Izvestiya: Mathematics
 RUS  ENG JOURNALS   PEOPLE   ORGANISATIONS   CONFERENCES   SEMINARS   VIDEO LIBRARY   PACKAGE AMSBIB
 General information Latest issue Forthcoming papers Archive Impact factor Guidelines for authors Submit a manuscript Search papers Search references RSS Latest issue Current issues Archive issues What is RSS

 Izv. RAN. Ser. Mat.: Year: Volume: Issue: Page: Find

 Izvestiya: Mathematics, 2023, Volume 87, Issue 5, Pages 941–946DOI: https://doi.org/10.4213/im9356e (Mi im9356)

Symmetries and conservation laws of the Liouville equation

V. V. Zharinov

Steklov Mathematical Institute of Russian Academy of Sciences, Moscow, Russia

### § 1. Introduction

Symmetries and conservation laws of the Liouville equation are studied in the frames of the algebra-geometrical approach to partial differential equations. Interest in the Liouville equation and its properties has intensified recently due to its important role in functional mechanics proposed by I. V. Volovich (see [1]–[3]).

### 1.1. Main ingredients

The article is quite computational, to make it more readable we first clarify the designations. Namely, below:

The main functional algebras are:

Remark 1. A number $r\in\mathbb{Z}_+$ is called the order of a function $f(t,\mathrm{x},\mathbf{u})\in\mathcal{A}$, $\operatorname{ord} f= r$, if the partial derivative $\partial_{u_\mathrm{i}}f\ne0$ for some index $\mathrm{i}\in\mathbb{I}$, $|\mathrm{i}|=r$, while the partial derivatives $\partial_{u_\mathrm{i}}f=0$ for all indexes $\mathrm{i}\in\mathbb{I}$, $|\mathrm{i}|>r$.

The main linear operators are:

### 1.2. The Liouville equation

The Liouville equation has the form

$$$$L\phi=\partial_t \phi+\partial_{x^\mu}(v^\mu\cdot \phi) =\frac{d\phi}{dt} +\operatorname{div}\mathrm{v}\cdot\phi=0,$$ \tag{1}$$
where $\phi(t,\mathrm{x})\in\Phi$ is an unknown function, $\mathrm{v} \colon \mathrm{X}\to\mathrm{X}$ is a given vector field.

Theorem 1. The general solution $\phi(t,\mathrm{x})\in\Phi$ of the Liouville equation (1) is implicitly defined by the equality

$$\begin{equation*} K(\varkappa,s^1,\dots,s^m)=0, \end{equation*} \notag$$
where $K$ is an arbitrary function, while $\varkappa(t,\mathrm{x},\phi)$, $\mathrm{s} =(s^1(t,\mathrm{x}),\dots,s^m(t,\mathrm{x}))$ are the first integrals of the system of the ordinary differential equations
$$\begin{equation*} \frac{dt}1=\frac{dx^1}{v^1}=\dots=\frac{dx^m}{v^m} =-\frac{d\phi}{\operatorname{div}\mathrm{v}\cdot\phi}. \end{equation*} \notag$$

The proof of this and the analogous theorems below follows from the theory of differential equations in partial derivatives of the first order (see [4], for example).

The situation simplifies when the vector field $\mathrm{v}$ is divergence free, the general solution can be found in the explicit form. Indeed, if $\operatorname{div}\mathrm{v}=0$, then the Liouville equation (1) takes the form

$$$$L\phi=\frac{d\phi}{dt}=\partial_t\phi+v^\mu\partial_{x^\mu}\phi=0,$$ \tag{2}$$
and has the general solution $\phi=\rho(s)$, where $\rho$ is an arbitrary function, while $\mathrm{s}=(s^1(t,\mathrm{x}),\dots,s^m(t,\mathrm{x}))$ are the first integrals of the shorten system
$$\begin{equation*} \frac{dt}1=\frac{dx^1}{v^1}=\dots=\frac{dx^m}{v^m}. \end{equation*} \notag$$

### § 2. Symmetries

The Lie algebra of symmetries of the Liouville equation (1) is defined as

$$$$\operatorname{Sym}(L)=\{\zeta=\operatorname{ev}_g \mid g\in\mathcal{A}, \ \mathbf{L} g =D_tg+D_\mu(v^\mu g)=0\}\simeq\operatorname{Ker}\mathbf{L},$$ \tag{3}$$
where the evolution derivation $\operatorname{ev}_g=D_\mathrm{i} g\cdot\partial_{u_\mathrm{i}}$, the evolution Lie bracket
$$\begin{equation*} [\operatorname{ev}_g,\operatorname{ev}_h]=\operatorname{ev}_{\{g,h\}}, \qquad \{g,h\}=\operatorname{ev}_gh-\operatorname{ev}_hg, \quad g,h\in\mathcal{A} \end{equation*} \notag$$
(see [5]–[7], for example). In particular, the linear space $\operatorname{Ker}\mathbf{L}$ here possesses the induced structure of the Lie algebra with the Lie bracket $\{\,{\cdot}\,,{\cdot}\,\}$, where
$$\begin{equation*} \operatorname{Ker}\mathbf{L}\times\operatorname{Ker}\mathbf{L}\ni(g,h) \mapsto\{g,h\}\in\operatorname{Ker}\mathbf{L}. \end{equation*} \notag$$
In more detail, the equation $\mathbf{L} g=0$ is written as $Lg-W_\mathrm{i}\cdot\partial_{u_\mathrm{i}}g=0$, where
$$\begin{equation*} \begin{gathered} \, Lg =\partial_tg+v^\mu\cdot\partial_{x^\mu}g+\operatorname{div}\mathrm{v}\cdot g, \\ W_\mathrm{i}=\sum_{\mathrm{j}+\mathrm{k}=\mathrm{i},\, \mathrm{j}\ne0}\binom{\mathrm{i}}{\mathrm{k}} \partial_{x^\mathrm{j}}v^\mu\cdot u_{\mathrm{k}+(\mu)} +\sum_{\mathrm{j}+\mathrm{k}=\mathrm{i}} \binom{\mathrm{i}}{\mathrm{k}}\partial_{x^\mathrm{j}}\operatorname{div}\mathrm{v} \cdot u_\mathrm{k}, \end{gathered} \end{equation*} \notag$$
$\partial_{\mathrm{x}^\mathrm{j}}=(\partial_{x^1})^{j^1}\circ\dots\circ(\partial_{x^m})^{j^m}$, $\mathrm{j}=(j^1,\dots,j^m)\in\mathbb{I}$.

Proposition 1. The linear operator $\mathbf{L}=L-W_\mathrm{i}\cdot\partial_{u^\mathrm{i}} \colon \mathcal{A}\to\mathcal{A}$ is the order-nonincreasing, that is, $\operatorname{ord}(\mathbf{L} g)\leqslant\operatorname{ord} g$ for any $g\in\mathcal{A}$.

Theorem 2. The equation $\mathbf{L} g=0$ has the general solution $g(t,\mathrm{x},\mathbf{u})\in\mathcal{A}$ implicitly defined by the equality

$$\begin{equation*} \mathbf{K}(\mathrm{s},\mathbf{w},\varkappa)=0, \end{equation*} \notag$$
where $\mathbf{K}$ is an arbitrary function of a finite order, while $\mathrm{s}(t,x)$, $\mathbf{w}(t,x,\mathbf{u})$, and $\varkappa(t,x,\mathbf{u},g)$, are the first integrals of the system
$$\begin{equation*} \frac{dt}{1}=\frac{dx^\mu}{v^\mu}=-\frac{du_\mathrm{i}}{W_\mathrm{i}} =-\frac{dg}{\operatorname{div}\mathrm{v}\cdot g}, \qquad \mu\in\mathbf{m}, \quad \mathrm{i}\in\mathbb{I} \quad (\textit{without summation}!), \end{equation*} \notag$$
here $\mathrm{i}=(i^\mu)\in\mathbb{I}$, $\mathbf{u}=(u_\mathrm{i})$, $\mathbf{w}=(w_\mathrm{i})\in\mathbb{R}_\mathbb{I}$.

Remark 2. Note that the functions $\mathrm{s}(t,x)$ here are solutions of the Liouville equation (2).

Again, the situation simplifies when the vector field $\mathrm{v}$ is divergence free. Indeed, if $\operatorname{div}\mathrm{v}=0$, the equation $\mathbf{L} g=0$ takes the form

$$\begin{equation*} \mathbf{L} g=\partial_tg+v^\mu\cdot\partial_{x^\mu}g -W'_\mathrm{i}\cdot\partial_{u_\mathrm{i}}g=0, \qquad W'_\mathrm{i}=\sum_{\mathrm{j}+\mathrm{k}=\mathrm{i},\, \mathrm{j}\ne0} \binom{\mathrm{i}}{\mathrm{k}} \partial_{x^\mathrm{j}}v^\mu\cdot u_{\mathrm{k}+(\mu)}, \end{equation*} \notag$$
and has the general solution $g=\rho(\mathrm{s},\mathbf{w})$, where $\rho$ is an arbitrary function of a finite order, while $\mathrm{s}(t,x)$, $\mathbf{w}(t,x,\mathbf{u})$ are the first integrals of the shorten system
$$\begin{equation*} \frac{dt}{1}=\frac{dx^\mu}{v^\mu}=-\frac{du_\mathrm{i}}{W'_\mathrm{i}} =\frac{du_0}0, \qquad \mu\in\mathbf{m}, \quad \mathrm{i}\in\mathbb{I}\setminus\{0\} \quad \text{(without summation!)}. \end{equation*} \notag$$

### § 3. Conservation laws

The linear space of conservation laws for the Liouville equation $L\phi=0$ is defined as the factor-space

$$$$\mathrm{CL}(L)=\{\rho\in\mathcal{A} \mid D_t\rho\in\operatorname{Div}\mathcal{A}^\mathbf{m}\} \bigm/ \{\rho\in\operatorname{Ker}\delta_u\},$$ \tag{4}$$
where $\operatorname{Div}\mathrm{J}=D_\mu J^\mu$, $\mathrm{J}=(J^\mu)\in\mathcal{A}^\mathbf{m}$, the variational derivative
$$\begin{equation*} \delta_u \colon \mathcal{A}\to\mathcal{A}, \qquad \rho\mapsto\delta_u\rho =(-D)_\mathrm{i}(\partial_{u_\mathrm{i}}\rho) \end{equation*} \notag$$
(see [5]–[8], for example).

Theorem 3. There is defined the isomorphism of the linear spaces

$$$$\delta_u \colon \mathrm{CL}(L)\simeq\{\chi\in\operatorname{Ker}\mathbf{D}_t \mid \chi_*=\chi^*\}, \qquad [\rho]\mapsto\chi=\delta_u\rho,$$ \tag{5}$$
where $[\rho]=\rho+\operatorname{Ker}\delta_u$, $\chi\in\mathcal{A}$,

Note that the linear operator $\mathbf{D}_t \colon \mathcal{A}\to\mathcal{A}$ is the order-nonincreasing.

Theorem 4. The equation $\mathbf{D}_t\chi=0$ has the general solution $\chi=R(\mathrm{s},\mathbf{w})$, where $R$ is an arbitrary function of a finite order, while $\mathrm{s}(t,x)$, $\mathbf{w}(t,x,\mathbf{u})$ are the first integrals of the system

$$\begin{equation*} \frac{dt}{1}=\frac{dx^\mu}{v^\mu}=-\frac{du_\mathrm{i}}{W'_\mathrm{i}} =\frac{du_0}0, \qquad \mu\in\mathbf{m}, \quad \mathrm{i}\in\mathbb{I}\setminus\{0\} \quad \textit{(without summation!)}, \end{equation*} \notag$$
here $\mathrm{i}=(i^\mu)\in\mathbb{I}$, $\mathbf{u}=(u_\mathrm{i})$, $\mathbf{w}=(w_\mathrm{i})\in\mathbb{R}_\mathbb{I}$.

Remark 3. Note, $\mathbf{D}_t=\mathbf{L}|_{\operatorname{div}\mathrm{v}=0}$.

### § 4. Examples

Example 1. Let $\mathrm{x}=(x^1,x^2)\in\mathrm{X}=\mathbb{R}^2$, $x^1=q$, $x^2=p$, $H=\frac{p^2}{2m}$, then $\mathrm{v}=(v^1,v^2)$, $v^1=\partial_{x^2}H=\frac{p}{m}$, $v^2=-\partial_{x^1}H=0$ (see Example 6.1 in [9]), so here

$$\begin{equation*} L=\frac d{dt}=\partial_t+\frac pm\, \partial_q, \qquad \mathbf{L}=\mathbf{D}_t=L-W_{j,k}\cdot\partial_{u_{j,k}}, \quad W_{j,k}=\frac{k}{m} u_{j+1,k-1}, \end{equation*} \notag$$
where $\mathrm{i}=(j,k)\in\mathbb{I}=\mathbb{Z}_+\times\mathbb{Z}_+$.

In particular, the equation $\mathbf{D}_t\chi=0$ has the general solution $\chi=R(r,s,\mathbf{w})$, where $R(r,s,\mathbf{w})$ is an arbitrary function of a finite order, while $r(t,q,p)$, $s(t,q,p)$, and $\mathbf{w}(t,q,p,\mathbf{u})$ are the first integrals of the system

$$\begin{equation*} \frac{dt}1=\frac{m\,dq}p=\frac{dp}0=-\frac{du_{j,k}}{W_{j,k}}, \qquad j,k\in\mathbb{Z}_+ \quad \text{(without summation!)}, \end{equation*} \notag$$
here $\mathrm{i}=(j,k)\in\mathbb{I}=\mathbb{Z}_+\times\mathbb{Z}_+$, $\mathbf{u}=(u_{j,k})$, $\mathbf{w}=(w_{j,k})$.

Proposition 2. In the above settings, the first integrals $r$, $s$, and $\mathbf{w}$ are:

where $\tau=\frac{t}{m}$.

Corollary 1. The Lie algebra $\operatorname{Sym}(L)\simeq\operatorname{Ker}\mathbf{D}_t$.

Corollary 2. The linear space $\mathrm{CL}(L)\simeq\{\chi\in\operatorname{Ker}\mathbf{D}_t \mid \chi_*=\chi^*\}$. In addition,

$$\begin{equation*} r_*=r^*=0, \qquad s_*=s^*=0, \qquad (w_{j,k})_*=(-1)^{j+k}(w_{j,k})^*. \end{equation*} \notag$$

Remark 4. One may say that $\mathrm{CL}(L)\subset\operatorname{Sym}(L)$.

Example 2. Let again $\mathrm{x}=(x^1,x^2)\in\mathrm{X}=\mathbb{R}^2$, $x^1=q$, $x^2=p$, but the Hamiltonian $H=\frac{p^2}{2m}+\frac{\varkappa q^2}2$. In this case, $\partial_{x^2}H=\frac{p}{m}$, $\partial_{x^1}H=\varkappa q$, so the vector field $\mathrm{v}=\bigl(\frac{p}{m},-\varkappa q\bigr)$ (see Example 6.3 in [9]), so here

$$\begin{equation*} L=\frac d{dt}=\partial_t+\frac{p}{m}\,\partial_q-\varkappa q\,\partial_p , \qquad \mathbf{L}=\mathbf{D}_t=L-W_{j,k}\,\partial_{u_{j,k}}, \end{equation*} \notag$$
where $\mathrm{i}=(j,k)\in\mathbb{I}=\mathbb{Z}_+\times\mathbb{Z}_+$, $W_{j,k}=-j\varkappa u_{j-1,k+1}+\frac{k}{m} u_{j+1,k-1}$, $j,k\in\mathbb{Z}_+$.

The equation $\mathbf{D}_t\chi=0$ has the general solution $\chi=R(r,s,\mathbf{w})$, where $R$ is an arbitrary function of a finite order, while $r(t,q,p)$, $s(t,q,p)$, and $\mathbf{w}(t,q,p,\mathbf{u})$ are the first integrals of the system

$$\begin{equation*} \frac{dt}1=\frac{mdq}p=-\frac{dp}{\varkappa q}=-\frac{du_{j,k}}{W_{j,k}}, \qquad j,k\in\mathbb{Z}_+ \quad \text{(without summation!)}, \end{equation*} \notag$$
here $\mathrm{i}=(j,k)\in\mathbb{I}=\mathbb{Z}_+\times\mathbb{Z}_+$, $\mathbf{u}=(u_{j,k})$, $\mathbf{w}=(w_{j,k})$.

In the above setting, the first integrals $r$, $s$ are as follows (see Example 6.3 in [9]):

The first integrals $w_{j,k}$ are defined by the equations

$$\begin{equation*} -j\varkappa u_{j-1,k+1}+\dot u_{j,k}+\frac{k}{m} u_{j+1,k-1}=0, \qquad j,k\in\mathbb{Z}_+, \quad \dot u=\frac{du(t)}{dt}. \end{equation*} \notag$$
Simple but tedious calculations give:

(0) $j+k=0$:

(1) $j+k=1$: hence (2) $j+k=2$: hence (3) $j+k=3$: hence and so on, for $j+k=4,5,\dots$ .

Again, the statements from the previous example are true. Namely:

#### Bibliography

1. I. V. Volovich, Time irreversibility problem and functional formulation of classical mechanics, arXiv: 0907.2445v1
2. I. V. Volovich, “Bogoliubov equations and functional mechanics”, Theoret. and Math. Phys., 164:3 (2010), 1128–1135
3. I. V. Volovich, “Functional stochastic classical mechanics”, $p$-Adic Numbers Ultrametric Anal. Appl., 7:1 (2015), 56–70
4. R. Courant and D. Hilbert, Methods of mathematical physics, v. II, Wiley Classics Library, Partial differential equations, Reprint of the 1962 original, John Wiley, New York, 1989
5. P. J. Olver, Applications of Lie groups to differential equations, Grad. Texts in Math., 107, Springer-Verlag, New York, 1986      ; 2nd ed., 1993
6. Symmetries and conservation laws for differential equations of mathematical physics, Transl. Math. Monogr., 182, eds. I. S. Krasil'shchik and A. M. Vinogradov, Amer. Math. Soc., Providence, RI, 1999
7. V. V. Zharinov, Lecture notes on geometrical aspects of partial differential equations, Ser. Soviet East European Math., 9, World Sci. Publ., River Edge, NJ, 1992
8. V. V. Zharinov, “Conservation laws of evolution systems”, Theoret. and Math. Phys., 68:2 (1986), 745–751
9. V. V. Zharinov, Dynamics of wave packets in the functional mechanics, arXiv: 2203.12028v1

 Statistics & downloads: Abstract page: 333 Russian version PDF: 6 English version PDF: 50 Russian version HTML: 55 English version HTML: 136 References: 87 First page: 12