0
$\begingroup$

I am trying to find the action associated with the Lagrangian density $$ \mathcal{L} = \frac{1}{2}\left( \frac{\partial\phi}{\partial x} \right)^2 + \frac{1}{2}m^2\phi^2. \tag{1} $$ I am supposed to use the discrete expansions $$\phi_j = \frac{1}{\sqrt{Na}}\sum_p \tilde{\phi}_pe^{ipja} = \frac{1}{\sqrt{Na}}\sum_{-p} \tilde{\phi}_{-p}e^{-ipja}. \tag{2} $$

So, first I find the Lagrangian, using $$ L = \int dx \mathcal{L} = a \sum_j \mathcal{L} = \frac{a}{2}\sum_j \left[ \left( \frac{\phi_{j+1}-\phi_j}{a} \right)^2 + m^2\phi_j^2 \right] \tag{3} $$ where $j$ labels the 1D lattice sites and $a$ is the equilibrium distance between each site.
Now I plug in the expansion for $\phi_j$ into the Lagrangian, and where $\phi_j$ is squared, I use one copy of the middle ($+p$) term in eq (2) and one copy of the right ($-p$) term in eq (2), multiplied together. This is motivated by the form of the action I am supposed to get in the end.
When I do the substitution into $L$, I end up with $$ L = \frac{1}{2} \sum_p \tilde{\phi_p}\tilde{\phi_{-p}}\left[ \frac{2}{a^2}\left( 1-\cos{pa} \right) +m^2 \right] \tag{4}. $$

Now to get the action, I know that $$ S = \int L dt, \tag{5}$$ but I have no idea where time is supposed to come into this problem at all. When integrating the Lagrangian density to get the Lagrangian, I know that I had to realize that the integral over one spatial dimension becomes, in the discrete case, a sum over the positions $x_j$ times the lattice constant $a$, or just a sum over $j$, again times $a$. In addition, the spatial derivative in the Lagrangian becomes a discrete difference, as I have shown above.

Furthermore, the expression that I obtained for the Lagrangian $L$ is exactly what my textbook says I should obtain for the action $S$! Is this somehow the result of the problem not having any obvious time-dependence? So, in total, I suppose I want to know how the action relates to the Lagrangian in the case of a problem that doesn't involve time.

Just for clarity, I am going to write the question as phrased in the textbook (QFT for the Gifted Amateur):

Exercise 17.5 (a): Consider a one-dimensional system with Lagrangian $$ \mathcal{L} = \frac{1}{2}\left( \frac{\partial \phi(x)}{\partial x} \right)^2 + \frac{m^2}{2} \left[ \phi(x) \right]^2. $$ The choice of sign makes this a Euclidean theory. Descretize this theory (that is, put it on a lattice) by defining $$ \phi_j = \frac{1}{\sqrt{Na}} \sum_p \tilde{\phi}_p e^{ipja}, $$ where $j$ labels the lattice site, $a$ is the lattice spacing, and $N$ is the number of lattice points. Using the method in exercise 17.3 show that the action may be written $$ S = \frac{1}{2} \sum_p \tilde{\phi}_{-p} \left( \frac{2}{a^2}-\frac{2}{a^2}\cos{pa} + m^2 \right) \tilde{\phi}_p, $$ and read off the propagator for this theory.

The "method in exercise 17.3" is just what I described in between eq (3) and eq (4), where you expand $\phi_j$ in terms of its Fourier transforms $\tilde{\phi}_p$ and $\tilde{\phi}_{-p}$. Problem 17.3 also is the one that shows that the free propagator is $\frac{i}{2}$ times the inverse of the quadratic term in the momentum-space action, which is why this problem is asking us to find the action in the first place.

$\endgroup$
  • $\begingroup$ @Qmechanic: His question says nothing about the continuum limit. $\endgroup$ – Daniel Jul 26 '15 at 21:37
  • $\begingroup$ Wouldn't taking the limit as $a$ goes to zero just be undoing the discretization that I was told to implement? $\endgroup$ – Physics_Plasma Jul 26 '15 at 21:39
  • $\begingroup$ @MarkMitchison I added an exact copy of the problem at the end of my question. The Lagrangian density given seems to indicate that $\phi$ is just a function of x, not of x,t. $\endgroup$ – Physics_Plasma Jul 26 '15 at 22:07
  • 1
    $\begingroup$ The question (v5) seems to be caused by a somewhat misleading exercise formulation. Exercise 17.5 (a) in QFT for the Gifted Amateur deals with a 0+1 dimensional system with Euclidean signature (where Euclidian time is called $x$), not a 1+1 dimensional system with Euclidean signature. $\endgroup$ – Qmechanic Jul 26 '15 at 22:20
  • $\begingroup$ @Qmechanic: I quoted your comment in the EDIT of my answer, I don't know how to link you for more credit. Feel free to do that, if you wish $\endgroup$ – Daniel Jul 26 '15 at 22:46
1
$\begingroup$

The question seems okay to me. It's more or less giving you the time-independent Klein-Gordon equation for a 1D lattice, right?

I think a clarification on how to discretise the derivative would go miles on this question.

If you write $$ \frac{\partial \phi}{\partial x} = \frac{\phi_{j+\frac{1}{2}} - \phi_{j-\frac{1}{2}}}{a} $$ then you can easily just substitute the Fourier expansion they give you and use the Dirac delta function, which in discrete space is given by $$ \delta_{p,q} = \frac{1}{Na} \sum_j{e^{i(p-q)ja}} $$

After cleaning everything up I think you get the end result.

$\endgroup$
  • $\begingroup$ Why would you use $j+\frac{1}{2}$ and $j−\frac{1}{2}$ as oppose to what I used in the squared expression of (3), which is my version of the discretized derivative? $\endgroup$ – Physics_Plasma Sep 16 '15 at 1:33
  • $\begingroup$ They're somewhat equivalent. From a numerical methods point of view, the difference is you've got a forward difference derivative whereas I've used a central difference which generally has an increased order of accuracy. $\endgroup$ – Aran Sep 23 '15 at 19:50
0
$\begingroup$

I'd say the problem doesn't make much sense as it stands. Without terms that include a time derivative, the system would just immediately jump the the configuration with the smallest $L$, as the speed with which it does so is not "punished" (does not increase the action).

In other words, the solution with the smallest action is the one with $L(t)=L_\text{min}$ at all times, except perhaps at $t=t_\text{initial}$.

In a sense though, $x$ and $t$ are just labels, so you could reinterpret your problem, considering $x$ as your time coordinate. $\mathcal L$ would be your Lagrangian and $L$ your action, just as your book says.

EDIT: The problem does actually talk about the Lagrangian $\mathcal L$. So this is not the Lagrangian density. Then $L$ is actually already your action. Also note what Qmechanic wrote in his comment, which I quote below for completeness

The question (v5) seems to be caused by a somewhat misleading exercise formulation. Exercise 17.5 (a) in QFT for the Gifted Amateur deals with a 0+1 dimensional system with Euclidean signature (where Euclidian time is called x), not a 1+1 dimensional system with Euclidean signature.

$\endgroup$
  • $\begingroup$ so you're saying that in the case of a problem with just one variable, let's say $x$, the action is the integral of the Lagrangian over that one variable, even if that variable isn't time? $\endgroup$ – Physics_Plasma Jul 26 '15 at 22:09
  • $\begingroup$ @Physics_Plasma: "Action" is defined as the integral of the Lagrangian over time. So to be technically correct, you would interpret $x$ as "time". Indeed, as Qmechanic points out in a comment to your question $x$ is indeed the time in your problem. $\endgroup$ – Daniel Jul 26 '15 at 22:41
  • $\begingroup$ Ok, thank you. The book explicitly mentions at some point that it will "follow convention" and call both $L$ and $\mathcal{L}$ "Lagrangians", so I didn't take its word choice in this question too literally when it didn't explicitly say "Lagrangian density". $\endgroup$ – Physics_Plasma Jul 26 '15 at 22:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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