A drop of water in your palm holds roughly $10^{21}$ molecules, each made of three atoms surrounded by a swarm of electrons. To understand how matter behaves, physicists solve equations for these electrons. The trouble is that the number of equations grows like an avalanche: for N electrons, you need to store $N^2$ numbers. For a drop of water, that's an astronomical figure, and no supercomputer can handle it.
The authors of the new work found a loophole. Not all electrons in a molecule "talk" to each other equally. Those far apart barely influence each other, like people at opposite ends of a huge hall. So they can be grouped together and described by a single, "averaged" wave. The method is called chromatic superposition (from the Greek for "color") because it resembles graph coloring: atoms with no strong connection get the same color and merge into one mathematical object. This radically compresses the data — sometimes by hundreds of times.
But how do you quickly translate calculations into this compressed language? That's where the Krylov method comes in — a trusty old tool for working with large matrices. It lets you build the needed quantities step by step without storing the entire matrix. Thanks to this, even for a system with a million water molecules, the computation scales linearly: more molecules mean proportionally more time, not explosive growth. On small problems, the new method already beats previous linear algorithms by a factor of ten, and on large ones it opens doors that were previously unimaginable.
The practical benefits are enormous: from drug development (where you need to know exactly how a protein molecule interacts with water) to creating materials for clean energy. When a computer can honestly, without crude simplifications, track the motion of tens of thousands of atoms, we start to see details that used to slip away. And all this thanks to an idea: sometimes you can assemble the whole from large colored patches rather than from each grain of sand individually.
If we tried to write down the wave function of just one gram of water as floating-point numbers, we'd need more bits than there are atoms in the observable universe. The color superposition method sidesteps this nightmare, compressing the description to a manageable size.