经典力学教材称精确可解模型均可分离，是否存在可积但不可分离模型？

Great question—this is a common sticking point because introductory classical mechanics textbooks often prioritize separable systems when teaching exact solvability, but the landscape is broader than that. Let’s break this down clearly:

1. 是否存在可积但不可分离的模型？ #

Absolutely yes. The key distinction here is that separability is a sufficient condition for Liouville integrability, but not a necessary one.

Liouville integrability only requires that for an n-degree-of-freedom system, there exist n independent, mutually commuting constants of motion (in involution). Separability, on the other hand, demands that the Hamiltonian can be split into a sum of terms each depending on only one pair of conjugate variables (or a subset), allowing you to solve the system by integrating each variable separately.

Classic examples of integrable but non-separable systems include:

• The Kovalevskaya Top: A rigid body rotating about a fixed point, with specific moments of inertia ($I_1=I_2=2I_3$) and no external torque. Its Hamiltonian cannot be separated in any of the standard coordinate systems (Euler angles, etc.), but it satisfies Liouville’s integrability condition with three independent, commuting integrals of motion. This is one of only three exactly solvable rigid body problems (alongside Euler and Lagrange tops), and it’s explicitly non-separable.
• Henon-Heiles System (at certain energy levels): When the total energy is below a critical threshold, this 2-degree-of-freedom system is Liouville integrable, but its Hamiltonian cannot be separated into functions of individual $(q,p)$ pairs. It’s a popular example in chaos theory because it becomes non-integrable above that energy.

2. 经典力学教材中“所有精确可解模型均为可分离模型”的说法准确吗？ #

This statement is incomplete, not universally true. Introductory textbooks focus on separable systems because they’re easier to teach and demonstrate basic techniques (like separation of variables in Hamilton-Jacobi equations). But exact solvability (which aligns with Liouville integrability for Hamiltonian systems) doesn’t require separability.

The Kovalevskaya Top I mentioned earlier is a perfect counterexample: it’s exactly solvable (you can derive closed-form solutions for its motion), but it’s non-separable. Other examples include certain versions of the Calogero-Moser system, a many-body system with inverse-square interactions that’s integrable but can’t be solved via variable separation.

The confusion arises because separability is a straightforward way to construct integrable systems, and many common integrable systems are separable. But the set of integrable systems is larger than just the separable ones.

内容的提问来源于stack exchange，提问作者Jiang-min Zhang