INTRODUCTIONThe glow discharge lamp used in glow discharge optical emission spectrometry (GDOES) combines a high-speed sputtering source and an intense emission source in one. This gives the lamp its power, compactness and speed to analyse and depth profile practically any solid material compatible with vacuum. Paradoxically, it is this combination of sputtering and emission that makes the technique so difficult to describe theoretically. The sputtering supplies atoms from the sample into the glow region of the source where the optical lines are emitted. The intensity of an emission line therefore varies with the concentration of the element in the sample, with the sputtering rate and with the emission process. It is difficult to know, when the intensity changes, which is the cause of the change : concentration, sputtering or emission.The analyst is therefore left largely with empirical means to obtain quantitative results. These empirical means are complicated by the inter-relatedness of the lamp parameters : current, voltage and pressure.'.' One can fix only one of the these three variables at a time and vary the other two. Hence, one is constantly unsure which of the two varying parameters-or perhaps a combination of both-is responsible for any variation in sputtering or emission.Several empirically based quantitative schemes are now available for GDOES.3-'3 Of particular note here are: (1) the method of Bengston et al., developed at the Swedish Institute for Metals Research (SIMR), of correcting intensities for voltage and current changes in the lamp;" (2) the method developed separately by Cazet and Hocquaux at the Institut de Recherches de la Sid6 rurgie Frangaise (IRSID), of varying the argon flow to ensure that current and voltage do not change with depth;14 and (3) the approach of Payling and Jones at BHP of keeping pressure constant to ensure that the emission yield is ~onstant.''*'~ These methods were discussed in more detail in Part I." The first two methods assume that pressure does not have a significant effect on emission yield but that current and voltage do, while the third method assumes that pressure is the only significant parameter affecting emission yield.The SIMR and BHP models each give a good description of how the intensity of a given emission line from one matrix varies as a function of current or voltage. This is illustrated in Fig. 1 for iron from brass standard 306C. The models work well with one matrix because they are clearly expressing different versions of the same current-voltage-pressure relationship inside the lamp.These varying approaches pose several obvious questions : how does emission yield really vary with current, voltage and pres?ure, and, from that, which quantitative scheme really is the best one? How does one separate the effects of parameters that are correlated with each other? How does one design an experiment, a decisive experiment, that shows precisely how the intensities in the GDOES lamp depend separately on current, voltage and pressure? Deep down these ques...