`` If the brain were so simple we could understand it, we would be so simple we couldn't.''Lyall Watson
Thermal donors remain one of the great unsolved problems in defect physics. First discovered over 45 years ago, their microscopic structure still remains a mystery. They are a family of up to 17 double donor defects  that appear in oxygen-rich Si (TD1 TD17) when annealed in the temperature range 300-550 C [227,228,229]. Higher temperature anneals lead to their destruction but also the growth of other, `new donor' centres. In addition, higher temperature anneals lead to quartz precipitation, and associated build-up of Sii and prismatic dislocation loops. Therefore Si is often subjected to rapid high temperature anneals to break up these oxygen clusters followed by quenching before other defect centres have time to form.
There is incentive to understand thermal donors from two camps. Firstly decreased tolerance chip design means that future processing will demand lower temperature anneals and less processing steps, which means that a thorough understanding of the thermal donors will become increasingly important. Secondly, despite being probably the most studied point defects in semiconductor materials to date they have remained intransient, and thus understanding them provides one of the greatest challenges to the point defect research community.
Experimental examination has revealed no other impurity atoms in these centres apart from Oi, and so it appears that they can only consist of Oi, possibly Sii or V, and lattice Si. Any complete model appears doomed to conflict with some of the experimental evidence, given the amount of material collected over the years. Hence we attempt to postulate here a model which agrees with the majority of results currently available; there are no doubt exceptions in the literature.
In this chapter we consider a number of different models for the thermal donor based around only Oi and lattice Si, and show that a model consisting of four Oi atoms, based around a `di-y-lid' core, can account for many of the experimental observations. We discuss possible formation methods for these defects.