betaFeedback

WorksPhysics and Astronomy

Classical Dynamical Symmetries and Geometry of Trajectories

Abstract

It is shown that all spherical symmetric potentials are capable of producing dynamical symmetries in classical one-body motions, thanks to the inevitable existence of symmetry axes associated with turning points for corresponding trajectories. This will definitely expand the class of maximally superintegrable one-body motions in central potentials that until now was considered to include only the Newtonian and Hookean cases. A simple method is proposed to identify and characterize these dynamical symmetries.

Introduction

In classical and non-relativistic mechanics, the one-body motion in a spherical symmetric potential is certainly a well known and long studied problem. However, in a general case, the question of whether such system has an additional dynamical (or “hidden”) symmetry has remained unanswered until now. This is because of ignorance of what might be the true origin of such a symmetry and lack of a simple method to reveal and describe it. In this regard, the general method of Lie’s transformations of differential equations is disappointing as it leads either to intractable equations when using contact transformations or to a well known result (mechanical similarity) when using point transformations: see [1]. Discovering a dynamical symmetry could therefore appear as a matter of luck. Yet, the two famous examples of the Newtonian and the Hookean systems suggest a strong and simple correlation between the appearance of an internal symmetry and the geometrical properties of the trajectories. For the Newtonian system with the attractive interaction potential V(r)=−K/rV(r)=-K/r (K>0K>0), its dynamical symmetry is characterized by the celebrated Laplace-Runge-Lenz (LRL) vector, Refs. [2, 3, 4], which is found to be along a symmetry axis for each trajectory. Together with the angular momentum, this vector generates a group of transformation of trajectories of the same species, currently identified as SO(4)SO(4) for ellipses or SO(3,1)SO(3,1) for hyperbolas. For the Hookean system with the harmonic potential V(r)=K′r2/2V(r)=K'r^2/2 (K′>0K'>0), its dynamical symmetry is characterized by one of the two vectors defining the symmetry axes of trajectories, and that vector also takes part in a group of transformation of the corresponding trajectories (ellipses): see Ref. [5]. Taking this correlation as a guiding idea, we show below how it works as well for any other spherical symmetric potential and we develop a simple method to identify and characterize the resulting dynamical symmetry.

The correlation between dynamical symmetries and geometrical properties of trajectories for central potentials has already been considered by various authors in the past, Ref. [6, 7, 8]. These authors can be considered as precursors of the subject although they treated it in a rather vague way. The authors of Ref. [6] gave some importance to a characteristic point of trajectories from which a symmetry could emerge, such as the perihelion for ellipses, but added in a footnote that “Every point of a trajectory can be considered as a characteristic point, except of course if the trajectory is a circle”, while the author of Ref. [7] said that “O4O_4 and SU3SU_3 are dynamical symmetries for a central potential essentially because motion takes place in a fixed plane”. In fact, all these articles have mainly focused on mathematical aspects of symmetries from the point of view of group theory (with possible extension to quantum mechanics), and none of them has emphasized what justifies the very existence of a symmetry group, namely its role with respect to trajectories which consists in a continuous mapping of trajectories of a same species, see Ref. [1].

Main known facts about trajectories

The motion considered here is that of a particle PP in an inertial frame having as origin the source OO of a spherical symmetric potential acting on PP. We will use the following notations: r\boldsymbol{r} is the position vector of PP relative to OO, and r=∥r∥r=\lVert\boldsymbol{r}\rVert; v\boldsymbol{v} and p=mv\boldsymbol{p}=m\boldsymbol{v} are its velocity and momentum vectors, respectively, and v=∥v∥v=\lVert\boldsymbol{v}\rVert, p=∥p∥=mvp=\lVert\boldsymbol{p}\rVert=mv; finally, implicit summation on repeated indices is also assumed.

Let us recall some known facts and formulas about trajectories.

Thanks to the spherical symmetry of the potential, the angular momentum of the particle, L=r×p\boldsymbol{L}=\boldsymbol{r}\times\boldsymbol{p}, is a conserved vector, and consequently its trajectory corresponding to a given value of L\boldsymbol{L} lies in a plane perpendicular to the latter and containing OO. As the orientation of axes can be freely chosen, OzOz is taken along LL, the plane of motion being then the plane (x,y)(x,y). The position of PP in that plane may be defined as usual by the distance r=OPr=OP and the polar angle θ\theta of r\mathbf{r} relative to the axis OxOx. Since z=0z=0, pz=0p_z=0, only the component LzL_z of the angular momentum is non-zero, and we have

Lz=∥L∥=L=mr2θ˙=constL_z=\lVert\mathbf{L}\rVert=L=mr^2\dot{\theta}=\mathrm{const}

from which the Kepler’s second law (law of equal areas) is deduced. In the following, we exclude the case L=0L=0 corresponding to motions along straight lines passing through OO. The potential being independent of time, the Hamiltonian (or energy) of PP, H=p22m+V(r)H=\frac{p^2}{2m}+V(r), is also a conserved quantity EE. The radial and orthoradial components of p\mathbf{p} being respectively pr=mr˙p_r=m\dot{r} and pθ=mrθ˙=L/rp_\theta=mr\dot{\theta}=L/r, we reexpress this Hamiltonian as

H=mr˙22+U(r),whereU(r)=L22mr2+V(r)(1)H=\frac{m\dot{r}^{2}}{2}+U(r),\quad\text{where}\quad U(r)=\frac{L^2}{2mr^2}+V(r) \tag*{(1)}

is the effective potential. Writing dH/dt=0dH/dt=0, the distance rr is found to vary in time according to the equation

p˙r=mr¨=−dUdr(2)\dot{p}_r=m\ddot{r}=-\frac{dU}{dr} \tag*{(2)}

From Eq. (1), permissible motions of PP satisfy

pr22m=E−U(r)≥0\frac{p_r^2}{2m}=E-U(r)\ge0

Turning points are those for which pr=0p_r=0 or U(r)=EU(r)=E. Such points are inevitably present in closed orbits but may also exist for non-closed and unbounded trajectories. At these points, prp_r changes sign. The modulus and the polar components of the velocity are

v=2m(E−V(r)),vθ=Lmr,v=\sqrt{2m\left(E-V(r)\right)},\quad v_\theta=\frac{L}{mr},
vr=±2m(E−V(r))−L2mr2(3)v_r=\pm\sqrt{2m\left(E-V(r)\right)-\frac{L^2}{mr^2}} \tag*{(3)}

the sign of vrv_r changing at turning points only. The unit vector of the tangent, T=v/v\mathbf{T}=\mathbf{v}/v, has polar components Tr=vr/vT_r=v_r/v, Tθ=vθ/v=L/(mrv)T_\theta=v_\theta/v=L/(mrv). According to Eq. (3), at a given rr, only the component TrT_r may be different by a simple change of sign. Let us then suppose that the curve representing the trajectory intersects itself at some point. This might occur for closed orbits, when the particle turns many times around OO. As the tangent can have only two orientations for a given rr, this point is a double point (a node). Note also that vrv_r is linked to the derivative of the polar equation r(θ)r(\theta) since

drdθ=r˙/θ˙=mr2r˙L(4)\frac{dr}{d\theta}=\dot{r}/\dot{\theta}=\frac{mr^2\dot{r}}{L} \tag*{(4)}

and that turning points satisfy as well drdθ=0\frac{dr}{d\theta}=0. If a trajectory is symmetrical relatively to an axis passing through OO and defined by θ=θ0\theta=\theta_0, we must have r(θ0+α)=r(θ0−α)r(\theta_0+\alpha)=r(\theta_0-\alpha) which means that drdθ(θ0)=0\frac{dr}{d\theta}(\theta_0)=0 (assuming r(θ0)r(\theta_0) well defined) : the point of the trajectory defined by θ=θ0\theta=\theta_0 and r=r(θ0)r=r(\theta_0) is a turning point.

Using Eq. (4), the polar angle of the radius vector r\mathbf{r} relatively to some initial position OP0=r0er0\mathbf{OP}_0=r_0\mathbf{e}_{r_0} is expressed as

θ=θ(r)=θ0+L22m∫r0r±dRR2E−U(R)(5)\theta=\theta(r)=\theta_0+\sqrt{\frac{L^2}{2m}}\int_{r_0}^{r}\frac{\pm dR}{R^2\sqrt{E-U(R)}} \tag*{(5)}

where the sign of the integrand should be chosen such that θ˙=L/(mr2)\dot{\theta}=L/(mr^2) is always strictly positive : during its motion, the particle PP tends to rotate around OO in the same direction of rotation. The complete curve of a closed trajectory necessarily surrounds the source OO. Such a trajectory is bounded and lies inside a circular crown defined by rm≤r≤rMr_m\le r\le r_M. Points PmP_m and PMP_M where rr takes its respective extreme values rmr_m and rMr_M are turning points. It can be shown that OPm\mathbf{OP}_m and OPM\mathbf{OP}_M define symmetry axes of the trajectory : see Appendix § 6. This feature is indeed the direct consequence of the spherical symmetry of the potential, which allows us to express all kinematical quantities as functions of the single variable rr (apart from the various constants) : see Eqs. (3) and (5).

The polar equation r(θ)r(\theta) of a closed orbit is necessarily a periodic function where θ\theta should be considered as varying from −∞-\infty to +∞+\infty. If the angular period of the motion is larger than 2π2\pi, the curve of the orbit makes more than one tour around OO during a period, and inevitably intersects itself in at least one double point at an intermediate distance between rmr_m and rMr_M. It can be shown that such double points also define symmetry axes of the orbit : see Appendix § 6.

Consider a trajectory confined in a circular crown rm≤r≤rMr_m\le r\le r_M and find out on which condition it is closed. Because θ˙>0\dot{\theta}>0, the particle turns indefinitely around the source OO and its trajectory is alternately tangent to the circles of radius rmr_m and rMr_M at turning points. When moving from a turning point P1P_1 at r=rmr=r_m to the next turning point P2P_2 at r=rMr=r_M, the vector r\mathbf{r} rotates by the angle

Δθ=L2m∫rmrMdrr2E−U(r)(6)\Delta\theta=\frac{L}{\sqrt{2m}}\int_{r_m}^{r_M}\frac{dr}{r^2\sqrt{E-U(r)}} \tag*{(6)}

see (2.5), and rotates by the same angle when moving from point P2P_2 to the next turning point P3P_3 at r=rmr = r_m, so that the angle between the two vectors OP1\mathbf{OP}_1 and OP3\mathbf{OP}_3 is 2Δθ2\Delta\theta. The trajectory is closed if after an integer MM of such rotations the final turning point is the same as the first one. As these rotations are equivalent to a rotation of angle 2πQ2\pi Q around OO, QQ being integer, we must have 2MΔθ=2πQ2M\Delta\theta= 2\pi Q or Δθ=Qπ/M\Delta\theta= Q\pi/M, which means that the trajectory is closed if and only if Δθ\Delta\theta is commensurable with π\pi. It has been demonstrated by Bertrand, Ref. [9], that only two attractive potentials fulfill this requirement at any distance from the source and for any corresponding permissible values of EE and LL, with (mod. π\pi) Δθ=π\Delta\theta= \pi for the Newtonian potential V(r)=−KN/rV(r) = -K_N/r (KN>0K_N > 0) and Δθ=π/2\Delta\theta= \pi/2 for the Hookean potential V(r)=KHr2/2V(r) = K_Hr^2/2 (KH>0K_H > 0). For these two cases only, all bound trajectories are closed.

A closed orbit has, at least, one couple of turning points, and the separation angle Δθ\Delta\theta between their corresponding axes of symmetry is certainly commensurable with π\pi. Note that if the tangent is required to be unique in each point of the curve, the period must be exactly 2π2\pi. The orbit is then realized in only one tour and is a simple curve. If c≥1c \ge1 is the number of couples of turning points like (P1,P2)(P_1, P_2) defined above, we then have c(2Δθ)=2πc(2\Delta\theta) = 2\pi, and Δθ=π/c\Delta\theta= \pi/c.

Further, it can be shown that the condition for a potential to provide, at any distance from the source, stable circular orbits and closed quasi-circular trajectories, is to be of the form V(r)=a/rηV(r) = a/r^\eta, with η=2−α2<2\eta= 2 - \alpha^2 < 2, α\alpha being a rational number, and aη<0a\eta< 0: see Ref. [10] §3.6. For these closed quasi-circular orbits, we have Δθ=π/2−η\Delta\theta= \pi/\sqrt{2-\eta}. Thus, consider first an attractive potential of the form V(r)=−K/rαV(r) = -K/r^\alpha with K>0K > 0 and α>0\alpha> 0. If Δθ\Delta\theta is to be an intrinsic property of this potential, independent of the parameters of orbits, we can compute Eq. (6) in the limit E→−0E \to-0, see Refs. [9], [11]. The result is, see Ref. [12],

Δθ=∫01dxxα−x2=π2−α(7)\Delta\theta= \int_0^1 \frac{dx}{\sqrt{x^\alpha- x^2}} = \frac{\pi}{2-\alpha} \tag*{(7)}

and is valid only if α<2\alpha< 2. We thus get 1≤c=2−α<21 \le c = 2-\alpha< 2. Then c=1=αc = 1 = \alpha and Δθ=π\Delta\theta= \pi. In the case of potentials such as V(r)=K′rζ/2V(r) = K'r^\zeta/2 with K′>0K' > 0, ζ>0\zeta> 0, the computation of Eq. (6) in the limit E→+∞E \to+\infty leads to

Δθ=∫01dx1−x2=π2(8)\Delta\theta= \int_0^1 \frac{dx}{\sqrt{1-x^2}} = \frac{\pi}{2} \tag*{(8)}

which now gives c=2=ζc = 2 = \zeta. Hence the conclusion of Bertrand’s theorem. It is worth noticing here that the values Eq. (7) and (8) are obtained exactly from (6) without any approximation, for the respective potentials −K/r-K/r (α=1\alpha= 1) and K′r2/2K'r^2/2 (ζ=2\zeta= 2), while both values of Δθ\Delta\theta are only reached in the limits defined above for other corresponding potentials. This reinforces the above conclusion.

Finally, let us mention that a good idea of the shape of trajectories can also be obtained using the well known formula

κ=−Lm2rv3F(r),with F(r)=−dVdr(r)\kappa= -\frac{L}{m^{2}rv^{3}}\mathcal{F}(r), \quad\text{with } \mathcal{F}(r) = -\frac{dV}{dr}(r)

giving the curvature κ\kappa of a trajectory: depending on whether the force is attractive or repulsive everywhere, the curvature is always positive or negative, respectively.

How to characterize a hidden symmetry ?

A dynamical or hidden symmetry must be understood as a continuous group of transformations linking solutions of a set as wide as possible. As shown in Ref. [5], such a symmetry, if it exists, can be characterized in a minimal way by a unitary vector UU, called there the eccentricity vector, but which we will now call more appropriately a symmetry vector. Note that the authors of Refs. [6, 7, 8] used the same formalism as our. Their symmetry vector is currently called the Fradkin-Bacry-Ruegg-Souriau’s perihelion vector. A symmetry vector is a first integral that does not depend explicitly on time and can be used to construct with HH and LL a complete set of independent first integrals of the problem. The role of a symmetry is to provide continuous transformations of a trajectory into another one of the same species by means of Poisson brackets, while defining in some way a common geometrical property of those trajectories, and the vector UU has to be chosen in this perspective. The motion being planar, that vector is more likely in the same plane, i.e., orthogonal to LL. With this requirement, we have shown in Ref. [5] that the components of UU have the simple Poisson brackets, see also Ref. [8]

{H,Ui}=∂H∂xr∂Ui∂pr−∂H∂pr∂Ui∂xr=0,{Li,Uj}=ϵijkUk,{Ui,Uj}=0(9)\{H,U_i\} = \frac{\partial H}{\partial x_r}\frac{\partial U_i}{\partial p_r} - \frac{\partial H}{\partial p_r}\frac{\partial U_i}{\partial x_r} = 0,\quad\{L_i,U_j\} = \epsilon_{ijk}U_k,\quad\{U_i,U_j\} = 0 \tag*{(9)}

the indices i,j,ki,j,k taking each the values 1, 2, 3 standing for xx, yy, zz, respectively; ϵijk\epsilon_{ijk} is the complete skew-symmetric Levi-Civita tensor for three dimensions, with ϵ123=1\epsilon_{123}=1. Choosing a symmetry vector as unitary has the advantage to account for any kind of dynamical symmetry, and so could be considered as the true indicator of a symmetry of trajectories. As an operator in phase space UU is to be considered as a field vector in that space, and treated in exactly the same way as the angular momentum LL. On the other hand, at a given point in phase space, the geometrical role of UU is to define a symmetry axis of the trajectory corresponding to that point.

Next, let us define W=1LL×UW = \frac{1}{L}L \times U which is a vector forming with UU and ℓ=L/L\ell= L/L an orthonormal basis. It is well known that the components of the vectors of such a basis are linked by the formulas

Wi=ϵijkljUk,Ui=ϵijkWjℓk,ℓi=ϵijkUjWk,ϵijkUk=Wiℓj−Wjℓi,ϵijkWk=ℓiUj−ℓjUi,ϵijkℓk=UiWj−UjWi,UiUj+WiWj+ℓiℓj=δij(10)W_i=\epsilon_{ijkl}jU_k,\quad U_i=\epsilon_{ijk}W_j\ell_k,\quad\ell_i=\epsilon_{ijk}U_jW_k,\\ \epsilon_{ijk}U_k=W_i\ell_j-W_j\ell_i,\quad\epsilon_{ijk}W_k=\ell_iU_j-\ell_jU_i,\quad\epsilon_{ijk}\ell_k=U_iW_j-U_jW_i,\\ U_iU_j+W_iW_j+\ell_i\ell_j=\delta_{ij} \tag*{(10)}

The vector W\mathbf{W} being also orthogonal to L\mathbf{L}, its components satisfy

{Li,Wj}=ϵijkWk=ℓiUj−ℓjUi,{Wi,Wj}=0\{L_i,W_j\}=\epsilon_{ijk}W_k=\ell_iU_j-\ell_jU_i,\quad\{W_i,W_j\}=0

To express the transformation of LL under the action of U\mathbf{U}, we need to compute some Poisson brackets. From Eq. (9), we get

{L,Uj}=12L{L2,Uj}=12L{LiLi,Uj}=LiL{Li,Uj}=ϵijkℓiUk=−Wj(11)\{L,U_j\}=\frac{1}{2L}\{L^2,U_j\}=\frac{1}{2L}\{L_iL_i,U_j\}=\frac{L_i}{L}\{L_i,U_j\}=\epsilon_{ijk}\ell_iU_k=-W_j \tag*{(11)}

In the same way, we obtain

{L,Wj}=LiL{Li,Wj}=ϵijkℓiWk=ϵjkiWkℓi=Uj\{L,W_j\}=\frac{L_i}{L}\{L_i,W_j\}=\epsilon_{ijk}\ell_iW_k=\epsilon_{jki}W_k\ell_i=U_j

Using Eqs. (11) and (10), we have

{ℓi,Uj}={Li,Uj}/L−Li{L,Uj}/L2=ϵijkUk/L+LiWj/L2=WiLj/L2\{\ell_i,U_j\}=\{L_i,U_j\}/L-L_i\{L,U_j\}/L^2=\epsilon_{ijk}U_k/L+L_iW_j/L^2=W_iL_j/L^2

whence

{Ui,Wj}=ϵjmn{Ui,ℓmUn}=ϵjmnUn{Ui,ℓm}=−ϵjmnUnWmLi/L2=LiLj/L3\{U_i,W_j\}=\epsilon_{jmn}\{U_i,\ell_mU_n\}=\epsilon_{jmn}U_n\{U_i,\ell_m\}\\ =-\epsilon_{jmn}U_nW_mL_i/L^2=L_iL_j/L^3

Under an infinitesimal transformation generated by a symmetry vector A(r,p)\mathbf{A}(\mathbf{r},\mathbf{p}), the infinitesimal variation of LL is given by

δL={αiAi,L}(12)\delta L=\{\alpha_iA_i,L\} \tag*{(12)}

where αi\alpha_i are the components of an infinitesimal constant vector α\boldsymbol{\alpha} (i.e., independent of the canonical variables). From the dimensional analysis of Eq. (12), αiAi\alpha_iA_i must be homogeneous to an angular momentum. If A=U\mathbf{A}=\mathbf{U}, then the αi\alpha_i are to be homogeneous to LL. If the αi\alpha_i are chosen dimensionless, A\mathbf{A} must be homogeneous to LL and for this reason cannot be a unitary vector. At this stage, LL is the only first integral that is homogeneous to an angular momentum, and we are naturally led to define a new symmetry vector by A=LW\mathbf{A}=L\mathbf{W}. With this choice, we obtain the new set of relations

{Li,Aj}=ϵilkAk,{Ai,Aj}=−ϵijkLk,{Li,Lj}=ϵijkLk(13)\{L_i,A_j\}=\epsilon_{ilk}A_k,\quad\{A_i,A_j\}=-\epsilon_{ijk}L_k,\quad\{L_i,L_j\}=\epsilon_{ijk}L_k \tag*{(13)}

which is homomorphic to that defining the Lie algebras of the groups SO(3,1)SO(3,1), or the Lorentz group L(3,1)L(3,1) or the group SL(2,C)SL(2,\mathbb{C}): see, e.g., Ref. [13]. Indeed, it appears that the latter choice yields a simpler law of transformation for LL, as seen below. Defining B=LW\mathbf{B}=L\mathbf{W}, we have

{Li,Bj}=ϵijkBk,{Bi,Bj}=−ϵijkLk,{Ai,Bj}=Lδij(14)\{L_i,B_j\}=\epsilon_{ijk}B_k,\quad\{B_i,B_j\}=-\epsilon_{ijk}L_k,\quad\{A_i,B_j\}=L\delta_{ij} \tag*{(14)}

With finite αi\alpha_i, the transformation TA\mathcal{T}_A generated by TA=α⋅AT_A=\boldsymbol{\alpha}\cdot\mathbf{A} changes LL into

L′=TA(L)=L+{TA,L}+12{TA,{TA,L}}+13!{TA,{TA,{TA,L}}}+⋯L'=\mathcal{T}_A(L)=L+\{T_A,L\}+\frac{1}{2}\{T_A,\{T_A,L\}\}+\frac{1}{3!}\{T_A,\{T_A,\{T_A,L\}\}\}+\cdots

Let us define ξ=∥α∥\xi=\lVert\boldsymbol{\alpha}\rVert, n=α/ξ\mathbf{n}=\boldsymbol{\alpha}/\xi. Then,

TA(L)={TA,L}=αi{Ai,L}=α⋅B=ξLn⋅W,T_A(L)=\{T_A,L\}=\alpha_i\{A_i,L\}=\boldsymbol{\alpha}\cdot\mathbf{B}=\xi L\mathbf{n}\cdot\mathbf{W},
TA2(L)={TA,{TA,L}}=αiαj{Ai,Bj}=ξ2L,T_A^2(L)=\{T_A,\{T_A,L\}\}=\alpha_i\alpha_j\{A_i,B_j\}=\xi^2L,
TA3(L)={TA,{TA,{TA,L}}}=ξ2{TA,L}=Lξ3n⋅WT_A^3(L)=\{T_A,\{T_A,\{T_A,L\}\}\}=\xi^2\{T_A,L\}=L\xi^3\mathbf{n}\cdot\mathbf{W}

and so on. Hence, we finally obtain the law of transformation for LL

L′=L[cosh⁡ξ+n⋅Wsinh⁡ξ](15)L'=L\left[\cosh\xi+\mathbf{n}\cdot\mathbf{W}\sinh\xi\right] \tag*{(15)}

which is a simple scaling. Let us remark that, generally speaking, the only invariants under the transformations generated by the Lie algebra Eq. (13) are the two Casimir operators C1=L2−A2C_1=L^2-A^2 (A2=AiAiA^2=A_iA_i) and C2=L⋅AC_2=\mathbf{L}\cdot\mathbf{A}, which are both zero here. Thus, L2L^2 and A2A^2 are not invariants, only their difference is. This explains the result L′≠LL'\ne L in Eq. (15).

Note also that any operator like TAT_A acts as a differentiation (it verifies Leibniz rule). Consequently, F1F_1, F2F_2, F3∘F1F_3\circ F_1 being any regular functions of canonical coordinates its exponential TAT_A has the two important properties:

TA(F1F2)=TA(F1)TA(F2),andTA(F3(F1))=F3(TA(F1))T_A(F_1F_2)=T_A(F_1)T_A(F_2),\quad\text{and}\quad T_A(F_3(F_1))=F_3(T_A(F_1))

which in particular ensure the covariance under TAT_A of the common polar equation of trajectories. For example, in the Kepler problem, they allow us to write that the eccentricity of an ellipse with parameters EE and LL is transformed into the eccentricity of another ellipse with parameters EE and L′=TA(L)L'=T_A(L).

Let us also give the transformation of the components of LL and AA. We have

TA(Li)=αj{Aj,Li}=ξnjϵjikAk=ξ[Li n⋅W−Wi n⋅L],whence{TA,n⋅L}=0TA2(Li)=ξ2nℓnjϵjik{Aℓ,Ak}=ξ2njnℓϵjikϵℓmkLm=ξ2[Li−ni n⋅L]\begin{aligned} T_A(L_i) &= \alpha_j\{A_j,L_i\}=\xi n_j\epsilon_{jik}A_k=\xi\left[L_i\,\mathbf{n}\cdot\mathbf{W}-W_i\,\mathbf{n}\cdot\mathbf{L}\right], \quad\text{whence}\\ \{T_A,\mathbf{n}\cdot\mathbf{L}\} &= 0\\ T_A^2(L_i) &= \xi^2 n_\ell n_j\epsilon_{jik}\{A_\ell,A_k\}=\xi^2 n_j n_\ell\epsilon_{jik}\epsilon_{\ell mk}L_m=\xi^2\left[L_i-n_i\,\mathbf{n}\cdot\mathbf{L}\right] \end{aligned}

where the formula ϵjikϵℓmk=δjℓδim−δjmδiℓ\epsilon_{jik}\epsilon_{\ell mk}=\delta_{j\ell}\delta_{im}-\delta_{jm}\delta_{i\ell} has been used in the last equation. Then,

TA3(Li)=ξ2TA(Li),TA2n+1(Li)=ξ2nTA(Li)with n≥1,TA4(Li)=ξ2TA2(Li),TA2n(Li)=ξ2n−2TA2(Li)with n≥2,\begin{aligned} T_A^3(L_i)&=\xi^2T_A(L_i), \quad T_A^{2n+1}(L_i)=\xi^{2n}T_A(L_i) \quad\text{with } n\geq1,\\ T_A^4(L_i)&=\xi^2T_A^2(L_i), \quad T_A^{2n}(L_i)=\xi^{2n-2}T_A^2(L_i) \quad\text{with } n\geq2, \end{aligned}

and finally,

Li′=TA(Li)=Li+(cosh⁡ξ−1)[Li−ni n⋅L]+sinh⁡ξ[Li n⋅W−Wi n⋅L]=Li[cosh⁡ξ+n⋅Wsinh⁡ξ]−n⋅L[ni(cosh⁡ξ−1)+Wisinh⁡ξ](16)\begin{aligned} L'_i=T_A(L_i)&=L_i+(\cosh\xi-1)\left[L_i-n_i\,\mathbf{n}\cdot\mathbf{L}\right]+\sinh\xi\left[L_i\,\mathbf{n}\cdot\mathbf{W}-W_i\,\mathbf{n}\cdot\mathbf{L}\right]\\ &=L_i\left[\cosh\xi+\mathbf{n}\cdot\mathbf{W}\sinh\xi\right]-\mathbf{n}\cdot\mathbf{L}\left[n_i(\cosh\xi-1)+W_i\sinh\xi\right] \tag*{(16)} \end{aligned}

The transformation of the components AiA_i is obtained from Eq. (16), simply replacing LiL_i and L\mathbf{L} with AiA_i and A\mathbf{A}. Then, using TA(Ai)=TA(LUi)=TA(L)TA(Ui)T_A(A_i)=T_A(LU_i)=T_A(L)T_A(U_i), we find

Ai′=Ai[cosh⁡ξ+n⋅Wsinh⁡ξ]−n⋅A[ni(cosh⁡ξ−1)+Wisinh⁡ξ]andUi′=Ui−n⋅Uni(cosh⁡ξ−1)+Wisinh⁡ξcosh⁡ξ+n⋅Wsinh⁡ξ(17)\begin{aligned} A'_i&=A_i\left[\cosh\xi+\mathbf{n}\cdot\mathbf{W}\sinh\xi\right]-\mathbf{n}\cdot\mathbf{A}\left[n_i(\cosh\xi-1)+W_i\sinh\xi\right]\quad\text{and}\\ U'_i&=U_i-\mathbf{n}\cdot\mathbf{U}\frac{n_i(\cosh\xi-1)+W_i\sinh\xi}{\cosh\xi+\mathbf{n}\cdot\mathbf{W}\sinh\xi} \tag*{(17)} \end{aligned}

In the same way, it is easy to obtain

Bi′=Bi+ni(cosh⁡ξ−1) n⋅B+niLsinh⁡ξandWi′=Wi+ni(cosh⁡ξ−1) n⋅W+nisinh⁡ξcosh⁡ξ+n⋅Wsinh⁡ξ(18)\begin{aligned} B'_i&=B_i+n_i(\cosh\xi-1)\,\mathbf{n}\cdot\mathbf{B}+n_iL\sinh\xi\quad\text{and}\\ W'_i&=\frac{W_i+n_i(\cosh\xi-1)\,\mathbf{n}\cdot\mathbf{W}+n_i\sinh\xi}{\cosh\xi+\mathbf{n}\cdot\mathbf{W}\sinh\xi} \tag*{(18)} \end{aligned}

taking into account the relations {Ai,Bj}=Lδij\{A_i,B_j\}=L\delta_{ij} (Eq. (14)) and TA(Bj)=L′TA(Wj)T_A(B_j)=L'T_A(W_j).

As shown in Appendix § 7, Eq. (18) can be interpreted as a Lorentz boost. However, the orthonormality of the reference basis U,W,ℓ\mathbf{U},\mathbf{W},\boldsymbol{\ell} is preserved under TAT_A, Eqs. (16), (17) and (18) defining the new reference basis U′,W′,ℓ\mathbf{U}^{\prime}, \mathbf{W}^{\prime}, \ell. Hence, TAT_A generally changes the orientation of the plane of motion, a change that should be described by a rotation whose axis and angle are to be expressed in terms of n\mathbf{n} and ξ\xi. This is also done in Appendix B. In addition, it is seen from Eqs. (15), (16) and (17) that TAT_A also causes a change of scale along the reference axes defined by A\mathbf{A}, B\mathbf{B} and L\mathbf{L}. As they have the same norm LL, the latter are multiplied by the same factor ρ=L′/L=cosh⁡ξ+n⋅Wsinh⁡ξ\rho= L^{\prime}/L = \cosh\xi+ \mathbf{n} \cdot\mathbf{W} \sinh\xi. Thus, the action of TAT_A on the said three vectors is equivalent to the product of a rotation RAR_A by ρ\rho. If n\mathbf{n} is along the initial W\mathbf{W}, i.e., if n=±W\mathbf{n} = \pm\mathbf{W}, the reference basis remains unchanged while the scale factor is exp⁡(±ξ)\exp(\pm\xi).

Unfortunately, we cannot derive a general form of the transformation of the coordinates XX and YY as such a transformation is specific to the dynamics of the motion. Actually, the previous developments are all we can do independently of the particular (spherical symmetric) potential under study.

As emphasized in Ref. [1], the transformations of dynamical symmetries alone are unable to link trajectories corresponding to different values of energy, because their generators have zero Poisson brackets with the Hamiltonian. As shown in the same reference, this lack is overcome for potentials in single power law, thanks to their additional symmetry called “mechanical similarity”. This symmetry allows us to complete the group of transformations that is necessary to obtain a full connection between trajectories of a given set by varying both variables EE and LL. Finding an equivalent symmetry for other kinds of spherically symmetric potentials is an open question.

The choice of reference axes

Thanks to rotational invariance, the reference axes in the plane of motion can be freely chosen. However, on one hand, we all know that whatever the problem, the best way to present its solution in the most intelligible and transparent form is to take into account from the outset some of their common features, if any. On the other hand, we are convinced that a dynamical symmetry, as currently expressed mathematically, is a direct consequence of the existence of common geometrical characteristics of a set of trajectories. Hence, we are immediately led to look for such characteristics and, of course, accounting for the previous section, a possible axis of symmetry that could serve as a reference. Before that, it is necessary to point out some subtleties.

First, we must mention a fact that seems obvious but is essential in our study : whatever the shape of the potential V(r)V(r) and the range of variations of rr, it is always possible to find a value of EE satisfying the inequality E≥U(r)E \ge U(r) (defining the allowed motions), with the consequence that the equation E=U(r)E = U(r) of turning points has at least one solution for rr which, unfortunately, is hard or impossible to obtain in closed form in the general case. Using elementary arguments, the existence of turning points can be proved for the attractive power-law potentials −K/rα-K/r^{\alpha} and Krα/2Kr^{\alpha/2} (K>0K > 0, α>0\alpha> 0), and for power-law potentials that are everywhere repulsive. As explained in Appendix § 6, every turning point is associated with an axis of symmetry, thanks to spherical symmetry. Thus, the search for symmetry axes requires that of turning points. This can be done through the polar equation r(θ)r(\theta) one derives from Eqs. (2) or (5), the parameters of which are determined by the values of EE and LL. Giving to θ\theta all permissible values, this polar equation represents the complete geometrical curve along which take place all motions having initial conditions that are different but give the same values for EE and LL. This curve will be also called a trajectory hereafter. The said motions can be linked by varying continuously their initial conditions, hence the said polar equation already covers an infinity of possible motions. Due to this diversity of initial conditions, it may happen that some of these motions do not have a turning point, and no symmetry axis, whereas, according to the above remark, such a point does exist and may be present on the complete geometrical curve (it corresponds to dr/dθ=0dr/d\theta= 0). It may also happen that, depending on the values of EE and LL, some trajectories do not have a symmetry axis. The following positive arguments can be set against this last negative remark. As we know, the existence of an axis of symmetry for a class of trajectories implies that of additional first integrals. But the latter, when expressed in terms of canonical variables, owe their status solely to the fact that their Poisson brackets with the Hamiltonian are zero : they are not attached to the class of trajectories from which they have been highlighted, they apply to all possible trajectories and motions. In addition, for a given potential, all possible trajectories are deduced from the same fundamental equation and must have similar or analytically related polar equations, and these necessarily bear a trace of the existence of a turning point.

Then, consider a trajectory (complete geometrical curve) having a turning point P0P_0, where the particle has the vector position r0=r0er0\mathbf{r}_0 = r_0\mathbf{e}_{r0} and momentum p0=p0eθ0\mathbf{p}_0 = p_0\mathbf{e}_{\theta0} such that r0⋅p0=0\mathbf{r}_0 \cdot\mathbf{p}_0 = 0. Its axis of symmetry is along r0\mathbf{r}_0 and passes through OO. In the 2-plane of motion, we redefine the unit vectors of rectangular axes as Ex=er0\mathbf{E}_x = \mathbf{e}_{r0} and Ey=eθ0\mathbf{E}_y = \mathbf{e}_{\theta0}. Note that the angular momentum is completely defined by r0\mathbf{r}_0 and p0\mathbf{p}_0, as L=r0p0L = r_0p_0. Of course, the two unit vectors Ex\mathbf{E}_x and Ey\mathbf{E}_y are constant in time and, as such, are good candidates to represent a (non-zero) symmetry vector. Let us choose U=Ex\mathbf{U} = \mathbf{E}_x. To make this vector a field vector in phase space, we have to express it in terms of the canonical variables r\mathbf{r} and p\mathbf{p} of a point of the trajectory, in the same way the angular momentum is written as L=r×p\mathbf{L} = \mathbf{r} \times\mathbf{p}. This is easily done by simply inverting the trivial expressions

r=XEx+YEy,p=PxEx+PyEy\mathbf{r} = X\mathbf{E}_x + Y\mathbf{E}_y,\qquad\mathbf{p} = P_x\mathbf{E}_x + P_y\mathbf{E}_y

Using

L=XPy−YPx,Py=prsin⁡ψ+pψcos⁡ψ=ursin⁡ψ+Lrcos⁡ψL = X P_y - Y P_x,\qquad P_y = p_r \sin\psi+ p_\psi\cos\psi= \frac{u}{r}\sin\psi+ \frac{L}{r}\cos\psi

where ψ\psi is the polar angle of r\mathbf{r} relatively to r0\mathbf{r}_0, as defined in Eq. (A.1), we get

Ex=1L[Pyr−Yp],\mathbf{E}_x = \frac{1}{L}\left[P_y\mathbf{r}-Y\mathbf{p}\right],

or

Ex=1Lr[(usin⁡ψ+Lcos⁡ψ)r−r2sin⁡ψp](19)\mathbf{E}_x = \frac{1}{Lr}\left[(u\sin\psi+L\cos\psi)\mathbf{r}-r^2\sin\psi\mathbf{p}\right] \tag*{(19)}

The second member of Eq. (19) is a function of canonical variables, ψ\psi being expressed as in Eq. (A.1) with L=p2r2−u2L=\sqrt{p^2r^2-u^2} and E=p2/(2m)+V(r)E=p^2/(2m)+V(r). It is obviously a first integral as Ex\mathbf{E}_x is a constant of motion. Note that XX, YY, PxP_x, PyP_y and ψ\psi are to be considered now as invariant under rotations, expressed in terms of the scalars r2r^2, p2p^2 and u=p⋅ru=\mathbf{p}\cdot\mathbf{r}. Using

p=1r2[ur−r×L]\mathbf{p}=\frac{1}{r^2}\left[u\mathbf{r}-\mathbf{r}\times\mathbf{L}\right]

and the polar equation r=R(ψ)r=R(\psi), Eq. (19) can be recast in the useful form

Ex=1LR(ψ)[sin⁡ψ r×L+Lcos⁡ψ r](20)\mathbf{E}_x=\frac{1}{LR(\psi)}\left[\sin\psi\,\mathbf{r}\times\mathbf{L}+L\cos\psi\,\mathbf{r}\right] \tag*{(20)}

Instead of obtaining a trivial relation when projecting r\mathbf{r} onto Ex\mathbf{E}_x, this trick allows us to “retrieve” the polar equation:

rcos⁡ψ=r⋅Ex=r2cos⁡ψR(ψ),hencer=R(ψ)r\cos\psi=\mathbf{r}\cdot\mathbf{E}_x=\frac{r^2\cos\psi}{R(\psi)},\qquad\text{hence}\quad r=R(\psi)

At this point, it is worth remarking that the expressions in Eqs. (19) and (20) are covariant under the symmetry group, a property that could not have been obtained using as a reference any other constant vector independent of the dynamics.

Defining an axis of symmetry, Ex\mathbf{E}_x must play a role similar to that of the LRL vector in the Kepler problem (V(r)=−K/rV(r)=-K/r). Let us check Eq. (19) by considering a few examples. The first one is just that of the Kepler problem. In the case of ellipses, we have

R(ψ)=r01+ecos⁡ψ,withr0=L2mK,e=1−2r0∣E∣K,u=eLrsin⁡ψr0R(\psi)=\frac{r_0}{1+e\cos\psi},\qquad\text{with}\quad r_0=\frac{L^2}{mK},\quad e=\sqrt{1-\frac{2r_0|E|}{K}},\quad u=\frac{eLr\sin\psi}{r_0}

and we find

KeEx=−ump+(p2m−Kr)r(21)K_e\mathbf{E}_x=-\frac{u}{m}\mathbf{p}+\left(\frac{p^2}{m}-\frac{K}{r}\right)\mathbf{r} \tag*{(21)}

which is exactly the LRL vector. The same result is obtained considering instead a hyperbola.

The second example is that of the Hookean (or harmonic) potential V(r)=Kr2/2V(r)=Kr^2/2 with K>0K>0. The more general trajectory is an ellipse centered in OO, whose cartesian equation is X2/a2+Y2/b2=1X^2/a^2+Y^2/b^2=1. It appears more convenient to the parametrisation with the eccentric anomaly χ\chi: X=acos⁡χX=a\cos\chi, Y=bsin⁡χY=b\sin\chi, which leads to

χ˙=Km,L=abKm,Py=bKmcos⁡χ,\dot{\chi}=\sqrt{\frac{K}{m}},\quad L=ab\sqrt{Km},\quad P_y=b\sqrt{Km}\cos\chi,

and

Ex=1aKm[Kmcos⁡χr−sin⁡χp]\mathbf{E}_x=\frac{1}{a\sqrt{Km}}\left[\sqrt{Km}\cos\chi\mathbf{r}-\sin\chi\mathbf{p}\right]

Using the notations of Ref. [5], we find

cos⁡θ=p2−Kmr2(p2−Kmr2)2+4Kmu2=cos⁡2χ,\cos\theta=\frac{p^2-Kmr^2}{\sqrt{(p^2-Kmr^2)^2+4Kmu^2}}=\cos2\chi,
sin⁡θ=2uKm(p2−Kmr2)2+4Kmu2=sin⁡2χ\sin\theta=\frac{2u\sqrt{Km}}{\sqrt{(p^2-Kmr^2)^2+4Kmu^2}}=\sin2\chi

Then, by setting λ=Kma2\lambda=Kma^2, χ=θ/2\chi=\theta/2, we retrieve the opposite of one of the two “eccentricity vectors” proposed in Ref. [5] for this case.

These two examples demonstrate the efficiency of this simple method in finding the expression of symmetry vectors in terms of canonical variables, once the polar equation of trajectories is known.

Let us now consider the instructive example of the trajectories provided by the potential V(r)=−K/r2V(r)=-K/r^2. Assuming L2≠2mKL^2\ne2mK, their polar equations can be derived from the general expression

R(θ)=Rcos⁡χ(θ−ϕ)withχ2=1−2mKL2,R2=Lχ22mER(\theta)=\frac{R}{\cos\chi(\theta-\phi)}\quad\text{with}\quad\chi^2=1-\frac{2mK}{L^2},\quad R^2=\frac{L\chi^2}{2mE}

θ\theta being the polar angle of rr relatively to its initial value r(0)r(0) and ϕ\phi being constant. Note that χ\chi, RR and ϕ\phi are real or complex numbers according to the signs of χ2\chi^2 and EE, subject to the condition R(θ)>0R(\theta)>0. There are three main cases:

  1. χ2>0\chi^2>0, E>0E>0; then

R(θ)=r0cos⁡ω(θ−ϕ)withω=1−2mKL2,r0=Lω2mE,ϕ realR(\theta)=\frac{r_0}{\cos\omega(\theta-\phi)} \quad\text{with}\quad \omega=\sqrt{1-\frac{2mK}{L^2}},\quad r_0=\frac{L\omega}{\sqrt{2mE}},\quad\phi\ \text{real}
  1. χ2<0\chi^2<0, E<0E<0; then

R(θ)=r0cosh⁡Ω(θ−ϕ)withΩ=2mKL2−1,r0=LΩ2m∣E∣,ϕ realR(\theta)=\frac{r_0}{\cosh\Omega(\theta-\phi)} \quad\text{with}\quad \Omega=\sqrt{\frac{2mK}{L^2}-1},\quad r_0=\frac{L\Omega}{\sqrt{2m\lvert E\rvert}},\quad\phi\ \text{real}
  1. χ2<0\chi^2<0, E>0E>0. Then χ=±iΩ\chi=\pm i\Omega, R=±ir0R=\pm i r_0 with r0=LΩ2mEr_0=\frac{L\Omega}{\sqrt{2mE}}, and we are led to set Ωϕ=±Ωϕ′−iπ/2\Omega\phi=\pm\Omega\phi'-i\pi/2 with ϕ′\phi' real and positive, to take into account all possible initial conditions. This gives

R(θ)=±r0sinh⁡Ω(θ±ϕ′)(22)R(\theta)=\frac{\pm r_0}{\sinh\Omega(\theta\pm\phi')} \tag*{(22)}

Case 1) is mainly that of unbounded trajectories. The complete geometrical curve is obtained by varying ψ=θ−ϕ\psi=\theta-\phi between −π/(2ω)-\pi/(2\omega) and +π/(2ω)+\pi/(2\omega). It has a single turning point at ψ=0\psi=0, that is reached or not during a real motion according to whether the initial angle θ(0)\theta(0) is less or greater than ϕ\phi (remember that θ˙>0\dot{\theta}>0). Independently of this possibility, the polar equation is symmetrical relatively to the value θ=ϕ\theta=\phi, corresponding to an axis of symmetry which is opportunately taken as axis OxOx. In this case, we find the symmetry vector

Ex=cos⁡ωψLr0[sin⁡ψ r×L+Lcos⁡ψ r],withψ=±1ωcos⁡−1(r0r)E_x=\frac{\cos\omega\psi}{Lr_0}\left[\sin\psi\,\mathbf{r}\times L+L\cos\psi\,\mathbf{r}\right], \quad\text{with}\quad \psi=\pm\frac{1}{\omega}\cos^{-1}\left(\frac{r_0}{r}\right)

Case 2) mainly corresponds to bounded trajectories. In a real motion, the particle always falls towards the source in a spiral movement. The complete geometrical curve is obtained by varying ψ=θ−ϕ\psi=\theta-\phi between −∞-\infty and +∞+\infty. It has also a single turning point at ψ=0\psi=0. Even for this special geometry, we have a symmetry vector:

Ex=cosh⁡ΩψLr0[sin⁡ψ r×L+Lcos⁡ψ r]withψ=±1Ωcosh⁡−1(r0r)E_x=\frac{\cosh\Omega\psi}{Lr_0}\left[\sin\psi\,\mathbf{r}\times L+L\cos\psi\,\mathbf{r}\right] \quad\text{with}\quad \psi=\pm\frac{1}{\Omega}\cosh^{-1}\left(\frac{r_0}{r}\right)

Case 3) gathers two subcases. The first one with the plus sign in Eq. (22) corresponds to bounded trajectories falling spirally towards the source as θ\theta varies from 00 to +∞+\infty, according to the polar equation r=r0/sinh⁡Ω(θ+ϕ′)r=r_0/\sinh\Omega(\theta+\phi'). The second one with the minus sign in Eq. (22) is that of unbounded trajectories, rr going from r(0)r(0) to +∞+\infty as θ\theta varies from 0 to ϕ′\phi', according to the polar equation r=r0/sinh⁡Ω(ϕ′−θ)r = r_0/\sinh\Omega(\phi' - \theta). None of these trajectories have a turning point, nor their complete geometrical curves for which ψ=ϕ′+θ\psi= \phi' + \theta for the first one varies between 0 and +∞+\infty and ψ=θ−ϕ′\psi= \theta- \phi' for the second one varies between −∞-\infty and 0. In each case, the value ψ=0\psi= 0 refers to an asymptote which could be considered as a reminiscence of the axis of symmetry observed for case 2) when transforming a cosh into a sinh in the polar equation, as seen above. These asymptotes are now taken as reference axes and we have

Ex=1Lr[sin⁡ψ r×L+Lcos⁡ψ r]withψ=±1Ωsinh⁡−1(r0r)E_x = \frac{1}{Lr}\left[\sin\psi\,\mathbf{r}\times\mathbf{L}+L\cos\psi\,\mathbf{r}\right]\quad\text{with}\quad\psi= \pm\frac{1}{\Omega}\sinh^{-1}\left(\frac{r_0}{r}\right)

This example shows that, as a function of canonical variables, a symmetry vector may have different analytical forms according to the trajectories to which it refers. This fact could explain those singularities of the symmetry vector which have been predicted by various authors, see e.g. Ref. [14, 15].

Another instructive example is provided by the potential

V(r)=−K1r+K2r2,K1>0,K2>0,V(r)=-\frac{K_1}{r}+\frac{K_2}{r^2},\quad K_1>0,\quad K_2>0,

which has been studied in a particular context in Ref. [16]. The corresponding effective potential

U(r)=L′22mr2−K1r,whereL′2=L2+2mK2U(r)=\frac{L'^2}{2mr^2}-\frac{K_1}{r},\quad\text{where}\quad L'^2=L^2+2mK_2

differs from that of the Kepler problem only by the replacement of LL by L′L'. However, we still have ψ˙=L/(mr2)\dot{\psi}=L/(mr^2). Consequently, the polar equation of trajectories is found to be

R(ψ)=r01+ecos⁡(βψ)(23)R(\psi)=\frac{r_0}{1+e\cos(\beta\psi)} \tag*{(23)}

with β=L′/L>1\beta=L'/L>1, r0=L′2/(mK1)r_0=L'^2/(mK_1), e=1+2EL′2/(mK12)e=\sqrt{1+2EL'^2/(mK_1^2)}.

Bounded trajectories exist for E<0E<0. They are closed only for discrete values of LL for which β\beta is a rational number larger than 1. Otherwise, they are not closed and, as time passes, the associated motion fills the entire circular crown r0/(1+e)≤r≤r0/(1−e)r_0/(1+e)\leq r\leq r_0/(1-e). Depending on the case, we encounter a finite or infinite number of turning points on one tour. Obviously, we can choose any of those on the circle r=r0/(1+e)r=r_0/(1+e) (pericenters) to define the reference axis for which we take ψ=0\psi=0 and obtain the polar equation in the form (4.5) with −∞<ψ<+∞-\infty<\psi<+\infty. For E>0E>0, we have unbounded trajectories. The only property shared between all these curves is the symmetry of their polar equation, when ψ\psi is replaced by −ψ-\psi, which means that the axis OxOx is their common axis of symmetry. Using Eq. (20), we find for this problem the symmetry vector

Ex=1+ecos⁡βψLr0[sin⁡ψ r×L+Lcos⁡ψ r]\mathbf{E}_x = \frac{1+e\cos\beta\psi}{Lr_0}\left[\sin\psi\,\mathbf{r}\times\mathbf{L}+L\cos\psi\,\mathbf{r}\right]

Unbounded trajectories have the same geometry, and the dynamical symmetry operations can transform continuously any of them into another having the same value of energy.

Let us now consider central potentials which are everywhere repulsive. All corresponding trajectories are unbounded but bear each a sole turning point associated with an axis of symmetry. Even this case has a symmetry group as illustrated by the following two examples.

The repulsive newtonian potential V(r)=K/rV(r)=K/r with K>0K>0 yields hyperbolas of polar equation

r=R(ψ)=r0ecos⁡ψ−1r=R(\psi)=\frac{r_0}{e\cos\psi-1}

with r0=L2/(mK)r_0=L^2/(mK), e=1+2EL2/(mK)e=\sqrt{1+2EL^2/(mK)}. This case has the LRL-like vector

KeEx=−ump+(p2m+Kr)rKe\mathbf{E}_x=-\frac{u}{m}\mathbf{p}+\left(\frac{p^2}{m}+\frac{K}{r}\right)\mathbf{r}

obtained from Eq. (21) by simply replacing KK with −K-K.

The “repulsive harmonic” potential V(r)=−Kr2/2V(r)=-Kr^2/2 with K>0K>0 also yields hyperbolas of polar equation

R(ψ)=r01+ecos⁡2ψR(\psi)=\frac{r_0}{\sqrt{1+e\cos2\psi}}

with r0=L/mEr_0=L/\sqrt{mE}, e=1+L2K/(mE2)e=\sqrt{1+L^2K/(mE^2)}. In the (attractive) Hookean case, the symmetry vector is an eigenvector of the constant symmetric two-rank tensor Tij=mKxixj+pipjT_{ij}=mKx_ix_j+p_ip_j. Here we also find a constant symmetric two-rank tensor: Tij=−mKxixj+pipjT_{ij}=-mKx_ix_j+p_ip_j, one of its two eigenvectors being along the axis of symmetry of hyperbolas. This vector is used to define a symmetry vector, analogous to the eccentricity vector given in Ref. [5].

Conclusion

The main idea that emerges from our study is that the dynamical or hidden symmetries discovered for the first time for the Newtonian and the Hookean potentials, are not a specific property of the latter, but instead are inherent to all the spherical symmetric potentials. This feature, that seems to have been widely accepted without solid basis after the works of Refs. [6, 7], must now be consider as well established. In the case of the Newtonian and the Hookean potentials, their true origin seems to have been masked by the fact that those potentials are the only ones providing closed orbits at any distance from the source. Indeed, this important property has often been considered, without any certainty, as the cause of the occurrence of the observed internal symmetries, or vice versa: see Ref. [13] §6.11. Thus, we must dissociate the existence of a dynamical symmetry from that of closed trajectories. This assertion is reinforced by the fact that even central potentials that are everywhere repulsive also provide a dynamical symmetry though they cannot produce any closed trajectory.

Let us review the different arguments leading to our conclusion.

An immediate consequence of the spherical symmetry of the potential is that, in the 2-plane of motion, the distance rr from the source is finally the single variable that determines two important geometrical characteristics of a trajectory, namely its limits and the existence of turning points, expected for all such potentials. Thanks again to the spherical symmetry, turning points define symmetry axes that may be considered as the true origin of a dynamical symmetry. Let us remark that a symmetry relative to an axis is generally reflected in the expression of the polar equation representing a given set of trajectories, and is therefore a common property of the latter. A natural way to characterize such axis is by means of its unit vector, which has been called here a symmetry vector. This vector is lying in the 2-plane of motion and is constant in time. For a nn-dimensional configuration space, its n−1n-1 independent components can be chosen as the first integrals forming, together with angular momentum and energy, a complete set of 2n−12n-1 fundamental and independent first integrals for the motion under study. Poisson brackets between those first integrals yield first integrals that cannot be independent of them. So, we may expect that performing Poisson brackets between fundamental first integrals leads to relations similar to that defining a Lie algebra. Here, the underlying idea is that there must exist a continuous group of transformations, the dynamical group, connecting all trajectories having a common symmetry, and which is generated by the said algebra. This is indeed realized, and the Lie relations obtained with a unit symmetry vector are very simple and apply to any situation. With a slightly modified symmetry vector, the obtained Lie algebra is homomorphic to that of the SO(3,1)SO(3,1) group.

It is worth mentioning here that the axial symmetry resulting from the inoperative change of ψ\psi into −ψ-\psi cannot of course be deduced from the equations of motion using the formalism of Lie’s transformations, since the latter applies only to continuous symmetries. The dynamical group does not describe directly this operation either, but takes into account the presence of an axis of symmetry for each trajectory.

To summarize, for the problem under study, what is often called “hidden symmetry” metry” and expected to be accounted for by a dynamical symmetry group is just a mathematical expression of the symmetry of the trajectories (or at least some of them), which goes beyond the simple case of the Newtonian and the Hookean potentials. Let us also emphasize here the status of a symmetry group as a stabilizer of the common structure of the trajectories under study, Ref. [1].

Examples have been given, showing that dynamical symmetries are not only present for potential in a simple power law in rr, just as in the Kepler and the Hookean problems, but also occur in problems with composite (or inhomogeneous) potentials. At this point, It must be remembered that energy, which is one of the parameters defining trajectories, cannot be modified by a dynamical group as defined above. Therefore, one has to enlarge this group in finding, for each case, an additional symmetry providing, if it exists, a complete mapping of trajectories, just as “mechanical similarity” does for potentials in single power law.

Let us make some final observations. As is well known, having three independent and Poisson-involutive first integrals, namely the energy EE, the norm LL of the angular momentum and its component LzL_z, the motion of a particle in a spherical symmetric potential is said to be an integrable problem, in the Liouville sense. But, as shown in this article, for any potential V(r)V(r), its spherical symmetry provides us with two additional and independent first integrals. For this reason, the said problem should be re-qualified as maximally superintegrable.

Finally, it should be noted that finding a continuous dynamical group for the system considered in this article appears rather natural, as explained below. An obvious but essential condition for the existence of such a group is the possibility of a continuous (or analytic) link between trajectories (complete geometrical curves). This possibility does exist for our system, and the said link is supposed to be realized mathematically by the transformations of the group, just as canonical transformations. Changing a trajectory into another of the same species remains to change continuously the values of the above-mentioned fundamental first integrals. But the latter being Poisson-involutive, the transformations they generate through Poisson brackets are unable to achieve this change. The rotational group also cannot change the values of EE and LL. The only possible conclusion is that there must exist at least one additional first integral which has non-zero Poisson brackets with the fundamental involutive first integrals and thus generates transformations changing in particular the value of LL (changing the value of EE is more complicated). In the present case, we have found two new first integrals expanding the action of the rotational group alone while carrying a common geometrical property of trajectories. From this observation, it is also natural to ask whether a dynamical group could exist for any integrable system whose solutions can be linked continuously. If so, such an integrable system would also be de facto superintegrable. Then, the simple condition of a continuous link between solutions would appear independent of and, conceptually, more important than the very nature of the system under study, that nature manifesting only in the representation of the group and the degree of superintegrability, specific to that system. This last conjecture is also based on the fact that the dynamical group considered in this article can be described by a single formalism, whatever the potential.

All the above observations might be of interest to researchers working on maximally superintegrability, Refs. [17, 18, 19, 20, 21], and its extension to systems having non-spherical symmetric potentials, see Refs. [22, 23].

Axes of symmetry of a trajectory

In the demonstration of his theorem, Bertrand started by saying that turning points are defining symmetry axes of trajectories, a statement which, in our opinion, deserves more explanation than that is usually given in textbooks, e.g. in Ref. [24] §14.

Let us consider a trajectory inside the circular crown rm≤r≤rMr_m \le r \le r_M, making more than one tour around OO and intersecting itself in a double point P0P_0 at the intermediate distance r0r_0. Let BaB_a and BbB_b be the two branches of the curve crossing at P0P_0. The polar angle of a radius vector r=rer\mathbf{r}=r\mathbf{e}_r relatively to OP0→=r0er0\overrightarrow{OP_0}=r_0\mathbf{e}_{r_0} is, mod. 2π2\pi,

ψ=ψ(r)=L22m∫r0r±dRR2E−U(R)(24)\psi=\psi(r)=\sqrt{\frac{L^2}{2m}}\int_{r_0}^{r}\frac{\pm dR}{R^2\sqrt{E-U(R)}} \tag*{(24)}

It is important to realize that, apart the parameters EE and LL, this angle is a function of the single variable rr. At any point, the tangent has only two possible orientations, according to the sign in (A.1). We can suppose that the plus sign and the minus sign are for BaB_a and for BbB_b, respectively. Consequently, BaB_a is a branch where rr is increasing from rmr_m to rMr_M and BbB_b a branch where rr is decreasing from rMr_M to rmr_m. It is then clear that any circle with center OO and radius r1r_1 such that rm≤r1≤rMr_m \le r_1 \le r_M intersects BaB_a at a point MaM_a and BbB_b at a point MbM_b. These two points being equidistant from OO, their corresponding polar angles, as given by (A.1), are simply opposite (mod. 2π2\pi). Hence, r(ψ)=r(−ψ)r(\psi)=r(-\psi), which means that OP0OP_0 is an axis of symmetry. The same kind of reasoning can be applied as well to the limit case where P0P_0 is a turning point. In this case, the branches BaB_a and BbB_b are respectively ending and beginning at P0P_0, and both have the same tangent at this point. We thus conclude that every turning point also defines an axis of symmetry. Moreover, if the trajectory crosses again an axis of symmetry defined by a double point in another simple point P1P_1 at a polar angle θ0\theta_0, the latter must be a turning point. This is so because r(θ)r(\theta) should be symmetric relatively to the value θ0\theta_0, and since P1P_1 is a simple point, we must have dr/dθ(θ0)=0dr/d\theta(\theta_0)=0.

To be more precise, let I1I_1, I2I_2 and I3I_3 three successive turning points at r=rmr=r_m, and S1S_1, S2S_2 the two turning points at r=rMr=r_M, following after I1I_1 and I2I_2, respectively. Then, the part of the curve between I1I_1 and I2I_2 is symmetric with respect to OS1OS_1, and the part of the curve between I2I_2 and I3I_3 is symmetric with respect to OS2OS_2. Moreover, those two parts of the curve are symmetric with respect to OI2OI_2. As time passes, the same scenario is repeated with other turning points. Then, if the angle Δθ\Delta\theta between OI1OI_1 and OS1OS_1, Eq. (6), is not commensurable with π\pi, the trajectory fills the entire space of the circular crown. and the previous axes of symmetry being in infinite number cannot be distinguished, except maybe one of them, because the polar equation of the trajectory must reflects the said symmetry. The trajectory has a finite number of turning points and corresponding axes of symmetry if and only if Δθ\Delta\theta is commensurable with π\pi. In that case, it is closed. In general, this case arises only for exceptional values of EE and LL. As we know, only the Newtonian and the Hookean potentials produce closed orbits for all corresponding admissible values of these parameters.

Turning points also exist for bounded trajectories inside a disk r≤r0r \le r_0 or unbounded trajectories outside that disk, likely in a limited number for the latter case, for r=r0r = r_0. So, their polar equation must also reflects some symmetry, a feature that can be used to define a symmetry vector also for such trajectories.

About the transformation TA\mathcal{T}_A

In Special Relativity, all inertial frames are linked by Lorentz transformations These transformations form of a Lie group, the Lorentz group, generated by a Lie algebra homomorphic to that of the group SL(2,C)SL(2,C). With an appropriate choice of their spatial reference axes, two inertial frames R\mathcal{R} and R′\mathcal{R}', having a relative velocity V=Vn\mathbf{V}=V\mathbf{n} along an axis defined by the unit vector n\mathbf{n} (VV is positive or negative), can always be connected by a Lorentz boost along n\mathbf{n}. This means that given an event whose measured data “time-position” are (t,r)(t,\mathbf{r}) in R\mathcal{R} and (t′,r′)(t',\mathbf{r}') in R′\mathcal{R}', these datas can be compared by means of the relations

ct′=ctcosh⁡ξ+n⋅rsinh⁡ξn⋅r′=n⋅rcosh⁡ξ+ctsinh⁡ξn×r′=r′−n(n⋅r′)=n×r=r−n(n⋅r)\begin{aligned} ct' &= ct\cosh\xi+ \mathbf{n}\cdot\mathbf{r}\sinh\xi\\ \mathbf{n}\cdot\mathbf{r}' &= \mathbf{n}\cdot\mathbf{r}\cosh\xi+ ct\sinh\xi\\ \mathbf{n}\times\mathbf{r}' &= \mathbf{r}'-\mathbf{n}(\mathbf{n}\cdot\mathbf{r}')=\mathbf{n}\times\mathbf{r}=\mathbf{r}-\mathbf{n}(\mathbf{n}\cdot\mathbf{r}) \end{aligned}

with ξ=tanh⁡−1(Vc)\xi=\tanh^{-1}\left(\frac{V}{c}\right)

cc being the speed of light in vacuum, or

ct′=ctcosh⁡ξ+n⋅rsinh⁡ξ,r′=r+n(cosh⁡ξ−1)(n⋅r)+nctsinh⁡ξ(25)ct'=ct\cosh\xi+\mathbf{n}\cdot\mathbf{r}\sinh\xi,\quad\mathbf{r}'=\mathbf{r}+\mathbf{n}(\cosh\xi-1)(\mathbf{n}\cdot\mathbf{r})+\mathbf{n}ct\sinh\xi \tag*{(25)}

The Lorentz boost, as defined by Eq. (C.1), applies as well to any 4-vector Q\mathbf{Q} in Minkowski 4-space. Let Q0Q_0 and Q\mathbf{Q} its time component and spatial components in R\mathcal{R}. In R′\mathcal{R}' they are Q0′Q'_0 and Q′\mathbf{Q}' and we have

Q0′=Q0cosh⁡ξ+n⋅Qsinh⁡ξ,(26)Q'_0=Q_0\cosh\xi+\mathbf{n}\cdot\mathbf{Q}\sinh\xi, \tag*{(26)}
Q′=Q+n(cosh⁡ξ−1)(n⋅Q)+nQ0sinh⁡ξ\mathbf{Q}'=\mathbf{Q}+\mathbf{n}(\cosh\xi-1)(\mathbf{n}\cdot\mathbf{Q})+\mathbf{n}Q_0\sinh\xi

The similarity between Eqs. (15), (18) and Eq. (C.2) leads us to interpret the transformation TA\mathcal{T}_A as a Lorentz boost acting on the fictitious 4-vector B\mathbf{B} having $L as time component and B\mathbf{B} as its spatial part. This 4-vector has the particularity to be lightlike, because its invariant squared norm is zero:

B2=B02−B2=L2−L2=0B^2 = B_0^2 - \mathbf{B}^2 = L^2 - \mathbf{L}^2 = 0

Next, let us define the rotation RA:(U,W,ℓ)→(U′,W′,ℓ′)R_A:(\mathbf{U},\mathbf{W},\ell)\to(\mathbf{U}',\mathbf{W}',\ell') induced by TAT_A. It is easy to obtain the angle of rotation γ\gamma, using the property tr⁡RA=1+2cos⁡γ\operatorname{tr}R_A=1+2\cos\gamma. We have then

1+2cos⁡γ=U′⋅U+W′⋅W+ℓ′⋅ℓ,1+2\cos\gamma=\mathbf{U}'\cdot\mathbf{U}+\mathbf{W}'\cdot\mathbf{W}+\ell'\cdot\ell,

whence

cos⁡γ=1ρ[1+(n⋅W)2(cosh⁡ξ−1)+n⋅Wsinh⁡ξ]\cos\gamma=\frac{1}{\rho}\left[1+(\mathbf{n}\cdot\mathbf{W})^2(\cosh\xi-1)+\mathbf{n}\cdot\mathbf{W}\sinh\xi\right]

with

ρ=cosh⁡ξ+n⋅Wsinh⁡ξ\rho=\cosh\xi+\mathbf{n}\cdot\mathbf{W}\sinh\xi

and

sin⁡γ=±1−(n⋅W)2ρ[sinh⁡ξ+n⋅W(cosh⁡ξ−1)]\sin\gamma=\pm\frac{\sqrt{1-(\mathbf{n}\cdot\mathbf{W})^2}}{\rho}\left[\sinh\xi+\mathbf{n}\cdot\mathbf{W}(\cosh\xi-1)\right]

The components NXN_X, NYN_Y and NZN_Z, relative to the basis (U,W,ℓ)(\mathbf{U},\mathbf{W},\ell), of the unit vector N\mathbf{N} defining the axis of rotation are subsequently found using the well known transformation formula

a′=acos⁡γ+(N×a)sin⁡γ+(1−cos⁡γ)(N⋅a)N\mathbf{a}'=\mathbf{a}\cos\gamma+(\mathbf{N}\times\mathbf{a})\sin\gamma+(1-\cos\gamma)(\mathbf{N}\cdot\mathbf{a})\mathbf{N}

We have

U′⋅U=cos⁡γ+(1−cos⁡γ)NX2=1−(n⋅U)2cosh⁡ξ−1ρ,\mathbf{U}'\cdot\mathbf{U}=\cos\gamma+(1-\cos\gamma)N_X^2=1-(\mathbf{n}\cdot\mathbf{U})^2\frac{\cosh\xi-1}{\rho},

hence

NX2=(n⋅ℓ)21−(n⋅W)2,N_X^2=\frac{(\mathbf{n}\cdot\ell)^2}{1-(\mathbf{n}\cdot\mathbf{W})^2},

accounting for the identity (n⋅U)2+(n⋅W)2+(n⋅ℓ)2=n2=1(\mathbf{n}\cdot\mathbf{U})^2+(\mathbf{n}\cdot\mathbf{W})^2+(\mathbf{n}\cdot\ell)^2=\mathbf{n}^2=1. In the same way, we obtain

ℓ′⋅ℓ=cos⁡γ+(1−cos⁡γ)NZ2=1−(n⋅ℓ)2cosh⁡ξ−1ρ,\ell'\cdot\ell=\cos\gamma+(1-\cos\gamma)N_Z^2=1-(\mathbf{n}\cdot\ell)^2\frac{\cosh\xi-1}{\rho},

hence

NZ2=(n⋅U)21−(n⋅W)2.N_Z^2=\frac{(\mathbf{n}\cdot\mathbf{U})^2}{1-(\mathbf{n}\cdot\mathbf{W})^2}.

Since NX2+NY2+NZ2=1N_X^2+N_Y^2+N_Z^2=1, we must have NY=0N_Y=0. This can be checked as follows. We have

W′⋅W=cos⁡γ+(1−cos⁡γ)NY2=1ρ[1+(n⋅W)2(cosh⁡ξ−1)+n⋅Wsinh⁡ξ]=cos⁡γ,W'\cdot W=\cos\gamma+(1-\cos\gamma)N_Y^2=\frac{1}{\rho}\left[1+(\boldsymbol{n}\cdot\boldsymbol{W})^2(\cosh\xi-1)+\boldsymbol{n}\cdot\boldsymbol{W}\sinh\xi\right]=\cos\gamma,

hence NY=0N_Y=0.

Further, using

U′⋅W=NZsin⁡γ=−n⋅Uρ[sinh⁡ξ+n⋅W(cosh⁡ξ−1)],U'\cdot W=N_Z\sin\gamma=-\frac{\boldsymbol{n}\cdot\boldsymbol{U}}{\rho}\left[\sinh\xi+\boldsymbol{n}\cdot\boldsymbol{W}(\cosh\xi-1)\right],
U′⋅ℓ=NZNX(1−cos⁡γ)=−(n⋅ℓ)(n⋅U)cosh⁡ξ−1ρU'\cdot\ell=N_ZN_X(1-\cos\gamma)=-(\boldsymbol{n}\cdot\ell)(\boldsymbol{n}\cdot\boldsymbol{U})\frac{\cosh\xi-1}{\rho}

we arrive at

NX=±n⋅ℓ1−(n⋅W)2,NZ=∓n⋅U1−(n⋅W)2,N_X=\pm\frac{\boldsymbol{n}\cdot\ell}{\sqrt{1-(\boldsymbol{n}\cdot\boldsymbol{W})^2}},\qquad N_Z=\mp\frac{\boldsymbol{n}\cdot\boldsymbol{U}}{\sqrt{1-(\boldsymbol{n}\cdot\boldsymbol{W})^2}},

and

Ni=±11−(n⋅W)2nj[ℓjUi−Ujℓi]=±11−(n⋅W)2ϵijknjWk=±(n×W)i1−(n⋅W)2.\begin{aligned} N_i&=\pm\frac{1}{\sqrt{1-(\boldsymbol{n}\cdot\boldsymbol{W})^2}}n_j[\ell_jU_i-U_j\ell_i] =\pm\frac{1}{\sqrt{1-(\boldsymbol{n}\cdot\boldsymbol{W})^2}}\epsilon_{ijk}n_jW_k\\ &=\pm\frac{(\boldsymbol{n}\times\boldsymbol{W})_i}{\sqrt{1-(\boldsymbol{n}\cdot\boldsymbol{W})^2}}. \end{aligned}

References

  1. [1]C. Carimalo, Symmetries and stability of motions in the Newtonian and the Hookean potentials, Theor. Appl. Mech. 49(1) (2022), 61–69.DOI
  2. [2]P. S. Laplace, Traité de Mécanique Céleste, 1799, Tome I, Première partie, Livre II, 165.
  3. [3]C. Runge, Vektoranalysis, Leipzig, 1919.
  4. [4]W. Lenz, Über den Bewegungsverlauf und Quantenzustände der gestörten Keplerbewegung, Z. Phys. 24 (1924), 197–207.
  5. [5]C. Carimalo, Dynamical symmetries behind Bertrand’s theorem, Am. J. Phys. 89 (2021), 1012–1017.DOI
  6. [6]H. Bacry, H. Ruegg, J-M Souriau, Dynamical Groups and Spherical Potentials in Classical Mechanics, Comm. math. Phys. 3 (1966), 323-333.
  7. [7]D. M. Fradkin, Existence of the Dynamical Symmetries O₄ and SU₃ for All Classical Central Potentials Problems, Progr. Theor. Phys. 37 (1967), 798-812.
  8. [8]N. Mukunda, Dynamical Symmetries and Classical Mechanics, Phys. Rev. 155 (1967), 1383-1386.DOI
  9. [9]J. Bertrand, Théorème relatif au mouvement d’un point attiré vers un centre fixe, C. R. Acad. Sci., Paris 77(16) (1873), 849–853.
  10. [10]H. Goldstein, J. L. Safko, C. P. Poole Jr., Classical Mechanics, 3rd ed., Pearson, 2014.
  11. [11]V. I. Arnold, Mathematical Methods in Classical Mechanics, 2nd ed., Springer-Verlag, 1989, Sec. 8.
  12. [12]C. Carimalo, Transmutation of Central Forces and Bertrand’s Theorem, Theor. Appl. Mech. 50 (2023), 1–11.DOI
  13. [13]H. Bacry, Lectures on Group Theory and Particle Theory, Gordon and Breach, 1977.
  14. [14]P.G.L. Leach, G.P. Flessas, Generalisations of the Laplace-Runge-Lenz vector, J. Nonlin. Math. Phys. 10 (2003), 340-423.
  15. [15]Y. Grandati, A. Bérard, H. Mohrbach, On Peres approach to Fradkin-Bacry-Ruegg-Souriau’s perihelion vector, Cent. Eur. J. Phys. 9 (2011), 88-95.arxiv.org/abs/0810.3780
  16. [16]C. Reed, An attractive/repulsive central-force orbit for student exercises, Eur. J. Phys. 42 (2021), 045007.DOI
  17. [17]F. Tremblay, A. V. Turbiner, P. Winternitz, An infinite family of solvable and integrable quantum systems on a plane, J. Phys. A, Math. Theor. 42 (2009), 242001.DOI
  18. [18]W. Miller Jr, S. Post, P. Winternitz, Classical and quantum superintegrability with applications, J. Phys. A, Math. Theor. 46 (2013), 423001.DOI
  19. [19]S. C. Anco, T. Meadows, V. Pascuzzi, Some new aspects of first integrals and symmetries for central force dynamics, J. Math. Phys. 57 (2016), 062901.DOI
  20. [20]I. Marquette, M. Sajedi, P. Winternitz, Fourth order superintegrable systems separating in Cartesian coordinates I. Exotic quantum potentials, J. Phys. A, Math. Theor. 50 (2017), 315201.DOI
  21. [21]S. C. Anco, A. Ballesteros, M. L. Gandarias, Global versus local superintegrability of nonlinear oscillators, Phys. Lett. A 383 (2019), 801-807.DOI
  22. [22]S. Post, P. Winternitz, An Infinite Family of Superintegrable Deformations of the Coulomb Potential, J. Phys. A : Math. and Theor. 43 (2010) 222001.DOI
  23. [23]I.A. Bizaev, A. V. Borisov, I. S. Mamaev, Superintegrable Generalizations of the Kepler and Hook Problems, Regular and Chaotic Dynamics, 19 (2014), 415-433.
  24. [24]L. D. Landau, E. M. Lifshitz, Mechanics, 2nd ed., Pergamon Press, 1969.

Paper details

vlnc:W_ws927ecm
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents