Micro signal: thepapernews
Schrodinger equation is famous for its difficulty, which is a big nightmare for quantum physicists. Ai said: it's not a big problem.
Scientists from the Department of Physics and the Department of Computer Science of the Free University of Berlin, Germany, recently jointly developed an artificial intelligence (AI) method to solve the ground state of Schrodinger equation in the field of quantum chemistry, which broke through the difficulties of traditional methods in accuracy and computational efficiency.
Related papers were published in Natural Chemistry on February 2 1, 65438. Frank Noé, who led this work, believes that this method may have a great impact on the future of quantum chemistry.
Quantum chemistry is the study of chemistry from a quantum perspective. Its main goal is to skip the time-consuming, laborious and expensive experiments and predict the chemical and physical properties of molecules and the results of chemical reactions only through the spatial distribution of atoms that make up molecules. This can be achieved by solving Schrodinger equation in theory, but it is extremely difficult in practice. So far, scientists have not found an efficient method to solve any molecule accurately.
Quantum chemistry is a chemistry without test tubes and flasks.
On the other hand, what kind of equation is Schrodinger equation? Why is it so hard to understand?
We know that in classical mechanics, the motion of an object can be described by Newton's second law (F=ma). But in the microscopic quantum world, it is different. The behavior of microscopic particles follows a set of much more complicated laws and must be described by Schrodinger equation.
1926, the famous Austrian physicist Schrodinger put forward a great equation, which laid the foundation stone for quantum mechanics. In the quantum world, the state of micro-system is not determined by the value of some specific mechanical quantities, but by the function of mechanical quantities, that is, wave function. How the probability distribution of mechanical quantities changes with time can be solved by solving Schrodinger equation of wave function.
This is why the core work of quantum chemistry is to solve the Schrodinger equation. The chemical properties of molecules and the results of chemical reactions are basically determined by the behavior of electrons around the nucleus.
Knock on the blackboard! Schrodinger has not only cats, but also equations.
Although the teacher has written the formula on the blackboard, it is still difficult to apply it to solve the problem. So far, human beings have only used Schrodinger equation to "understand" hydrogen atom, which is the simplest system composed of only one proton and one extranuclear electron. For other atoms, this understanding is greatly reduced.
For more complex systems like molecules, it is even more difficult for wave functions to capture how electrons interact subtly.
In fact, most quantum chemistry methods give up directly solving the wave function and only pursue the determination of the given molecular energy. This either requires approximate calculation and sacrifices the quality of prediction, or requires complex mathematical methods, which are difficult to be applied.
Austrian physicist Schrodinger
The greatest value of this research achievement of Free University of Berlin is that it breaks through the dilemma of accuracy and computational efficiency and provides unprecedented accuracy at acceptable computational cost.
Jan Hermann, another author of the paper, mentioned that the most popular calculation method in this field is density functional theory with high calculation efficiency, and their profound "quantum Monte Carlo" will not be inferior.
Different from the traditional method of splitting the wave function into relatively simple mathematical modules, the research team designed an artificial deep neural network, which can learn the complex distribution pattern of electrons around the nucleus. The so-called deep learning refers to a method of extracting structured information from massive data sets by using hierarchical machine learning algorithm.
Of course, AI can defeat Ke Jie by learning tens of millions of chess manuals by itself, and self-learning quantum mechanics is somewhat "overreaching". Physicists also integrate the basic physical characteristics of electronic wave functions such as Pauli exclusion principle into artificial neural networks, injecting quantum mechanics "soul" into AI, which has become the finishing touch.
Author: Yu