The search for tetraoxygen, which dates back to a study by Lewis in 1924, [1] is being actively pursued, owing to its fundamental interest, its potential role as a material of high energy density, [2] and, in particular, the current upsurge of interest in excited states of O 2 with relevance to atmospheric chemistry and to terrestrial and Venusian nightglow. [3] Numerous calculations at different levels of theory predict the existence of metastable O 4 molecules with a D 2d cyclic geometry [4] or a D 3h ªpinwheelº geometry. [5,6] Positive experimental detection of such species has not been reported, whereas weakly bonded (O 2 ) 2 van der Waals dimers, with dissociation enthalpies of less than 1 kcal mol À1 , have long been known and thoroughly characterized. [7] Indirect evidence for metastable tetraoxygen was derived from experiments involving electron transfer from Cs to O 4 , followed by the measurement of the kinetic energy spectrum of the O 2 fragments. The structure of the kinetic energy distribution was consistent with the intervention of some O 4 species that were not observable under the conditions of the experiment. [8] Strong evidence for metastable O 4 was derived from studies involving the photoionization of O 2 excited by a DC discharge. [9,10] Interestingly, the results of both experimental studies were interpreted as suggestive of the existence of a third metastable O 4 species in addition to the theoretically predicted molecules, namely, a relatively long lived complex between a ground-state O 2 molecule and an O 2 molecule in the excited c 1 S À u electronic state. [10] To obtain conclusive proof, we have carried out an experiment aimed at detecting intact O 4 . Our approach is also based, like previous studies, [8] on the neutralization of O 4 cations in the gas phase, but uses for the identification of the intact O 4 molecule neutralization reionization (NR) mass spectrometry [11] on a highly sensitive instrument that allowed the detection of other elusive atmospheric species, such as hydrogen trioxide [12] and the [H 2 O ´O 2 À ] charge-transfer complex. [13] The O 4 ions were generated in a chemical ionization (CI) source by association of O 2 molecules with O 2 primary ions, formed both in the X 2 P g ground state and in electronically excited states by electron impact. [14] The ions were then accelerated through 4 ± 8 keV, magnetically mass resolved, and probed by collisionally activated dissociation (CAD) mass spectrometry. Whereas the CAD spectrum of 16 O 4 from the ionization of 16 O 2 shows 16 O 2 as the charged fragment, 16 O 2 18 O 2 ions from 16 O 2 / 18 O 2 mixtures give 16 O 2 and 18 O 2 without the 16 O 18 O ions [15] that would suggest isotopic scrambling ( Figure 1). This shows that the O 4 ions probed contain two discrete O 2 units, each of which retains its original identity.In the NR experiments the O 4 ions were also accelerated to 4 ± 8 keV and mass-selected before undergoing two consecutive collision events in separate cells aligned along the beam path. In the first c...