In normal modes, we often refer the total potential energy of the system to be:

$$V = q^T B q$$

where $V$ is the total potential energy, $q$ is the coordinates of the system and $B$ is just some matrix.

In this problem I am working on, the total potential energy I have come out to is:

$$V =k/4[(-x_1+x_4+y_1-y_4)^2 + (x_2-x_3+y_2-y_3)^2+2(x_1-x_2)^2 + 2(x_3-x_4)^2 + 2(x_3-x_4)^2 + 2(x_3-x_4)^2 + 2(y_1-y_3)^2 + 2(y_2-y_4)^2] \tag{1}$$

Now, I am trying to find the matrix B, and the method I am familiar with is to compute:

$$V = {\begin{pmatrix} x_1 \\ y_1 \\\vdots \\x_4 \\y_4 \end{pmatrix}}^T \begin{pmatrix} B_{11} & B_{22} & \cdots & B_{18} \\\vdots & \vdots & \ &\vdots \\B_{81} & B_{82} & \cdots & B_{88} \end{pmatrix} \begin{pmatrix} x_1 \\ y_1 \\\vdots \\x_4 \\y_4 \end{pmatrix} \tag{2}$$

in algebraic form and then compare the coefficients of $(2)$ and $(1)$ in order to determine the individual elements $B_{ij}$ noting of course, that for $i \neq j$, $B_{ij} = B_{ji}$ (B is symmetric) as the potential energy is a quadratic form. However, this is a very tedious process and i was wondering if there were any alternative methods to find the matrix B? Thank you very much!

  • $\begingroup$ Note that for example, if you multiply out your expression for $V$ then the quadratic term in $x_1$ is $B_{11}x_1^2$. So $B_{11}$ is the co-efficient of $x_1^2$ which should be reasonable simple to eyeball from your expression for $V$. $\endgroup$
    – jacob1729
    Jun 8, 2019 at 10:45
  • $\begingroup$ Just to be clear: would you be able to do this quickly if your other form for $V$ looked like $V=3x^2+2xy+y^2+...$ (ie if you expanded all the brackets and grouped like terms, which does look a bit tedious in this case)? $\endgroup$
    – jacob1729
    Jun 8, 2019 at 10:46
  • $\begingroup$ @jacob1729 I see. But what about the other elements? are there alternatives? $\endgroup$
    – D. Soul
    Jun 8, 2019 at 23:46
  • $\begingroup$ @jacob1729 I actually already expanded out everything. which is sort of the "normal" but I was wondering if there are better alternatives $\endgroup$
    – D. Soul
    Jun 8, 2019 at 23:48

1 Answer 1


Given an expression in the form of your equation (1) you should be able to go directly to equation (2) without any extra work. Note: the following assumes in equation (1) everything is expanded fully.

Let's use a smaller example so I can type it more easily:

$$V(x,y) = (x-y)^2+2(x+y)^2 \tag{1}$$

this is easier to work with in the expanded form:

$$V(x,y) = 3x^2 + 2xy + 3y^2 \tag{1*}$$

for every mixed entry (in this case just $2xy$) we split it into two reversed terms: $2xy = xy + yx$ to get:

$$ V(x,y) = 3x^2 + xy + yx + 3y^2 $$

(you can do this step mentally, or just recognise where it's going to lead to factors of two in a moment).

But this is in the form:

$$V(x,y) = \begin{pmatrix} x & y \end{pmatrix}\begin{pmatrix} 3 & 1 \\ 1 & 3 \end{pmatrix} \begin{pmatrix} x \\ y \end{pmatrix} \tag{2}$$

where the entry in the $i-j$'th slot in the matrix is simply the co-efficient of $x_ix_j$ in our previous expression for $V$.

To sum up: diagonal entries are coefficients of squared terms, off diagonal entries are half the coefficients of mixed quadratic terms. The factor of two comes from when we split the $2xy=xy+yx$ up, you should be able to convince yourself that it's correct.

  • $\begingroup$ sorry forgot to upvote this $\endgroup$
    – D. Soul
    Jun 23, 2019 at 7:53

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.