Free gluons?

gluons are not colorless particles, so they are subject to the same rules of confinement as quarks. thus you will never see a free gluon in the lab.

on the other hand, when you smash a bunch of particles together hard enough, if you see 4 jets (since only 3 of them can be quarks), that s a sure sign that one of the particles spit out a gluon, as one of the out states of the reaction, instead of just a virtual particle binding the quarks together.

this experiment has been done.

 

Gluons have colour too so they are confined in a simlair manner to quarks, but they can in theory exist together without quarks as a colourless glueball (a collection of gluons) which would be a (flavour-neutral) meson.

 

glueballs

In theory, glueballs should behave very much like regular mesons. Glueballs have a couple of restrictions; they must be isoscalar, they must not have any flavor (just like jcsd said), they will not couple to photons well, and they will not have branching ratios compatible with SU(3) mesons. They will also tend to be supernumerary to the other members of their associated SU(3) meson nonets, and have widths that are unusually narrow for their mass.

It is predicted that the scalar ( 0+(0++) ) glueball will occur at about 1611 MeV, and the tensor ( 0+(2++) ) glueball will occur around 2232 MeV. I have been trying to extract two unknown peaks (practically just bumps) from E760 neutral data (Fermilab 1994) in ppbar annihilation, and they appear awfully close to those values. Unfortunately, the fit they used on their original data kind of leveled out the bumps, so I had to start from scratch on the analysis.

A couple of good existing candidates for the ground state (scalar) glueball are the f0(1500) and f0(1710) scalar mesons (thats the 1(3)P0 nonet). Check out the Physical Review D section on Non-qqbar Candidates (http://pdg.lbl.gov/). It basically says everything I've said plus all the details and other indications.