We investigate the thermodynamic properties of D-brane−anti-D-brane pairs and non-BPS D-branes on the basis of boundary string field theory. We calculate the finite temperature effective potential of N D-brane−anti-D-brane pairs in a non-compact background and in a toroidal background. In the non-compact background case, a phase transition occurs slightly below the Hagedorn temperature, and the D9-D9 pairs become stable. Moreover, the total energy at the critical temperature is a decreasing function of N as long as the 't Hooft coupling is very small. This leads to the conclusion that a large number N of D9-D9 pairs are created simultaneously near the Hagedorn temperature. In the toroidal background case (M 1,9−D × T D ), a phase transition occurs only if the Dp-Dp pair is extended in all the noncompact directions, as long as the 't Hooft coupling is very small. The total energy at the critical temperature also decreases as N increases. We also calculate the finite temperature effective potential of non-BPS D-branes, and we obtain similar results. Then, we consider the thermodynamic balance between open strings on these branes and closed strings in the bulk in the ideal gas approximation, and conclude that the total energy is dominated by the open strings. §1. IntroductionUnderstanding the properties of unstable D-brane systems, such as coincident D-brane−anti-D-brane pairs and non-BPS D-branes, 1) has been a subject of much interest (for a review see, e.g., Ref. 2)). Type IIB string theory contains D-brane−anti-D-brane pairs of odd dimension and non-BPS D-branes of even dimension, whereas type IIA string theory contains D-brane−anti-D-brane pairs of even dimension and non-BPS D-branes of odd dimension. The spectrum of open strings on these unstable branes contains a tachyon field T . In such a brane configuration, we have T = 0, and the potential of this tachyon field has a local maximum at T = 0. If we assume that the tachyon potential has a non-trivial minimum, it is hypothesized that the tachyon falls into it. Sen conjectured that the tensions of these branes and the negative potential energy of the tachyon exactly cancel at the potential minimum. 3) This implies that these unstable brane systems disappear at the end of the tachyon condensation.If tachyon condensation occurs in a topologically non-trivial way, there remain some topological defects, such as kinks and vortices. We can identify these topological defects as lower-dimensional D-branes and the topological charge as the RamondRamond charge of the resulting D-branes. 1),3) These D-brane charges can be classified using K-theory. 4),5) In particular, we can realize all the D-branes through tachyon condensation from the spacetime-filling branes, such as D9-D9 pairs and non-BPS * )