Abstract
After an introduction to inversion layers and a summary of the theoretical ideas concerning the expected effects of disorder in such systems, the observed behaviour near the threshold of conduction is reviewed. Although the system shows a classic metal-insulator transition, the detailed behaviour is not consistent with an independent-particle mobility edge model. A new model is proposed in which correlation causes the carriers to behave like a viscous liquid near the conduction threshold. The new model provides quantitative explanations of many observed properties. Other experiments are suggested by which it may be further tested.