A new method, well-tempered metadynamics, is introduced to calculate the free energy surface (FES) of a system using a set of collective variables (CVs). This method improves upon traditional metadynamics by allowing for controlled convergence and accurate estimation of the FES. The method introduces a bias potential that depends on the history of the system, enabling the simulation to explore physically relevant regions of the CV space more efficiently. The bias potential is designed to decrease over time, ensuring that the system converges to the correct FES while maintaining computational efficiency. The method is based on the concept of adaptive biasing, where the bias potential is adjusted dynamically to enhance sampling. The bias potential is constructed as a function of the history of the system, and its rate of change is controlled to ensure convergence. The method allows for the tuning of the biasing parameters to focus computational resources on the most relevant regions of the CV space. The method is tested on the alanine dipeptide system, where the FES is calculated as a function of the backbone dihedral angles. The results show that the method converges to the correct FES and provides accurate estimates of the free energy differences between metastable states. The method is also shown to reduce the error in the FES estimation, with the error decreasing as the simulation time increases. The method is compared to standard metadynamics and is shown to have better convergence properties and lower error. The method is also shown to be robust and effective in a wide range of scenarios, including cases where the transverse degrees of freedom are either fast or slow. The method is applicable to a variety of systems and can be extended to higher-dimensional cases, making it a versatile tool for studying complex systems. The method is also shown to be computationally efficient, allowing for the use of adaptive-bias methods in higher-dimensional cases.A new method, well-tempered metadynamics, is introduced to calculate the free energy surface (FES) of a system using a set of collective variables (CVs). This method improves upon traditional metadynamics by allowing for controlled convergence and accurate estimation of the FES. The method introduces a bias potential that depends on the history of the system, enabling the simulation to explore physically relevant regions of the CV space more efficiently. The bias potential is designed to decrease over time, ensuring that the system converges to the correct FES while maintaining computational efficiency. The method is based on the concept of adaptive biasing, where the bias potential is adjusted dynamically to enhance sampling. The bias potential is constructed as a function of the history of the system, and its rate of change is controlled to ensure convergence. The method allows for the tuning of the biasing parameters to focus computational resources on the most relevant regions of the CV space. The method is tested on the alanine dipeptide system, where the FES is calculated as a function of the backbone dihedral angles. The results show that the method converges to the correct FES and provides accurate estimates of the free energy differences between metastable states. The method is also shown to reduce the error in the FES estimation, with the error decreasing as the simulation time increases. The method is compared to standard metadynamics and is shown to have better convergence properties and lower error. The method is also shown to be robust and effective in a wide range of scenarios, including cases where the transverse degrees of freedom are either fast or slow. The method is applicable to a variety of systems and can be extended to higher-dimensional cases, making it a versatile tool for studying complex systems. The method is also shown to be computationally efficient, allowing for the use of adaptive-bias methods in higher-dimensional cases.