Simulation of the dynamics of dust-gas circumstellar discs is crucial in understanding the mechanisms of planet formation. The dynamics of small grains in the disc is stiffly coupled to the gas, while the dynamics of grown solids is decoupled. Moreover, in some parts of the disc the concentration of the dust is low (dust to gas mass ratio is about 0.01), while in other parts it can be much higher. These factors place high requirements on the numerical methods for disc simulations. In particular, when gas and dust are simulated with two different fluids, explicit methods require very small timestep (must be less than dust stopping time $t_{\rm stop}$ during which the velocity of a solid particle is equalized with respect to the gas velocity) to obtain solution, while some implicit methods requires high temporal resolution to obtain acceptable accuracy. Moreover, recent studies underlined that for Smoothed particle hydrodynamics (SPH) when the gas and the dust are simulated with different sets of particles only high spatial resolution $h<c_{\rm s} t_{\rm stop}$ guaranties suppression of numerical overdissipation due to gas and dust interaction. To address these problems, we developed a fast algorithm based on the ideas of (1) implicit integration of linear (Epstein) drag and (2) exact conservation of local linear momentum. We derived formulas for monodisperse dust-gas in two-fluid SPH and tested the new method on problems with known analytical solutions. We found that our method is a promising alternative for the previously developed two-fluid SPH scheme in case of stiff linear drag thanks to the fact that spatial resolution condition $h<c_{\rm s} t_{\rm stop}$ is not required anymore for accurate results.

Circumstellar discs, from which planetary systems are formed, consist of gas, dust and solids. Simulations of self-consistent dynamics of gas, dust and solids in circumstellar discs is a challenging problem. In the paper we present fast algorithms for computing the drag force (momentum transfer) between solid phase and gas. These algorithms (a) are universal and applicable to dust and solids with any sizes smaller than the mean free path of gas molecules, (b) can be used to calculate the momentum transfer between dust and gas instead of one-way effect, as it is done in many models, (c) can perform simulations, without a loss in accuracy, with the time step determined by gas-dynamic parameters rather than by drag force, and (d) are compatible with the widely used parallel algorithms for solving 3D equations of gas dynamics, hydrodynamic equations for dust, and the collisionless Boltzmann equation for large bodies. Preliminary results of supercomputer simulation of the gas-dust disc dynamics within the developed approach are reported.