Lets assume we have a very simple transformation in 1 Dimension from $(x, p_x)\rightarrow (y,p_y)$ given as $$\begin{aligned} y &= cx \\ p_y &= c^{-1} p_x \end{aligned}$$

Is this a strictly canonical transformation, an extended canonical transformation or a scaling transformation?

Let the Hamiltonian in $x,p_x$ be $$ H(x, p_x) = \frac{p_x^2}{2m} + V(x) $$

A quick test with Poisson brackets $$\{f,H \}_{(x,p_x)}\equiv \frac{\partial f}{\partial x}\frac{\partial H}{\partial p_x} - \frac{\partial f}{\partial p_x}\frac{\partial H}{\partial x} $$

yields $$\begin{aligned} \dot y &= \{y,H \}_{(x,p_x)} = \frac{c}{m}p_x = c \dot x \\ \dot p_y &= \{p_y,H \}_{(x,p_x)} = -c^{-1}V'(x) = c^{-1}\dot p_x \\ \end{aligned}$$

which looks like it satisfies the strict canonical transformation conditions.

Is this a special case of a scaling transformation that is at the same time a canonical transformation? And if it is a canonical transformation, what is a generating function for this transformation? I fail at matching it with the four common types and have a hard time applying the formalism for canonical transformations to it.


1 Answer 1


First of all, let my clarify my notation. In this answer I'll use

  • $(p,q)$ as coordinates before the transformation;
  • $(P,Q)$ as coordinates after the transformation.

In this notation, your transformation reads $$Q=cq \qquad P=c^{-1}p$$ this is indeed a scale transformation and it preserves the canonical structure of Hamilton's equations$^1$. In other words, there exists a Hamiltonian in the new set of coordinates that satisfies Hamilton's equations. According to some authors this is the definition of canonical transformation.

Others, the book I quoted below for example, define as canonical a transformation whose jacobian matrix is symplectic.$^2$ You can verify that the jacobian matrix is

$$\begin{pmatrix} c^{-1} & 0 \\ 0 & c \end{pmatrix}$$ is symplectic i.e.

$$\begin{pmatrix} c^{-1} & 0 \\ 0 &c\end{pmatrix}^T\begin{pmatrix} 0 & -1 \\ 1 & 0 \end{pmatrix}\begin{pmatrix} c^{-1} & 0 \\ 0 & c \end{pmatrix}=\begin{pmatrix} 0 & -1\\ 1 & 0 \end{pmatrix} $$

This is not just luck. More generally, a scale transformation is defined as follows$^3$ $$Q_i=c_i q \qquad P_i=d_i p_i \qquad\forall i$$ with $c_i,d_i\in\mathbb{R}$ such that $c_id_i=\lambda\in\mathbb{R}$.

If $\lambda=1$, it turns out it is canonical also in this sense$^4$ (symplectic jacobian matrix). Of course, in our case we have just one coordinate and $d=c^{-1}$, so our condition is satisfied.

Finally, regarding the generating function, consider a generating function of the second type $F_2(q,P)=cqP$ that is the generating function of identical transformation multiplied by $c$. Using the definition of generating functions $$Q=\frac{\partial F_2}{\partial P}=cq \iff Q=cq\\ p=\frac{\partial F_2}{\partial q}=cP\iff P=c^{-1}q$$ that is your transformation.

$^1$ See Analytical Mechanics. Fasano&Marmi. Chapter 10, section 10.2. Example 2.3.

$^2$ See Analytical Mechanics. Fasano&Marmi. Chapter 10, Definitions 1.2 and 2.2.

$^3$ No summation on repeated indexes is understood.

$^4$ See Analytical Mechanics. Fasano&Marmi. Chapter 10, section 10.2. Example 2.7

  • $\begingroup$ The reference and answer is very helpful. There is one point that is still murky for me. Which matrix is the symplectic one? The jacobian $$\begin{pmatrix} c^{-1} & 0 \\ 0 & c \end{pmatrix}$$ is clearly not skew-symmetric and thus not symplectic. I think I am misunderstanding something obvious here. $\endgroup$
    – Hans Wurst
    Commented Jul 18, 2022 at 18:36
  • 1
    $\begingroup$ I expanded that part. Symplectic does not mean skew symmetric. Identity matrix is symplectic, for example. As for the definition of symplectic, you can check footnote 2, definition 1.2. A $2n\times 2n$matrix $A$ is symplectic iff $A^T\mathcal{J}A=\mathcal{J}$ where $\mathcal{J}$ is $\begin{pmatrix} 0& -I_n \\ I_n&0\end{pmatrix} $ $\endgroup$ Commented Jul 18, 2022 at 19:05
  • $\begingroup$ Thank you for the clarification. $\endgroup$
    – Hans Wurst
    Commented Jul 18, 2022 at 19:55

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.