While studying Poisson brackets in classical mechanics and the derivation of $\dot{q_j}=\{q_j,H\}$ and $\dot{p_j}=\{p_j,H\}$ form of Hamilton's equations I encountered a surpsing identity, which led me to think that maybe I got something wrong about full time-derivative, which is as follows:

$$ \frac{d\,f(q_1,q_2,\dots,q_N,p_1,p_2,\dots,p_N,t)}{dt}=\sum_{j=1}^{N}\left( \frac{\partial f}{\partial q_j} \underbrace{\frac{\partial q_j}{\partial t}}_{=\dot{q_j}=\frac{\partial H}{\partial p_j}}+ \frac{\partial f}{\partial p_j} \underbrace{\frac{\partial p_j}{\partial t}}_{=\dot{p_j}=-\frac{\partial H}{\partial q_j}} \right)+\frac{\partial f}{\partial t} =\{f,H\}+\frac{\partial f}{\partial t} $$

Now if for example I use function $f$ as $q_k$, by saying: $f(q_1,q_2,\dots,q_N,p_1,p_2,\dots,p_N,t)=q_k$ then I get following:

$$ \frac{dq_k}{dt}=\sum_{j=1}^{N}\left( \underbrace{\frac{\partial q_k}{\partial q_j}}_{=\delta_{kj}} \frac{\partial H}{\partial p_j}- \underbrace{\frac{\partial q_k}{\partial p_j}}_{=0} \frac{\partial H}{\partial q_j} \right)+\frac{\partial q_k}{\partial t} =\{q_k,H\}+\frac{\partial q_k}{\partial t} = \underbrace{\frac{\partial H}{\partial p_k}}_{=\dot{q_k}} + \underbrace{\frac{\partial q_k}{\partial t}}_{\stackrel{?}{=}\dot{q_k}}\stackrel{?}{=}2\dot{q_k} $$

If I cross out the $\frac{\partial q_k}{\partial t}$ on both sides of equation $\{q_k,H\}+\frac{\partial q_k}{\partial t}=\frac{\partial H}{\partial p_k}+\frac{\partial q_k}{\partial t}$ then I recover the Hamilton's equation $\dot{q_k}=\{q_k,H\}$. But if I don't do this, and go forward with the $2\dot{q_k}$ that appears at the end I get this very surprising identity:

$$ \frac{dq_k}{dt}\stackrel{?}{=}2\dot{q_k} $$

or written another way:

$$ \frac{dq_k}{dt}\stackrel{?}{=}2\frac{\partial q_k}{\partial t} $$

My question is following: is this really true? If not, then what have I done wrong? If yes, then why it is not mentioned anywhere in the textbooks - wouldn't that be some other way to find the $\dot{q_k}$?

Note:The derivation goes exactly the same for $\dot{p_k}$.

EDIT: Thanks to your answers I marked now the wrong equalities with $\stackrel{?}{=}$ since I hate to see incorrect math typed out. But I still wanted to preserve this question as how it was written at first. Also in the first equation the full time derivative for $\dot{q_j}$ and $\dot{p_j}$ should be used like this:

$$ \frac{df}{dt}=\sum_{j=1}^{N}\left( \frac{\partial f}{\partial q_j} \underbrace{\frac{d q_j}{d t}}_{=\dot{q_j}=\frac{\partial H}{\partial p_j}}+ \frac{\partial f}{\partial p_j} \underbrace{\frac{d p_j}{d t}}_{=\dot{p_j}=-\frac{\partial H}{\partial q_j}} \right)+\frac{\partial f}{\partial t} $$


1 Answer 1


The problem is that you are equating too many things to $\dot{q_k}$.

Usually $\dot{q_k} = \frac{dq_k}{dt}$, a total derivative, as opposed to a partial derivative.

If $q_k$ has no explicit time-dependence, i.e. it does not depend directly on $t$ itself, then $\frac{\partial q_k}{\partial t} = 0.$

In this case, the Poisson bracket reduces to: $ \frac{dq_k}{dt} = \frac{\partial H}{\partial p_k} $, so now you can say $\dot{q_k} = \frac{\partial H}{\partial p_k} $.

If you now consider your $q_k$ to have a time dependence, so $q_k(t)$, the Poisson bracket becomes, as you have pointed out: $ \frac{dq_k}{dt} = \frac{\partial H}{\partial p_k}+\frac{\partial q_k}{\partial t}$, so $ \dot{q_k} =\frac{\partial H}{\partial p_k} +\frac{\partial q_k}{\partial t} $.

ADDITION after discussion in comments:

$\frac{d}{dt}f$ is a total derivative with respect to time, which means it picks out ALL time dependences of $f(t, x(t), y(t), z(t))$. Collectively, call $x,y,z$ as $\{x_i\}$.

Using the chain rule, $$ \frac{d}{dt} = \frac{\partial}{\partial x_i}\frac{\partial x_i}{\partial t} = \frac{\partial}{\partial x}\frac{\partial x}{\partial t}+\frac{\partial}{\partial y}\frac{\partial y}{\partial t}+\frac{\partial}{\partial z}\frac{\partial z}{\partial t}+\frac{\partial}{\partial t}\underbrace{\frac{\partial t}{\partial t}}_1 = \mathbf{u}\cdot\mathbf{\nabla}+\frac{\partial}{\partial t}.$$

In Hamiltonian mechanics, you parameterise a function in terms of its positions $\{q_i\}$ and its momenta $\{p_i\}$. Basically the $\{x_i\}$ from earlier are now $\{q_i, p_i\}$. $$ \frac{df}{dt}=\frac{\partial f}{\partial x_i}\frac{\partial x_i}{\partial t} = \frac{\partial f}{\partial q_i}\frac{\partial q_i}{\partial t}+\frac{\partial f}{\partial p_i}\frac{\partial p_i}{\partial t} + \frac{\partial f}{\partial t}.$$

The definition of the Poisson bracket is:

$$ \{f,H\}=\left(\frac{\partial f}{\partial q_i}\frac{\partial H}{\partial p_i}-\frac{\partial f}{\partial p_i}\frac{\partial H}{\partial q_i} \right). $$

Plugging in the Hamilton's equations:

$$ \{f,H\}=\left(\frac{\partial f}{\partial q_i}\frac{dq_i}{dt}+\frac{\partial f}{\partial p_i}\frac{dp_i}{dt} \right), $$ when $q_i$ and $p_i$ are the fundamental variables, so they are not functions of $x, y, z$ but only time. This means that, applying the formula from before, $\frac{\partial q_i}{\partial t} = \frac{dq_i}{dt}$, so: $$ \{f,H\}=\left(\frac{\partial f}{\partial q_i}\frac{\partial q_i}{\partial t}+\frac{\partial f}{\partial p_i}\frac{\partial p_i}{\partial t} \right). $$

As you can see $\frac{\partial f}{\partial t}$ does not enter the definition.

In conclusion,

$$ \frac{df}{dt}=\{f,H\}+\frac{\partial f}{\partial t}. $$

P.S. I have been using Einstein's summation convention: repeated indeces imply summation. $$\frac{\partial f}{\partial p_i}\frac{\partial p_i}{\partial t} = \sum_{i=1}^{N} \frac{\partial f}{\partial p_i}\frac{\partial p_i}{\partial t}.$$

  • $\begingroup$ Thank you very much, but in that case the textbook definition of Poisson bracket becomes problematic since it uses $\frac{\partial q_j}{\partial t}$ and it becomes zero. I mean this $\{f,H\}=\sum_{j=1}^{N}\left(\frac{\partial f}{\partial q_j}\underbrace{\frac{\partial q_j}{\partial t}}_{=0}+\frac{\partial f}{\partial p_j}\underbrace{\frac{\partial p_j}{\partial t}}_{=0} \right)$. Or we are having three different kinds of derivatives here, and I just didn't know about that?Oh wait $\endgroup$ Commented Nov 11, 2015 at 20:12
  • $\begingroup$ Oh wait, maybe it will be correct if I write it this way: $\{f,H\}=\sum_{j=1}^{N}\left(\frac{\partial f}{\partial q_j}\frac{dq_j}{dt}+\frac{\partial f}{\partial p_j}\frac{dp_j}{dt} \right)$ ? $\endgroup$ Commented Nov 11, 2015 at 20:18
  • $\begingroup$ Let me add something to my question, let's see if that answers more of your questions. $\endgroup$ Commented Nov 11, 2015 at 22:05
  • $\begingroup$ Am I correct that critical to understanding this problem is time dependence: when plugging Hamilton's equations $q_i(t)$ depends on only one variable inside the Poisson bracket: $\frac{dq(t)_i}{dt}=\frac{\partial q(t)_i}{\partial t}$, while in other situation $\frac{\partial q(NOT\,t)_i}{\partial t}=\frac{\partial q()_i}{\partial t}=0$ when using the function $f(q_1,q_2,\dots,q_N,p_1,p_2,\dots,p_N,t)=q_i()$ in formula $\frac{df}{dt}=\{f,H\}+\frac{\partial q()_i}{\partial t}$ ? $\endgroup$ Commented Nov 12, 2015 at 14:33
  • $\begingroup$ if q has no explicit time dependence, i.e. q(x) and NOT q(t) , even though x(t), then partial_t q = total_t q, yes. $\endgroup$ Commented Nov 13, 2015 at 13:42

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.