In ICP-OES (Wikipedia page), you introduce a sample into a very hot plasma and then run a spectral analysis on the resulting light emissions. Liquid samples are sprayed into the flame. From the Wikipedia article:
A peristaltic pump delivers an aqueous or organic sample into an analytical nebulizer where it is changed into mist and introduced directly inside the plasma flame.
From that I can imagine that you would probably dissolve solid samples (the technique is widely used in metallurgy and mineralogy) in something suitable to obtain a liquid sample. If that's correct, I'm wondering how the feeding apparatus is constructed to withstand possibly aggresive solvents, like nitric acid or aqua regia. The injection tube is apparently made from fused silica, I guess that's a probable material for the feed as well.
And how about gases? Are they analysed with ICP-OES as well? Are they simply pumped into the injection tube? For example, can we just introduce a stream of air into the plasma and get the emission spectrum for 78% nitrogen, 21% oxygen (I guess the argon will be hidden because the plasma gas is argon)? I guess since every sample will be beyond vaporized anyway by the intense heat, it can't make much difference, but are gas samples too "thin", i. e. do they contain too little atoms per volume for useful analysis?