PDES in the inverse problem of dynamics

Abstract

Authors

Mira-Cristiana Anisiu
Tiberiu Popoviciu Institute of Numerical Analysis, Romanian Academy

Keywords

References

PDF

(pdf file here)

Cite this paper as:

Anisiu M.C., PDES in the inverse problem of dynamics, Analysis and Optimization of Differential Systems, vol. 121, 2003, pag. 13-20

About this paper

Journal
Publisher Name
Print ISSN

Not available yet.

Online ISSN

Not available yet.

Google Scholar Profile
2003-Anisiu-PDEs

PDES IN THE INVERSE PROBLEM OF DYNAMICS

Mira-Cristiana AnisiuT. Popoviciu Institute of Numerical AnalysisRomanian Academy37 Republicii st., Cluj-Napoca, RO-3400 Romaniamira@math.ubbcluj.ro

Abstract

The basic equations are exposed for the following version of the inverse problem of dynamics: determine the two-dimensional potential compatible with a given family of orbits, traced by a material point. If the potential is known in advance, a nonlinear equation is satisfied by the function representing the family of orbits. Its solutions are studied in the presence of additional information on the family. The possibility of programming the motion of a material point in a preassigned region of the plane is also considered.

Keywords: inverse problem of dynamics, Szebehely's and Bozis' equations.

1. Introduction

The version of the inverse problem of dynamics discussed in this paper consists in finding the two-dimensional potential which governs the motion of a dynamical system, knowing a given family of orbits. The first outstanding results are due to Newton (1687), who found the force law compatible with Kepler's laws. The paper which gave a new impulse to the field of the inverse problems was that of Szebehely's (1974). Information on various other aspects (nonconservative systems, threedimensional ones, rotating frames, holonomic systems with n n nnn degrees of freedom etc) of the inverse problem is contained in Bozis (1995) and Anisiu (1998).

2. The main tools of the inverse problem

Given a family of curves
(2.1) f ( x , y ) = c (2.1) f ( x , y ) = c {:(2.1)f(x","y)=c:}\begin{equation*} f(x, y)=c \tag{2.1} \end{equation*}(2.1)f(x,y)=c
one looks for a potential for which this family is an orbit family of a particle. The problem is then to determine the potential V C 1 ( D ) V C 1 ( D ) V inC^(1)(D)V \in C^{1}(D)VC1(D) for which the equations of the motion of a unit mass particle are
(2.2) x ¨ = V x y ¨ = V y , (2.2) x ¨ = V x y ¨ = V y , {:(2.2)x^(¨)=-V_(x)quady^(¨)=-V_(y)",":}\begin{equation*} \ddot{x}=-V_{x} \quad \ddot{y}=-V_{y}, \tag{2.2} \end{equation*}(2.2)x¨=Vxy¨=Vy,
knowing a given family of orbits (2.1). The first order partial differential equation satisfied by the potential V V VVV was obtained by Szebehely (1974).
The system (2.2) admits the energy integral x ˙ 2 + y ˙ 2 = 2 ( E ( f ) V ) x ˙ 2 + y ˙ 2 = 2 ( E ( f ) V ) x^(˙)^(2)+y^(˙)^(2)=2(E(f)-V)\dot{x}^{2}+\dot{y}^{2}=2(E(f)-V)x˙2+y˙2=2(E(f)V), E ( f ) E ( f ) E(f)E(f)E(f) being constant on every orbit in (2.1).
Theorem 2.1 Let D R 2 D R 2 D subR^(2)D \subset \mathbb{R}^{2}DR2 be an open set and f C 2 ( D ) f C 2 ( D ) f inC^(2)(D)f \in C^{2}(D)fC2(D) such that f x 2 ( x , y ) + f y 2 ( x , y ) 0 f x 2 ( x , y ) + f y 2 ( x , y ) 0 f_(x)^(2)(x,y)+f_(y)^(2)(x,y)!=0f_{x}^{2}(x, y)+f_{y}^{2}(x, y) \neq 0fx2(x,y)+fy2(x,y)0 for each ( x , y ) D ( x , y ) D (x,y)in D(x, y) \in D(x,y)D. If the system (2.2) admits as orbits the curves of the family (2.1), then the function V V VVV satisfies the partial differential equation
(2.3) f x V x + f y V y 2 ( E ( f ) V ) f x 2 + f y 2 ( f x x f y 2 2 f x y f x f y + f y y f x 2 ) = 0 (2.3) f x V x + f y V y 2 ( E ( f ) V ) f x 2 + f y 2 f x x f y 2 2 f x y f x f y + f y y f x 2 = 0 {:(2.3)f_(x)V_(x)+f_(y)V_(y)-(2(E(f)-V))/(f_(x)^(2)+f_(y)^(2))(f_(xx)f_(y)^(2)-2f_(xy)f_(x)f_(y)+f_(yy)f_(x)^(2))=0:}\begin{equation*} f_{x} V_{x}+f_{y} V_{y}-\frac{2(E(f)-V)}{f_{x}^{2}+f_{y}^{2}}\left(f_{x x} f_{y}^{2}-2 f_{x y} f_{x} f_{y}+f_{y y} f_{x}^{2}\right)=0 \tag{2.3} \end{equation*}(2.3)fxVx+fyVy2(E(f)V)fx2+fy2(fxxfy22fxyfxfy+fyyfx2)=0
Proof. From (2.1) we obtain x ˙ f x + y ˙ f y = 0 x ˙ f x + y ˙ f y = 0 x^(˙)f_(x)+y^(˙)f_(y)=0\dot{x} f_{x}+\dot{y} f_{y}=0x˙fx+y˙fy=0, hence x ˙ x ˙ x^(˙)\dot{x}x˙ and y ˙ y ˙ y^(˙)\dot{y}y˙ will be
(2.4) x ˙ = ± k f y y ˙ = k f x (2.4) x ˙ = ± k f y y ˙ = k f x {:(2.4)x^(˙)=+-sqrtkf_(y)quady^(˙)=∓sqrtkf_(x):}\begin{equation*} \dot{x}= \pm \sqrt{k} f_{y} \quad \dot{y}=\mp \sqrt{k} f_{x} \tag{2.4} \end{equation*}(2.4)x˙=±kfyy˙=kfx
where the sign depends on the sense of the motion on the orbit, and k C 1 ( D , R + ) k C 1 D , R + k inC^(1)(D,R_(+))k \in C^{1}\left(D, \mathbb{R}_{+}\right)kC1(D,R+)is an arbitrary function. Differentiating again we have
x ¨ = k ( f x y f y f y y f x ) + ( k x f y k y f x ) f y / 2 y ¨ = k ( f x y f x f x x f y ) ( k x f y k y f x ) f x / 2 . x ¨ = k f x y f y f y y f x + k x f y k y f x f y / 2 y ¨ = k f x y f x f x x f y k x f y k y f x f x / 2 . {:[x^(¨)=k(f_(xy)f_(y)-f_(yy)f_(x))+(k_(x)f_(y)-k_(y)f_(x))f_(y)//2],[y^(¨)=k(f_(xy)f_(x)-f_(xx)f_(y))-(k_(x)f_(y)-k_(y)f_(x))f_(x)//2.]:}\begin{aligned} & \ddot{x}=k\left(f_{x y} f_{y}-f_{y y} f_{x}\right)+\left(k_{x} f_{y}-k_{y} f_{x}\right) f_{y} / 2 \\ & \ddot{y}=k\left(f_{x y} f_{x}-f_{x x} f_{y}\right)-\left(k_{x} f_{y}-k_{y} f_{x}\right) f_{x} / 2 . \end{aligned}x¨=k(fxyfyfyyfx)+(kxfykyfx)fy/2y¨=k(fxyfxfxxfy)(kxfykyfx)fx/2.
It follows
V x = k ( f x y f y f y y f x ) + ( k x f y k y f x ) f y / 2 V y = k ( f x y f x f x x f y ) ( k x f y k y f x ) f x / 2 V x = k f x y f y f y y f x + k x f y k y f x f y / 2 V y = k f x y f x f x x f y k x f y k y f x f x / 2 {:[-V_(x)=k(f_(xy)f_(y)-f_(yy)f_(x))+(k_(x)f_(y)-k_(y)f_(x))f_(y)//2],[-V_(y)=k(f_(xy)f_(x)-f_(xx)f_(y))-(k_(x)f_(y)-k_(y)f_(x))f_(x)//2]:}\begin{aligned} & -V_{x}=k\left(f_{x y} f_{y}-f_{y y} f_{x}\right)+\left(k_{x} f_{y}-k_{y} f_{x}\right) f_{y} / 2 \\ & -V_{y}=k\left(f_{x y} f_{x}-f_{x x} f_{y}\right)-\left(k_{x} f_{y}-k_{y} f_{x}\right) f_{x} / 2 \end{aligned}Vx=k(fxyfyfyyfx)+(kxfykyfx)fy/2Vy=k(fxyfxfxxfy)(kxfykyfx)fx/2
and
(2.5) f x V x + f y V y = k ( 2 f x y f x f y f x 2 f y y f y 2 f x x ) . (2.5) f x V x + f y V y = k 2 f x y f x f y f x 2 f y y f y 2 f x x . {:(2.5)f_(x)V_(x)+f_(y)V_(y)=-k(2f_(xy)f_(x)f_(y)-f_(x)^(2)f_(yy)-f_(y)^(2)f_(xx)).:}\begin{equation*} f_{x} V_{x}+f_{y} V_{y}=-k\left(2 f_{x y} f_{x} f_{y}-f_{x}^{2} f_{y y}-f_{y}^{2} f_{x x}\right) . \tag{2.5} \end{equation*}(2.5)fxVx+fyVy=k(2fxyfxfyfx2fyyfy2fxx).
From the energy integral and the relations (2.4), the function k k kkk is given by k = 2 ( V E ( f ) ) / ( f x 2 + f y 2 ) k = 2 ( V E ( f ) ) / f x 2 + f y 2 k=2(V-E(f))//(f_(x)^(2)+f_(y)^(2))k=2(V-E(f)) /\left(f_{x}^{2}+f_{y}^{2}\right)k=2(VE(f))/(fx2+fy2). We replace it in (2.5) and obtain Szebehely's equation (2.3).
Remark 2.1 In the paper Bozis (1983) the functions
(2.6) γ = f y f x Γ = γ γ x γ y (2.6) γ = f y f x Γ = γ γ x γ y {:(2.6)gamma=(f_(y))/(f_(x))quad Gamma=gammagamma_(x)-gamma_(y):}\begin{equation*} \gamma=\frac{f_{y}}{f_{x}} \quad \Gamma=\gamma \gamma_{x}-\gamma_{y} \tag{2.6} \end{equation*}(2.6)γ=fyfxΓ=γγxγy
were introduced for the family of curves (2.1). Using these notations, Szebehely's equation was written in the simpler form (Bozis, 1983)
(2.7) V x + γ V y + 2 Γ ( E V ) 1 + γ 2 = 0 . (2.7) V x + γ V y + 2 Γ ( E V ) 1 + γ 2 = 0 . {:(2.7)V_(x)+gammaV_(y)+(2Gamma(E-V))/(1+gamma^(2))=0.:}\begin{equation*} V_{x}+\gamma V_{y}+\frac{2 \Gamma(E-V)}{1+\gamma^{2}}=0 . \tag{2.7} \end{equation*}(2.7)Vx+γVy+2Γ(EV)1+γ2=0.
From the equation above, an inequality can be deduced because the kinetic energy of the particle B = E V B = E V B=E-VB=E-VB=EV is nonnegative.
Corollary 2.1 The potential V V VVV satisfies the inequality (Bozis and Ichtiaroglou, 1994)
(2.8) Γ ( V x + γ V y ) 0 . (2.8) Γ V x + γ V y 0 . {:(2.8)Gamma(V_(x)+gammaV_(y)) <= 0.:}\begin{equation*} \Gamma\left(V_{x}+\gamma V_{y}\right) \leq 0 . \tag{2.8} \end{equation*}(2.8)Γ(Vx+γVy)0.
In what follows the functions are considered to be sufficiently smooth.
The equation (2.7) has the disadvantage that it contains the energy E E EEE, which in general is not known in advance. Bozis (1984) has eliminated the energy, obtaining a second order partial differential equation.
Theorem 2.2 Denoting κ = 1 γ γ , λ = Γ y γ Γ x γ Γ , μ = λ γ + 3 Γ γ κ = 1 γ γ , λ = Γ y γ Γ x γ Γ , μ = λ γ + 3 Γ γ kappa=(1)/(gamma)-gamma,quad lambda=(Gamma_(y)-gammaGamma_(x))/(gamma Gamma),quad mu=lambda gamma+(3Gamma)/(gamma)\kappa=\frac{1}{\gamma}-\gamma, \quad \lambda=\frac{\Gamma_{y}-\gamma \Gamma_{x}}{\gamma \Gamma}, \quad \mu=\lambda \gamma+\frac{3 \Gamma}{\gamma}κ=1γγ,λ=ΓyγΓxγΓ,μ=λγ+3Γγ, the potential V V VVV satisfies the second order equation (Bozis, 1984)
(2.9) V x x + κ V x y + V y y = λ V x + μ V y . (2.9) V x x + κ V x y + V y y = λ V x + μ V y . {:(2.9)-V_(xx)+kappaV_(xy)+V_(yy)=lambdaV_(x)+muV_(y).:}\begin{equation*} -V_{x x}+\kappa V_{x y}+V_{y y}=\lambda V_{x}+\mu V_{y} . \tag{2.9} \end{equation*}(2.9)Vxx+κVxy+Vyy=λVx+μVy.
Proof. The energy E = E ( f ) E = E ( f ) E=E(f)E=E(f)E=E(f) can be eliminated from equation (2.7) by solving it with respect to E E EEE and inserting E E EEE into
(2.10) E y = γ E x , (2.10) E y = γ E x , {:(2.10)E_(y)=gammaE_(x)",":}\begin{equation*} E_{y}=\gamma E_{x}, \tag{2.10} \end{equation*}(2.10)Ey=γEx,
the slope function γ γ gamma\gammaγ being given by (2.6). Relation (2.10) holds because from E = E ( f ) E = E ( f ) E=E(f)E=E(f)E=E(f) we have E x = E f x E x = E f x E_(x)=E^(')f_(x)E_{x}=E^{\prime} f_{x}Ex=Efx and E y = E f y E y = E f y E_(y)=E^(')f_(y)E_{y}=E^{\prime} f_{y}Ey=Efy.
The main tools used in the inverse problem are: Szebehely's quasilinear first order partial differential equation (2.7), Bozis' linear second order partial differential equation (2.9), and the inequality (2.8). If one manages to obtain a solution of (2.9), from (2.7) one can find the value of the energy, and from (2.8) the plane region where the actual trajectories are allowed to take place.
Example 2.1 For the case of the family f ( x , y ) = x 2 + y 2 f ( x , y ) = x 2 + y 2 f(x,y)=x^(2)+y^(2)f(x, y)=x^{2}+y^{2}f(x,y)=x2+y2 of concentric circles, Broucke and Lass (1977) found (in polar coordinates r , θ r , θ r,thetar, \thetar,θ ) the potential V ( r , θ ) = g ( r ) + h ( θ ) / r 2 V ( r , θ ) = g ( r ) + h ( θ ) / r 2 V(r,theta)=g(r)+h(theta)//r^(2)V(r, \theta)=g(r)+h(\theta) / r^{2}V(r,θ)=g(r)+h(θ)/r2, where g g ggg and h h hhh are arbitrary functions. The energy function is in this case E = g ( r ) + r g ( r ) / 2 E = g ( r ) + r g ( r ) / 2 E=g(r)+rg^(')(r)//2E=g(r)+r g^{\prime}(r) / 2E=g(r)+rg(r)/2, and the allowed region obtained from (2.8) is given by g ( r ) 2 h ( θ ) / r 3 0 g ( r ) 2 h ( θ ) / r 3 0 g^(')(r)-2h(theta)//r^(3) >= 0g^{\prime}(r)-2 h(\theta) / r^{3} \geq 0g(r)2h(θ)/r30. Other cases in which Szebehely's equation is solvable are exposed in Grigoriadou et al (1999).
The relation (2.9) was derived only for the case Γ 0 Γ 0 Gamma!=0\Gamma \neq 0Γ0. From (2.6) one can express Γ Γ Gamma\GammaΓ in terms of the derivatives of f f fff as
Γ = 2 f x y f x f y f x x f y 2 f y y f x 2 f x 3 , Γ = 2 f x y f x f y f x x f y 2 f y y f x 2 f x 3 , Gamma=(2f_(xy)f_(x)f_(y)-f_(xx)f_(y)^(2)-f_(yy)f_(x)^(2))/(f_(x)^(3)),\Gamma=\frac{2 f_{x y} f_{x} f_{y}-f_{x x} f_{y}^{2}-f_{y y} f_{x}^{2}}{f_{x}^{3}},Γ=2fxyfxfyfxxfy2fyyfx2fx3,
hence the family (2.1) consists of straight lines if and only if Γ = 0 Γ = 0 Gamma=0\Gamma=0Γ=0; in view of (2.6) this condition may be written as
(2.11) γ γ x γ y = 0 (2.11) γ γ x γ y = 0 {:(2.11)gammagamma_(x)-gamma_(y)=0:}\begin{equation*} \gamma \gamma_{x}-\gamma_{y}=0 \tag{2.11} \end{equation*}(2.11)γγxγy=0

3. Families of straight lines

The problem of determining the potentials under whose action families of straight lines are described was considered by Bozis and Anisiu (2001).
Szebehely's equation (2.7) for a family of straight lines becomes
(3.12) V x + γ V y = 0 (3.12) V x + γ V y = 0 {:(3.12)V_(x)+gammaV_(y)=0:}\begin{equation*} V_{x}+\gamma V_{y}=0 \tag{3.12} \end{equation*}(3.12)Vx+γVy=0
Remark 3.1 In what follows we shall consider γ 0 γ 0 gamma!=0\gamma \neq 0γ0, because for γ = 0 γ = 0 gamma=0\gamma=0γ=0 we obtain V x = 0 V x = 0 V_(x)=0V_{x}=0Vx=0, hence V = v ( y ) V = v ( y ) V=v(y)V=v(y)V=v(y); this potential produces only the family of vertical straight lines x = c x = c x=cx=cx=c.
We differentiate (3.12) with respect to x x xxx, then to y y yyy, and express γ x γ x gamma_(x)\gamma_{x}γx and γ y γ y gamma_(y)\gamma_{y}γy from the obtained equations. From (2.11) we get
(3.13) V x x + ( 1 γ γ ) V x y + V y y = 0 (3.13) V x x + 1 γ γ V x y + V y y = 0 {:(3.13)-V_(xx)+((1)/(gamma)-gamma)V_(xy)+V_(yy)=0:}\begin{equation*} -V_{x x}+\left(\frac{1}{\gamma}-\gamma\right) V_{x y}+V_{y y}=0 \tag{3.13} \end{equation*}(3.13)Vxx+(1γγ)Vxy+Vyy=0
this corresponds to Bozis' partial differential equation for the straight lines situation. We insert γ γ gamma\gammaγ obtained from (3.12) into (3.13) and get a nonlinear partial differential equation
(3.14) V x V y ( V x x V y y ) = V x y ( V x 2 V y 2 ) (3.14) V x V y V x x V y y = V x y V x 2 V y 2 {:(3.14)V_(x)V_(y)(V_(xx)-V_(yy))=V_(xy)(V_(x)^(2)-V_(y)^(2)):}\begin{equation*} V_{x} V_{y}\left(V_{x x}-V_{y y}\right)=V_{x y}\left(V_{x}^{2}-V_{y}^{2}\right) \tag{3.14} \end{equation*}(3.14)VxVy(VxxVyy)=Vxy(Vx2Vy2)
which must be satisfied by all potentials creating (among other orbits) a family of straight lines.
Remark 3.2 Given an adequate γ γ gamma\gammaγ (i.e. satisfying (2.11)), there exists infinitely many potentials V ( x , y ) V ( x , y ) V(x,y)V(x, y)V(x,y), found from (3.12), creating the monoparametric family of straight lines. But for an adequate V V VVV (i.e. satisfying (3.14) and depending on both variables x x xxx and y y yyy ), there corresponds to it exactly one γ γ gamma\gammaγ, found from (3.12), hence one family of straight lines.
Focussing our attention on equation (3.14), we mention some of its obvious solutions:
(i) V = V ( x ) V = V ( x ) quad V=V(x)\quad V=V(x)V=V(x) or V = V ( y ) V = V ( y ) V=V(y)V=V(y)V=V(y);
(ii) V = V ( k 1 x + k 2 y ) , k 1 V = V k 1 x + k 2 y , k 1 V=V(k_(1)x+k_(2)y),k_(1)V=V\left(k_{1} x+k_{2} y\right), k_{1}V=V(k1x+k2y),k1 and k 2 k 2 k_(2)k_{2}k2 constants;
(iii) V = x 2 + y 2 V = x 2 + y 2 V=x^(2)+y^(2)V=x^{2}+y^{2}V=x2+y2.
In case (i), equation (3.12) can be satisfied only for the trivial case, V = V = V=V=V= const, as follows from Remark 3.1. For the class of potentials in case (ii), we obtain from (3.12) the family of straight lines given
by γ = k 1 k 2 = γ = k 1 k 2 = gamma=-(k_(1))/(k_(2))=\gamma=-\frac{k_{1}}{k_{2}}=γ=k1k2= const. In case (iii), the family of straight lines has a homogeneous of order 0 slope function γ = x y γ = x y gamma=-(x)/(y)\gamma=-\frac{x}{y}γ=xy. The special solution V = x 2 + y 2 V = x 2 + y 2 V=x^(2)+y^(2)V=x^{2}+y^{2}V=x2+y2 allows us to state that all central potentials V = V ( r ) V = V ( r ) V=V(r)V=V(r)V=V(r) are solutions of (3.14).
Other classes of potentials, some expressed in polar coordinates, are given in Bozis and Anisiu (2001).

4. Equations of the direct problem

Bozis' equation represents a relation between the function γ γ gamma\gammaγ and the potential V V VVV. It can be used to face the direct problem of Dynamics: given a potential V V VVV, find the families of orbits which can be generated. The nonlinear second order differential equation relating potentials and orbits in the form suitable for the direct problem (Bozis, 1995) is
(4.15) γ 2 γ x x 2 γ γ x y + γ y y = γ γ x γ y V y γ + V x ( γ x V x + ( 2 γ γ x 3 γ y ) V y + γ ( V x x V y y ) + ( γ 2 1 ) V x y ) (4.15) γ 2 γ x x 2 γ γ x y + γ y y = γ γ x γ y V y γ + V x γ x V x + 2 γ γ x 3 γ y V y + γ V x x V y y + γ 2 1 V x y {:[(4.15)gamma^(2)gamma_(xx)-2gammagamma_(xy)+gamma_(yy)=(gammagamma_(x)-gamma_(y))/(V_(y)gamma+V_(x))],[(-gamma_(x)V_(x)+(2gammagamma_(x)-3gamma_(y))V_(y)+gamma(V_(xx)-V_(yy))+(gamma^(2)-1)V_(xy))]:}\begin{align*} & \gamma^{2} \gamma_{x x}-2 \gamma \gamma_{x y}+\gamma_{y y}=\frac{\gamma \gamma_{x}-\gamma_{y}}{V_{y} \gamma+V_{x}} \tag{4.15}\\ & \left(-\gamma_{x} V_{x}+\left(2 \gamma \gamma_{x}-3 \gamma_{y}\right) V_{y}+\gamma\left(V_{x x}-V_{y y}\right)+\left(\gamma^{2}-1\right) V_{x y}\right) \end{align*}(4.15)γ2γxx2γγxy+γyy=γγxγyVyγ+Vx(γxVx+(2γγx3γy)Vy+γ(VxxVyy)+(γ21)Vxy)
This is obtained by rearranging Bozis' equation (2.9).
Due to its nonlinearity in γ γ gamma\gammaγ, it is difficult to be solved. Additional information may help in searching for solutions. The case of families produced by homogeneous potentials was considered by Bozis and Stefiades (1993), and by Bozis and Grigoriadou (1993); the problem was reduced to solving ordinary differential equations. Homogeneous families produced by inhomogeneous potentials were studied by Bozis et al (1997), as well as families of orbits with γ = γ ( x ) γ = γ ( x ) gamma=gamma(x)\gamma=\gamma(x)γ=γ(x), corresponding to families f ( x , y ) = y + h ( x ) = c f ( x , y ) = y + h ( x ) = c f(x,y)=y+h(x)=cf(x, y)=y+h(x)=cf(x,y)=y+h(x)=c (Bozis et al, 2000); in these two cases γ γ gamma\gammaγ was found as the common root of some algebraic equations in γ γ gamma\gammaγ, with coefficients depending on V V VVV and on its derivatives.
The additional condition satisfied by γ γ gamma\gammaγ may be put in the terms of a first order differential equation. Indeed, if f f fff is homogeneous of degree m m mmm, then γ γ gamma\gammaγ is homogeneous of degree 0 . This happens if and only if x γ x + y γ y = 0 x γ x + y γ y = 0 xgamma_(x)+ygamma_(y)=0x \gamma_{x}+y \gamma_{y}=0xγx+yγy=0. For the family f ( x , y ) = y + h ( x ) f ( x , y ) = y + h ( x ) f(x,y)=y+h(x)f(x, y)=y+h(x)f(x,y)=y+h(x) the corresponding γ γ gamma\gammaγ is given by γ = 1 h ( x ) γ = 1 h ( x ) gamma=(1)/(h^(')(x))\gamma=\frac{1}{h^{\prime}(x)}γ=1h(x) and satisfies the equation γ y = 0 γ y = 0 gamma_(y)=0\gamma_{y}=0γy=0.
More generally, we can suppose that we have additional information on the family of curves (2.1) given as a linear first order differential equation which is satisfied by γ , a ( x , y ) γ x + b ( x , y ) γ y = 0 γ , a ( x , y ) γ x + b ( x , y ) γ y = 0 gamma,a(x,y)gamma_(x)+b(x,y)gamma_(y)=0\gamma, a(x, y) \gamma_{x}+b(x, y) \gamma_{y}=0γ,a(x,y)γx+b(x,y)γy=0. In this case, as in the special cases mentioned above, if the potential satisfies a differential condition, the family γ γ gamma\gammaγ can be obtained as a common solution of two polynomial equations of degree at most seven, respectively twelve (Bozis et al, 2002). The coefficients of the polynomials in γ γ gamma\gammaγ are expressions containing the derivatives of V V VVV up to the forth order, and can be calculated using symbolic algebra programs.
Example 4.1 For the Hénon-Heiles potential V ( x , y ) = 1 2 x 2 + 8 y 2 + x 2 y + 16 3 y 3 V ( x , y ) = 1 2 x 2 + 8 y 2 + x 2 y + 16 3 y 3 V(x,y)=(1)/(2)x^(2)+8y^(2)+x^(2)y+(16)/(3)y^(3)V(x, y)=\frac{1}{2} x^{2}+8 y^{2}+ x^{2} y+\frac{16}{3} y^{3}V(x,y)=12x2+8y2+x2y+163y3 and a ( x , y ) = x , b ( x , y ) = y a ( x , y ) = x , b ( x , y ) = y a(x,y)=x,b(x,y)=ya(x, y)=x, b(x, y)=ya(x,y)=x,b(x,y)=y, the solution γ = x 4 y γ = x 4 y gamma=-(x)/(4y)\gamma=-\frac{x}{4 y}γ=x4y corresponding to the family y x 4 = c y x 4 = c yx^(-4)=cy x^{-4}=cyx4=c was found by Bozis and al (1997) as an example of a homogeneous family traced under the action of an inhomogeneous potential. The energy on the family is given by E = 1 24 c E = 1 24 c E=-(1)/(24 c)E=-\frac{1}{24 c}E=124c and the allowed region is ( x 2 + 8 y 2 + 12 y ) y 0 x 2 + 8 y 2 + 12 y y 0 (x^(2)+8y^(2)+12 y)y <= 0\left(x^{2}+8 y^{2}+12 y\right) y \leq 0(x2+8y2+12y)y0.

5. Programmed motion

It was proved in section 2 that during the motion of a material point of unit mass along an orbit of the family (2.1), the inequality
(5.16) B ( x , y ) 0 (5.16) B ( x , y ) 0 {:(5.16)B(x","y) >= 0:}\begin{equation*} B(x, y) \geq 0 \tag{5.16} \end{equation*}(5.16)B(x,y)0
must be observed, with
(5.17) B = E ( f ( x , y ) ) V ( x , y ) (5.17) B = E ( f ( x , y ) ) V ( x , y ) {:(5.17)B=E(f(x","y))-V(x","y):}\begin{equation*} B=E(f(x, y))-V(x, y) \tag{5.17} \end{equation*}(5.17)B=E(f(x,y))V(x,y)
This means that the motion is allowed along those members of the family (2.1) which are lying only inside some regions of the x y x y xyx yxy plane. The function B ( x , y ) B ( x , y ) B(x,y)B(x, y)B(x,y) is the kinetic energy of the material point of unit mass, as it moves on any of the orbits in the presence of the potential V ( x , y ) V ( x , y ) V(x,y)V(x, y)V(x,y).
According to Galliulin (1984), dynamical systems with programmed motion "are solved in such a way that the process occurring in these systems satisfies some preset requirements". The requirement that the motion takes place in the region (5.16) was considered at first by Bozis (1996). This type of programmed motion is the following: Given a preassigned region in the plane defined by (5.16), find a potential V V VVV which produces as trajectories of (2.2) the curves in the family (2.1).
Anisiu and Bozis (2000) studied this problem completely for a simpler family of functions, given by
(5.18) f ( x , y ) = y h ( x ) = c (5.18) f ( x , y ) = y h ( x ) = c {:(5.18)f(x","y)=y-h(x)=c:}\begin{equation*} f(x, y)=y-h(x)=c \tag{5.18} \end{equation*}(5.18)f(x,y)=yh(x)=c
where h h hhh is a nonlinear ( h ( x ) 0 h ( x ) 0 h^('')(x)!=0h^{\prime \prime}(x) \neq 0h(x)0 ) function of x x xxx. A simpler function b b bbb, which is nonnegative if and only if B B BBB is nonnegative, was considered. Under certain conditions on b b bbb, the function h h hhh (and, consequently, the family f f fff ) as well as the energy dependence function E ( f ) E ( f ) E(f)E(f)E(f) and the potential V ( x , y ) V ( x , y ) V(x,y)V(x, y)V(x,y) were determined. Let us denote
(5.19) b = B / ( 1 + γ 2 ) (5.19) b = B / 1 + γ 2 {:(5.19)b=B//(1+gamma^(2)):}\begin{equation*} b=B /\left(1+\gamma^{2}\right) \tag{5.19} \end{equation*}(5.19)b=B/(1+γ2)
The function b b bbb satisfies a second order partial differential equation for the general family of functions (2.1). This was derived at first by Bozis (1995) starting from the case of nonconservative forces. Here we shall obtain it using Szebehely's equation.
Theorem 5.1 Denoting κ = 1 γ γ , m = 2 γ y + γ x γ , n = 2 γ x 3 γ y γ κ = 1 γ γ , m = 2 γ y + γ x γ , n = 2 γ x 3 γ y γ kappa=(1)/(gamma)-gamma,quad m=2gamma_(y)+(gamma_(x))/(gamma),n=2gamma_(x)-(3gamma_(y))/(gamma)\kappa=\frac{1}{\gamma}-\gamma, \quad m=2 \gamma_{y}+\frac{\gamma_{x}}{\gamma}, n=2 \gamma_{x}-\frac{3 \gamma_{y}}{\gamma}κ=1γγ,m=2γy+γxγ,n=2γx3γyγ, p = 2 Γ y γ p = 2 Γ y γ p=(2Gamma_(y))/(gamma)p=\frac{2 \Gamma_{y}}{\gamma}p=2Γyγ, the function b b bbb satisfies the linear second order partial differential equation (Bozis, 1995)
(5.20) b x x + κ b x y + b y y = m b x + n b y + p b . (5.20) b x x + κ b x y + b y y = m b x + n b y + p b . {:(5.20)-b_(xx)+kappab_(xy)+b_(yy)=mb_(x)+nb_(y)+pb.:}\begin{equation*} -b_{x x}+\kappa b_{x y}+b_{y y}=m b_{x}+n b_{y}+p b . \tag{5.20} \end{equation*}(5.20)bxx+κbxy+byy=mbx+nby+pb.
Proof. In view of (5.17) and (5.19), Szebehely's equation (2.7) becomes
(5.21) V x + γ V y = 2 Γ b (5.21) V x + γ V y = 2 Γ b {:(5.21)V_(x)+gammaV_(y)=-2Gamma b:}\begin{equation*} V_{x}+\gamma V_{y}=-2 \Gamma b \tag{5.21} \end{equation*}(5.21)Vx+γVy=2Γb
and E = V + ( 1 + γ 2 ) b E = V + 1 + γ 2 b E=V+(1+gamma^(2))bE=V+\left(1+\gamma^{2}\right) bE=V+(1+γ2)b. But γ = E y E x γ = E y E x gamma=(E_(y))/(E_(x))\gamma=\frac{E_{y}}{E_{x}}γ=EyEx, so we obtain γ = V y + ( ( 1 + γ 2 ) b ) y V x + ( ( 1 + γ 2 ) b ) x γ = V y + 1 + γ 2 b y V x + 1 + γ 2 b x gamma=(V_(y)+((1+gamma^(2))b)_(y))/(V_(x)+((1+gamma^(2))b)_(x))\gamma=\frac{V_{y}+\left(\left(1+\gamma^{2}\right) b\right)_{y}}{V_{x}+\left(\left(1+\gamma^{2}\right) b\right)_{x}}γ=Vy+((1+γ2)b)yVx+((1+γ2)b)x, which can be written as
(5.22) γ V x V y = ( 1 + γ 2 ) ( b y γ b x ) 2 γ Γ b . (5.22) γ V x V y = 1 + γ 2 b y γ b x 2 γ Γ b . {:(5.22)gammaV_(x)-V_(y)=(1+gamma^(2))(b_(y)-gammab_(x))-2gamma Gamma b.:}\begin{equation*} \gamma V_{x}-V_{y}=\left(1+\gamma^{2}\right)\left(b_{y}-\gamma b_{x}\right)-2 \gamma \Gamma b . \tag{5.22} \end{equation*}(5.22)γVxVy=(1+γ2)(byγbx)2γΓb.
Solving the system of two equations (5.21) and (5.22) we get V x = γ ( b y γ b x ) 2 Γ b V y = γ b x b y V x = γ b y γ b x 2 Γ b V y = γ b x b y V_(x)=gamma(b_(y)-gammab_(x))-2Gamma b quadV_(y)=gammab_(x)-b_(y)V_{x}= \gamma\left(b_{y}-\gamma b_{x}\right)-2 \Gamma b \quad V_{y}=\gamma b_{x}-b_{y}Vx=γ(byγbx)2ΓbVy=γbxby. Writing the compatibility condition ( V x ) y = ( V y ) x V x y = V y x (V_(x))_(y)=(V_(y))_(x)\left(V_{x}\right)_{y}=\left(V_{y}\right)_{x}(Vx)y=(Vy)x we obtain (5.20).
We can rearrange (5.20) as a nonlinear equation of order two in γ γ gamma\gammaγ
2 b γ γ x y 2 b γ y y = 2 b γ x γ y ( 2 b y γ + b x ) γ x (5.23) ( 2 b x γ 3 b y ) γ y b x y γ 2 + ( b y y b x x ) γ + b x y 2 b γ γ x y 2 b γ y y = 2 b γ x γ y 2 b y γ + b x γ x (5.23) 2 b x γ 3 b y γ y b x y γ 2 + b y y b x x γ + b x y {:[2b gammagamma_(xy)-2bgamma_(yy)=-2bgamma_(x)gamma_(y)-(2b_(y)gamma+b_(x))gamma_(x)],[(5.23)-(2b_(x)gamma-3b_(y))gamma_(y)-b_(xy)gamma^(2)+(b_(yy)-b_(xx))gamma+b_(xy)]:}\begin{gather*} 2 b \gamma \gamma_{x y}-2 b \gamma_{y y}=-2 b \gamma_{x} \gamma_{y}-\left(2 b_{y} \gamma+b_{x}\right) \gamma_{x} \\ -\left(2 b_{x} \gamma-3 b_{y}\right) \gamma_{y}-b_{x y} \gamma^{2}+\left(b_{y y}-b_{x x}\right) \gamma+b_{x y} \tag{5.23} \end{gather*}2bγγxy2bγyy=2bγxγy(2byγ+bx)γx(5.23)(2bxγ3by)γybxyγ2+(byybxx)γ+bxy
which can be used in the case when the function b b bbb is known.
Considering the kinetic energy B = 1 B = 1 B=1B=1B=1 (or, equivalently, B = B = B=B=B= const) for all the curves of the family (2.1), equation (5.20) will become
(5.24) ( 1 + γ 2 ) ( γ 2 γ x x 2 γ γ x y + γ y y ) + 2 γ ( 1 γ 2 ) γ x 2 4 γ γ y 2 + 2 ( 3 γ 2 1 ) γ x γ y = 0 (5.24) 1 + γ 2 γ 2 γ x x 2 γ γ x y + γ y y + 2 γ 1 γ 2 γ x 2 4 γ γ y 2 + 2 3 γ 2 1 γ x γ y = 0 {:(5.24)(1+gamma^(2))(gamma^(2)gamma_(xx)-2gammagamma_(xy)+gamma_(yy))+2gamma(1-gamma^(2))gamma_(x)^(2)-4gammagamma_(y)^(2)+2(3gamma^(2)-1)gamma_(x)gamma_(y)=0:}\begin{equation*} \left(1+\gamma^{2}\right)\left(\gamma^{2} \gamma_{x x}-2 \gamma \gamma_{x y}+\gamma_{y y}\right)+2 \gamma\left(1-\gamma^{2}\right) \gamma_{x}^{2}-4 \gamma \gamma_{y}^{2}+2\left(3 \gamma^{2}-1\right) \gamma_{x} \gamma_{y}=0 \tag{5.24} \end{equation*}(5.24)(1+γ2)(γ2γxx2γγxy+γyy)+2γ(1γ2)γx24γγy2+2(3γ21)γxγy=0
which is the equation giving the totality of isotach orbits (Bozis, 1986).
For the special case of the family (5.18), the function z z zzz in Anisiu and Bozis (2000) is equal to γ γ -gamma-\gammaγ, and equation (5.20) becomes
( b x 2 b y z ) z = b x y z 2 + ( b y y b x x ) z b x y b x 2 b y z z = b x y z 2 + b y y b x x z b x y (b_(x)-2b_(y)z)z^(')=b_(xy)z^(2)+(b_(yy)-b_(xx))z-b_(xy)\left(b_{x}-2 b_{y} z\right) z^{\prime}=b_{x y} z^{2}+\left(b_{y y}-b_{x x}\right) z-b_{x y}(bx2byz)z=bxyz2+(byybxx)zbxy
which was derived there directly. The method presented in the cited paper gives the family of orbits, the energy and the potential, when a suitable function b b bbb defines the allowed region.
Example 5.1 For x > 0 x > 0 x > 0x>0x>0 and the function b = b 2 y 2 + b 1 y + b 0 b = b 2 y 2 + b 1 y + b 0 b=b_(2)y^(2)+b_(1)y+b_(0)b=b_{2} y^{2}+b_{1} y+b_{0}b=b2y2+b1y+b0, with b 2 = 3 / x , b 1 = 3 ( x 4 + 1 ) / x 2 , b 0 = ( 3 x 4 ( x 4 / 5 + 1 ) + 1 ) / x 3 b 2 = 3 / x , b 1 = 3 x 4 + 1 / x 2 , b 0 = 3 x 4 x 4 / 5 + 1 + 1 / x 3 b_(2)=-3//x,b_(1)=3(x^(4)+1)//x^(2),quadb_(0)=-(3x^(4)(x^(4)//5+1)+1)//x^(3)b_{2}=-3 / x, b_{1}=3\left(x^{4}+1\right) / x^{2}, \quad b_{0}=-\left(3 x^{4}\left(x^{4} / 5+1\right)+1\right) / x^{3}b2=3/x,b1=3(x4+1)/x2,b0=(3x4(x4/5+1)+1)/x3, the family of orbits of the form (5.18) y 1 x = c y 1 x = c y-(1)/(x)=cy-\frac{1}{x}=cy1x=c is described inside the region b ( x , y ) 0 b ( x , y ) 0 b(x,y) >= 0b(x, y) \geq 0b(x,y)0. In this case, E = c 3 E = c 3 E=c^(3)E=c^{3}E=c3 and the potential is V ( x , y ) = y 3 + 3 x 3 y 2 3 x 2 ( x 4 + 2 ) y + 1 5 x ( 3 x 8 + 18 x 4 + 20 ) V ( x , y ) = y 3 + 3 x 3 y 2 3 x 2 x 4 + 2 y + 1 5 x 3 x 8 + 18 x 4 + 20 V(x,y)=y^(3)+3x^(3)y^(2)-3x^(2)(x^(4)+2)y+(1)/(5)x(3x^(8)+18x^(4)+20)V(x, y)= y^{3}+3 x^{3} y^{2}-3 x^{2}\left(x^{4}+2\right) y+\frac{1}{5} x\left(3 x^{8}+18 x^{4}+20\right)V(x,y)=y3+3x3y23x2(x4+2)y+15x(3x8+18x4+20).

6. Final remarks

PDEs appear in connection with the inverse problem of dynamics; (2.7) is quasilinear and (2.9) is linear in the potential function V V VVV. The direct problem gives rise to the nonlinear equation (4.15). Other PDEs are produced by related problems: the study of potentials creating families of straight lines leads to equation (3.14), the programming of the
motion in certain regions of the plane to (5.20), (5.23), and the study of families traced with constant kinetic energy to (5.24).
Acknowledgement. This research was partially supported by the Ministry of Education and Research, by grant 343-CNCSIS 33444/2002.

References

[1] Anisiu, M.-C. : 1998, Nonlinear Analysis Methods with Application in Celestial Mechanics, University Press, Cluj-Napoca (in Romanian)
[2] Anisiu, M.-C. and Bozis, G. : 2000, Programmed motion for a class of families of planar orbits, Inverse Problems 16, 19-32
[3] Bozis, G. : 1983, Inverse problem with two-parametric families of planar orbits, Celest. Mech. 31, 129-143
[4] Bozis, G. : 1984, Szebehely's inverse problem for finite symmetrical material concentrations, Astronom. Astrophys. 134, 360-364
[5] Bozis, G. : 1986, Adelphic potentials, Astron. Astrophys. 160, 107-110
[6] Bozis, G. and Grigoriadou, S. : 1993, Families of planar orbits generated by homogeneous potentials, Celest. Mech. 57(3), 461-472
[7] Bozis,G. and Stefiades, Ap. : 1993, Geometrically similar orbits in homogeneous potentials, Inverse Problems 9(2), 233-240
[8] Bozis, G. and Ichtiaroglou, S.: 1994, Boundary curves for families of planar orbits, Celest. Mech. 58, 371-385
[9] Bozis, G. : 1995, The inverse problem of dynamics: basic facts, Inverse Problems 11, 687-708
[10] Bozis, G.: 1996, Two-dimensional programmed motion, Proceedings of the 2nd Hellenic Astronomical Conference, Thessaloniki, June 29-July 1, 1995 (eds. M. E. Contadakis et al), pp. 587-590
[11] Bozis, G., Anisiu, M.-C. and Blaga, C. : 1997, Inhomogeneous potentials producing homogeneous orbits, Astron. Nachr. 318, 313-318
[12] Bozis, G., Anisiu, M.-C. and Blaga, C. : 2000, A solvable version of the direct problem of dynamics, Rom. Astronom. J. 10(1), 59-70
[13] Bozis, G. and Anisiu, M.-C. : 2001, Families of straight lines in planar potentials, Rom. Astronom. J. 11(1), 27-43
[14] Bozis, G., Anisiu, M.-C. and Blaga, C. : 2002, Special families of orbits in the direct problem of dynamics, preprint
[15] Broucke, R. and Lass, H. : 1977, On Szebehely's equation for the potential of a prescribed family of orbits, Celest. Mech. 16, 215-225
[16] Galiullin, A. S. : 1984, Inverse Problem of Dynamics, p. 91, Mir Publishers, Moscow
[17] Grigoriadou, S., Bozis, G. and Elmabsout, B.: 1999, Solvable cases of Szebehely's equation, Celest. Mech. 74, 211-221
[18] Newton, I. : 1687, Philosophiae Naturalis Principia Mathematica, London
[19] Szebehely, V. : 1974, On the determination of the potential by satellite observations, in E. Proverbio (ed.) Proc. Intern. Meeting on Earth's Rotations by Satellite Observations, The University of Cagliari, Bologna, Italy, 31-35.
2003

Related Posts