A traditional line in inorganic-gas spectral analysis is based on using the emission from atoms and molecules on plasma excitation; the first paper on the topic was published almost a hundred years ago [I]. Originally, plasma sources were used only to determine gaseous impurities in air and certain simple gases: hydrogen, oxygen, nitrogen, helium, neon, and argon. More recently, they have been used for determining volatile inorganic compounds (hydrides, halides, oxides, and so on) either as impurities in pure gases or in complicated technological mixtures. The methods here form two groups: for binary or pseudobinary gas mixtures and for complex ones. In the first, there are methods in which there are no effects on the signal from the undetermined impurities, and the calibration curves uniquely reflect the relation between the contents for the test and major components. In the second, there are methods where the variations in the contents of undetermined or third components have substantial effects. This classification or third components have substantial effects. This classification differs from the traditional one, where methods are distinguished by purpose (pure gases and complex mixtures), but it seems justified because it is based on the plasma processes. Binary (Pseudobinary) Mixtures. Topics here have mainly been dealt with in [2] and in certain later papers [3-7], so here we nlerely briefly discuss the major results given in Tables 1 and 2 for three lines: determining components difficult to excite, determining ones readily excited, and analyzing molecular-gas mixtures having similar excitation potentials. The most sensitive form of spectral analysis concerns determining readily excited trace components in major ones difficult to excite, where the qualitative composition determines where Cs the detection limit, falls in the range from 10 -2 to i0 -s mol.% (I in Table i). The best results here are from exciting the spectrum in a high-frequency discharge in flowing gas in a tube i-2 mm in diameter near atmospheric pressure [3-5]. The random error as a rule does not exceed 5-10% relative but rises to 30% relative at the lower limit. Determining trace components difficult to excite in readily excited matrices. The scope for spectral analysis here is somewhat limited. The detection limits are usually no better than 10 -I mol.%, but the best results (II in Table I) have been obtained at low pressures (P ~ 1 kPa) in flowing gases with discharge tubes not more than i mm in diameter. Analyzing mixtures of molecular gases having comparable excitation potentials. Table 2 shows the scope here; as in the previous case, the detection limits are comparatively poor. There is a fairly simple method of lowering the detection limits: diluting the mixture by about a factor of 100 with pure helium , which sometimes increases the spectrum intensi~y bya factor of 5-10 [8]. There is usually no advantage from further dilution with helium, firstly because the intensity falls because the analytical species is reduced in concentration and secondly because impurities in the diluting gas begin to have an effect.