Let $H_1,H_2\in\mathcal{C}^1(\mathbb{R}^3;\mathbb{R})$ be two scalar fields. Consider a trajectory $\vec{x}(t)\in\mathbb{R}^3$ such that, for all observable $f\in\mathcal{C}^1(\mathbb{R}^3;\mathbb{R})$,

$$\frac{\mathrm{d}}{\mathrm{d}t}f(x)=\det\big(\nabla f,\nabla H_1, \nabla H_2\big)=\frac{\partial(f,H_1,H_2)}{\partial\vec{x}}.$$

This dynamical system recalls a Hamiltonian system with hamiltonian $H$ on the phase space $\lbrace(x,p)\in\mathbb{R}^2\rbrace$ such that for all observable $f\in\mathcal{C}^1(\mathbb{R}^2;\mathbb{R})$:

$$\frac{\mathrm{d}}{\mathrm{d}t}f(x)=\det\big(\nabla f,\nabla H\big)=\frac{\partial(f,H)}{\partial(x,p)}=\frac{\partial f}{\partial x}\frac{\partial H}{\partial p}-\frac{\partial f}{\partial p}\frac{\partial H}{\partial x}=\big\lbrace f,H\big\rbrace,$$

the Poisson bracket. Hence I would like to say that my dynamical system is a kind of "multi-hamiltonian" system. Is there any reference in which this kind of generalisation is studied?

Edit: it can be generalised to a system with $d-1$ scalar fields $(H_i)$ on $\mathbb{R}^d$ satisfying:

$$\frac{\mathrm{d}}{\mathrm{d}t}f(x)=\det\big(\nabla f,\nabla H_1,... \nabla H_{d-1}\big)=\frac{\partial(f,H_1,...,H_{d-1})}{\partial\vec{x}}.$$

  • $\begingroup$ How does your proposal behave under changes of coordinates? I ask to see if we can figure out if there is any further geometric meaning to it. $\endgroup$ Nov 16, 2020 at 1:12
  • $\begingroup$ A set of new coordinates $\vec{y}\in\mathbb{R}^3$ satisfying $\partial\vec{y}/\partial \vec{x}=1$ preserves the equation of evolution (due to the fact that $\partial(f,H_1,H_2)/\partial\vec{x}=\partial(f,H_1,H_2)/\partial{\vec{y}}\times\partial\vec{y}/\partial\vec{x}$). This is analogous to a conanical change of coordinates for a usual Hamiltonian system. $\endgroup$
    – G. Panel
    Nov 16, 2020 at 1:32
  • $\begingroup$ In particular, the orders of the $x$ dimensions seems tied to the orders of the $H$'s, so it might be better to think of it as something like $(x, p_{1}, p_{2})$ rather than $\vec x$ $\endgroup$ Nov 16, 2020 at 2:07
  • $\begingroup$ My first question would be, " can you reduce this down to a normal Hamiltonian system by eliminating the third variable"? $\endgroup$ Nov 16, 2020 at 2:09
  • $\begingroup$ If we assume that $\partial(x_1,x_2,H_2)/\partial\vec{x}=1$, then $\partial(f,H_1,H_2)/\partial\vec{x}=\partial(f,H_1,H_2)/\partial(x_1,x_2,H_2)=\partial(f,H_1)/\partial(x_1,x_2)$ (after development toward the third column), and the system is then hamilonian in this specific case. $\endgroup$
    – G. Panel
    Nov 16, 2020 at 3:02

1 Answer 1


Answered in PhysicsOverflow; reposting here for convenience:

Intriguingly enough, a quick search for "multi Hamiltonian physics" does not give any meaningful result; the mechanics described by the OP is Nambu mechanics, and is linked to nonassociative algebras appearing in e.g. M-theory.

In contrast to the construction by the OP, Nambu generalizes the Poisson bracket (rather than using a higher-dimensional matrix determinant) and writes the equations of motion $$\frac{df}{dt}=\{f,H_1,H_2,…,H_n\}$$

See https://arxiv.org/abs/hep-th/0212267 | https://arxiv.org/abs/1903.05673


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.