I am trying to prove that the matrix expression of the potential energy (Hessian matrix from a Taylor expansion in several variables of the potential) is diagonal considering small oscillations, when written in normal coordinates. But, the accounts do not seem to work out.

Consider a potential of a system of N degrees of freedom $V(q_{1},...,q_{N})=v(q_{i})$ and defining $\vec{q}-\vec{q}_{min}=\vec{\eta}$, where $\vec{q}_{min}$ is a minimum of stable potential:

$$V_{ij}=\frac{\partial V}{\partial q_{i}\partial q_{j}}(\vec{q}=\vec{q}_{min}).$$

The transformation between the generalized coordinates $q_{i}$ and the normal coordinates $\xi _{i}$ is:

$$\vec{q}=A\vec{\xi }+\vec{q}_{min} \Longleftrightarrow q_{i}=\sum_{j}^{}A_{ij}\xi_{j} + q_{i}^{min}$$

where $A$ is the matrix formed by the eigenvectors of the system:

$$\det(V-\lambda ^{(\alpha )}T)=0.$$

So I tried to write $V_{ij}$ in terms of $\xi$ :

$$\frac{\partial V}{\partial \xi _{i} }=\frac{\partial V}{\partial q_{i}} \frac{\partial q_{i}}{\partial \xi _{i}}=\frac{\partial V}{\partial q_{i}}\sum_{j}^{}A_{ij}\delta _{ij}=A_{ii}\frac{\partial V}{\partial q_{i}}$$

$$\frac{\partial^{2} V}{\partial \xi _{i}^{2} }=\frac{\partial }{\partial \xi _{i}} (A_{ii}\frac{\partial V}{\partial q_{i}})=A_{ii} \frac{\partial [\frac{\partial V}{\partial q_{i}}] }{\partial q_{i}} \frac{\partial q_{i}}{\partial \xi_{i}}=\frac{\partial^{2} V}{\partial q _{i}^{2} }A_{ii}^{2}$$

but for cross derivatives I get a nonzero result:

$$\frac{\partial V}{\partial \xi _{j}\partial \xi _{i}}=\frac{\partial V}{\partial \xi _{j}}(A_{ii}\frac{\partial V}{\partial q_{i}})=A_{ii}\frac{\partial[\frac{\partial V}{\partial q_{i}}] }{\partial q_{k}}\frac{\partial q_{k}}{\partial \xi_{j} }=A_{ii} \frac{\partial ^{2}V}{\partial q_{i} \partial q_{k}}\sum_{j}^{}A_{ij}\delta _{kj}=A_{ik}A_{ii} \frac{\partial ^{2}V}{\partial q_{i} \partial q_{k}}\neq 0$$ taking into account that we can exchange the index $k$ for $j$.

What could I do wrong? Where did I go wrong?


\begin{align*} &\text{with the potential energy $~V=V(\boldsymbol q)~$ and the kinetic enegry $~T=\frac 12\, \boldsymbol{\dot q}\cdot \boldsymbol{\dot q} ~$}\\ &\text{ you obtain this non linear copuled differential equations}\\\\ &\boldsymbol{\ddot{q}}=\underbrace{\left[\frac{\partial}{\partial \boldsymbol q}\left(\frac{\partial V}{\partial \boldsymbol{q}}\right)\right]}_{\boldsymbol K}\,\boldsymbol q=\boldsymbol{K(\boldsymbol q)}\,\boldsymbol{q}\tag 1 \end{align*} where

  • $~\boldsymbol q~$ are the $n$ generalized coordinates
  • $~\boldsymbol K~$ $n\times n~$ symmetric matrix

substitute $~\boldsymbol q=\boldsymbol A\,\boldsymbol{\xi}+\boldsymbol{\xi}_{\text{min}}$ into Eq. (1) you obtain: (where $~\boldsymbol A~$ is the matrix of the eigen vectors) \begin{align*} &\boldsymbol{A}\,\boldsymbol{\ddot{\xi}}=\left[\boldsymbol{K}\left(\boldsymbol A\,\boldsymbol{\xi}+\boldsymbol{\xi}_{\text{min}}\right)\right]\, \left(\boldsymbol A\,\boldsymbol{\xi}+\boldsymbol{\xi}_{\text{min}}\right)\overset{\text{linear}}{=} \boldsymbol K(\boldsymbol q=\boldsymbol\xi_{\text{min}})\,\boldsymbol A\,\boldsymbol\xi\\\\ &\boldsymbol{\ddot{\xi}}=\underbrace{\boldsymbol{A}^{-1}\,\boldsymbol{K}_L\,\boldsymbol A}_{\boldsymbol D}\,\boldsymbol\xi\tag 2\\ &\text{where}\\ &\boldsymbol D=\text{Diag}\left[\lambda_1~,\lambda_2\ldots~,\lambda_n\right]\\ &\text{and}\\ &\boldsymbol{K}_L=\boldsymbol{K}\bigg|_{\boldsymbol q=\boldsymbol\xi_{\text{min}}} \end{align*} thus Eq. (2) is a linear uncoupled differential equation


You were messy some of the indexes. First, remind that each column vector is normalized eigen vector of matrix $\bf{V}$, therefore $ \bf{V} \hat{A}_k = \lambda_k \hat{A}_k$. $\hat{A}_k$ is the $k_{th}$ column vectoe of mtrix $\bf{k}$, in terms of index form:

$$ \sum_j V_{ij} A_{jk} = \lambda_k A_{ik} $$

I try to re-write your equations:

$$ \frac{\partial V}{\partial \xi_i} = \sum_j \frac{\partial V}{\partial q_j}\frac{\partial q_j}{\partial \xi_i} = \sum_j \frac{\partial V}{\partial q_j}\frac{\partial }{\partial \xi_i} \sum_k A_{jk} \xi_k =\sum_j \frac{\partial V}{\partial q_j} \sum_k A_{jk} \delta_{ik} = \sum_j \frac{\partial V}{\partial q_j} A_{ji} $$

$$ \frac{\partial^{2} V}{\partial \xi _{i}^{2} }=\frac{\partial }{\partial \xi _{i}} \sum_j A_{ji}\frac{\partial V}{\partial q_{j}}= \sum_k \frac{\partial q_k}{\partial \xi _{i}} \frac{\partial}{\partial q_k} \left( \sum_j A_{ji}\frac{\partial V}{\partial q_{j}} \right) = \sum_k A_{ki} \frac{\partial}{\partial q_k} \left( \sum_j A_{ji}\frac{\partial V}{\partial q_{j}} \right) \\ = \sum_j\sum_k A_{ki}A_{ji} \frac{\partial^2 V}{\partial q_k \partial q_j} =\sum_j\sum_k A_{ki}A_{ji} V_{kj} = \lambda_i \sum_k A_{ki} A_{ki} = \lambda_i. $$

$$ \frac{\partial^{2} V}{\partial \xi _{p} \partial \xi _{q}}=\frac{\partial }{\partial \xi_ {p}} \sum_j A_{jq}\frac{\partial V}{\partial q_{j}}= \sum_k \frac{\partial q_k}{\partial \xi _{p}} \frac{\partial}{\partial q_k} \left( \sum_j A_{jq}\frac{\partial V}{\partial q_{j}} \right) = \sum_k A_{kp} \frac{\partial}{\partial q_k} \left( \sum_j A_{jq}\frac{\partial V}{\partial q_{j}} \right) \\ = \sum_j\sum_k A_{kp}A_{jq} \frac{\partial^2 V}{\partial q_k \partial q_j} = \sum_j\sum_k A_{kp}A_{jq} V_{kj} = \lambda_q \sum_k A_{kp} A_{kq} = 0 $$

  • $\begingroup$ Thanks, it has been very helpful, just a question, the eigenvalue equation of the system has the form $V \cdot \vec{a_{k}} = \lambda_ {k} T \cdot \vec {a_ {k}}$ , then the matrix A made up of the column vectors $\vec{a_{k}}$ and the diagonal matrix $ D = diag (\lambda_ {1}, .., \lambda_ {k}, ..,\lambda_ {N}) $ what is the matrix that is diagonalized in the form $ ADA^{-1}$?. $\endgroup$
    – Cast fj
    Feb 5 '21 at 15:48
  • $\begingroup$ $\bf{D} = \bf{A}^T \bf{V} \bf{A}$. The matrix $\bf{V}$ is diagonalized by orthogonal matrix $\bf{A}$ intp digonal matrix $\bf{D}$. Therefrore, the expansion of potential near the equilibrium point $q_{min}$ will have no cross terms. $V(q) = \sum_i\sum_j V_{ij} q_i q_j = \sum_i D_{ii} \xi_i^2$. This is call dynamical matrix. $\endgroup$
    – ytlu
    Feb 6 '21 at 14:06

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.