offer either, but nitrogen is cheap and, despite having a fluoride of its own, safe for a thermodynamic reason rather than a habit: 4 ClF₃ + 2 N₂ → 4 NF₃ + 2 Cl₂ is endothermic by about +124 kJ, so nitrogen will not react. Carbon dioxide will: 4 ClF₃ + 3 CO₂ → 3 CF₄ + 2 Cl₂ + 3 O₂ runs about −320 kJ per mole of CO₂. A CO₂ extinguisher aimed at a ClF₃ fire is adding fuel.

It ships as UN 1749, Division 2.3 toxic gas with subsidiary 5.1 oxidiser and 8 corrosive, in nickel or Monel cylinders under little more than 1 bar of its own vapour. Exposure limits are severe: IDLH 20 ppm, with a workplace ceiling of 0.1 ppm. The vapour is 3.2× denser than air, so it pools in pits and trenches, and its hydrolysis products are dominated by hydrogen fluoride — which means a ClF₃ casualty is treated as an HF casualty, calcium gluconate and all.

Making it, measuring it, specifying it

Otto Ruff and H. Krug made it first in 1930 by the route still used: Cl₂ + 3 F₂ → 2 ClF₃ at 250–300 °C in a nickel or Monel reactor, with fractional distillation to separate it from ClF and unreacted chlorine. The 11.8 °C boiling point makes that distillation easy, which is one reason ClF₃ rather than ClF₅ became the industrial interhalogen.

Analysis is done without glass wherever possible, since any moisture generates HF that will attack silica. Gas-phase FTIR is the workhorse: a C_2v triatomic-plus molecule of this shape has six fundamentals, all infrared-active, and the spectrum doubles as an in-line purity monitor for the usual contaminants — ClF, Cl₂, HF, ClO₂F and above all water, which electronic-grade specifications hold to the sub-ppm level. ¹⁹F NMR is the structural check: the T-shape should give axial and equatorial fluorines in a 2:1 ratio, but rapid exchange averages them to a single line at room temperature, and the two environments only separate on cooling.

There is a piece of metrology hiding here. Because ClF₃ fluorinates refractory materials completely and cleanly, it became a combustion aid in fluorine bomb calorimetry: burn a stubborn carbide, boride or silicide in ClF₃ inside a nickel bomb, measure the heat, and you have the formation enthalpy of a compound that will not burn in oxygen. It also strips uranium and its oxides as volatile UF₆, which is why enrichment plants keep it on hand to clean deposits out of equipment.

N-Stoff, a hole in a concrete floor, and what ClF3 is not

Nazi Germany chased it as N-Stoff, building a plant at Falkenhagen in Brandenburg for an incendiary meant to burn through fortifications. A few tens of tonnes were made before the Red Army took the site in 1945; it was never fielded, because the handling problem defeated the weapons problem. After the war the United States evaluated it as a storable rocket oxidiser under the name CTF. On paper it was ideal — dense at 1.77 g/cm³ against 1.44 for nitrogen tetroxide, storable at room temperature, hypergolic with everything. Nitrogen tetroxide won anyway: a propellant that ignites its own plumbing has no operational future.

John D. Clark's Ignition! (1972) preserves the incident everyone quotes: a ~1-tonne spill burned through 12 in (30 cm) of concrete and 3 ft (90 cm) of gravel beneath it. Note the direction. The liquid pooled, the fire ate downward through the floor, and nothing anyone could spray on it would help — water hydrolyses ClF₃ explosively (formally ClF₃ + 2 H₂O → 3 HF + HClO₂, in practice a mess of HF, HCl, Cl₂, ClO₂ and O₂), sand burns, and CO₂ burns. The correct response to a ClF₃ fire is to evacuate, cool the surroundings from a distance, and let it finish.

Three misconceptions are worth killing. It is not chlorine gas: Cl₂ is a middling oxidant that leaves dry silica alone, and ClF₃'s chemistry is fluorine chemistry wearing a chlorine badge. It is not pyrophoric: pyrophoric substances such as white phosphorus or silane are fuels that air ignites, whereas ClF₃ is indifferent to air because air offers it nothing to oxidise — it is the oxidiser, and the sand is the fuel. And it is not the strongest oxidiser known: ClF₅ and O₂F₂ both outrank it, and elemental fluorine has a higher oxidation potential. ClF₃'s real distinction is more unsettling than being strongest — it is aggressive enough to burn a bunker and tame enough to keep in a cylinder at room temperature.

Chlorine trifluoride against the species it is routinely confused with

Species	Boiling point	Shape and oxidation state	Behaviour toward silica and steel
Cl<sub>2</sub> (chlorine gas)	−34.0 °C	Linear, Cl 0	Attacks hot steel; dry silica is untouched. Not a fluorinating agent.
ClF<sub>3</sub> (chlorine trifluoride)	11.8 °C	T-shaped C<sub>2v</sub>, Cl +3	Ignites silica, concrete, glass wool and cloth; passivates nickel and Monel.
ClF<sub>5</sub> (chlorine pentafluoride)	−13.1 °C	Square pyramidal C<sub>4v</sub>, Cl +5	Stronger fluorinator still, but harder to synthesise and store.
F<sub>2</sub> (fluorine)	−188.1 °C	Linear, F 0	Stronger oxidant on paper, but cryogenic and less dense as a stored liquid (1.50 vs 1.77 g/cm³).
O<sub>2</sub>F<sub>2</sub> (dioxygen difluoride)	−57 °C (extrapolated), decomposes above −160 °C	Skewed C<sub>2</sub>, O in a rare +1	Fluorinates almost anything, but only exists at cryogenic temperature.
P<sub>4</sub> (white phosphorus)	280.5 °C	Tetrahedral, P 0	Genuinely <em>pyrophoric</em> — it is the fuel that air ignites, the opposite role.