rial has been developed based upon density-functional simulations rather than trial-and-error preparation procedures. In addition, we also expect that materials with N sp above 4.25 will also be good PC materials. This is in excellent agreement with the data compiled in Table 1. In this table we have collected all known phase-change alloys that show good properties in rewritable optical data storage (successful samples') and also two samples that are not suitable. Consistent with the arguments presented above we note from Table 1 that the N sp of successful PC materials is always larger than 4, and usually between 4.3 and 5. From Figure 3, we estimate that the critical value of N sp that distinguishes the rocksalt and the chalcopyrite structures is about 4.1.In summary, the DFT calculations presented here help to identify new and possibly superior phase-change materials. All phase-change materials reported are characterized by a cubic or near-cubic coordination, which is caused by the dominance of the p-electron bonding. The average number of valence electrons plays the crucial role in determining the structure of the semiconductor alloys studied here. For N sp = 4, we find that the stable structure of (Cu,Ag,Au)InTe 2 is the chalcopyrite phase with sp 3 -bonding, while for N sp = 4.5, the stable structure of (Cu,Ag,Au)SbTe 2 ternary alloys is the rocksalt structure with p-bonding. Results show that Te-based ternary materials will favor the suitable rocksalt structure if their N sp is bigger than 4.1. This simple criterion facilitates the search for new phase-change materials and opens the way to a more fundamental understanding of phase-change materials.