We have calculated the melting temperature of tungsten by two ab initio approaches. The first approach can be divided into two steps. In the first step, we simulate a large coexisting solid and liquid system by the classical embedded-atom method potential and obtain an approximate melting temperature. In the second step, we compute the accurate melting temperature by performing the ab initio free-energy corrections. The second approach is to perform a direct ab initio molecular-dynamics simulation for the coexisting solid and liquid system using the constant particle number, pressure, and enthalpy ensemble. In the second approach, the simulation is carried out entirely using a density-functional theory Hamiltonian, and no other approximations are imposed. However, the simulation is performed using a relatively small supercell. The results obtained from two ab initio approaches can provide a check for each other. Our results show that they are in good agreement with each other and also in reasonably good agreement with the experimental value. Computer simulation of solid-liquid equilibrium can be traced back to the 1950s.1 Thanks to the increasing power of modern supercomputers, it has recently become possible to calculate the melting properties of materials using accurate ab initio methods. There are two commonly used computational approaches introduced in previous studies to obtain the melting temperature of a material. The first one is the so-called thermodynamic integration approach. [2][3][4][5][6][7] In this approach, the Gibbs free energies are calculated ab initio for solid and liquid, and the melting transition is determined by the equality of the Gibbs free energies of two phases. The second approach is the direct simulation of the solid-liquid coexistence, i.e., the so-called coexistence approach. [8][9][10][11][12][13][14][15][16][17][18] In this approach, the temperature adjusts spontaneously during the simulation to provide a two-phase equilibrium that satisfies the equality of Gibbs free energies of solid and liquid. In this paper, we use a recently proposed [19][20][21][22] hybrid approach combining the above two approaches. An approximate melting temperature is first obtained by coexisting solid and liquid simulation using an empirical potential. Next, the ab initio melting temperature is obtained by applying free-energy corrections akin to thermodynamic integration in the limit of small perturbations. This approach has been applied to get the melting properties of Fe, 19 Cu, 20 Ta, 21 and Mo 22 for a wide pressure range. The advantage of this last approach is its moderate computational costs compared with the other two approaches.As the melting point is a quantity that is very sensitive to small inaccuracies in the Hamiltonian used, the use of accurate ab initio methods is desired. However, such methods are computationally demanding, and the accuracy is limited by sampling and system size convergence. The ability to internally check the accuracy of the results is therefore crucial. In the presen...