Access the full text.

Sign up today, get DeepDyve free for 14 days.

Nonlinear Sciences
, Volume 2018 (1801) – Jan 2, 2018

/lp/arxiv-cornell-university/basins-of-convergence-of-equilibrium-points-in-the-generalized-hill-2c4fLoVlOa

- ISSN
- 0218-1274
- eISSN
- ARCH-3338
- DOI
- 10.1142/S0218127417300439
- Publisher site
- See Article on Publisher Site

January 3, 2018 2:3 IJBC_01 International Journal of Bifurcation and Chaos c World Scienti c Publishing Company Basins of Convergence of Equilibrium Points in the Generalized Hill Problem Euaggelos E. Zotos Department of Physics, School of Science, Aristotle University of Thessaloniki, GR-541 24, Thessaloniki, Greece Corresponding author's email: evzotos@physics.auth.gr Received Received August 19, 2017; Revised October 10, 2017 The Newton-Raphson basins of attraction, associated with the libration points (attractors), are revealed in the generalized Hill problem. The parametric variation of the position and the linear stability of the equilibrium points is determined, when the value of the perturbation parameter varies. The multivariate Newton-Raphson iterative scheme is used to determine the attracting domains on several types of two-dimensional planes. A systematic and thorough numerical investigation is performed in order to demonstrate the in uence of the perturbation parameter on the geometry as well as of the basin entropy of the basins of convergence. The correlations between the basins of attraction and the corresponding required number of iterations are also illustrated and discussed. Our numerical analysis strongly indicates that the evolution of the attracting regions in this dynamical system is an extremely complicated yet very interesting issue. Keywords : Generalized Hill problem, Equilibrium points, Basins of attraction, Fractal basins boundaries, Basin entropy arXiv:1801.00710v1 [nlin.CD] 2 Jan 2018 January 3, 2018 2:3 IJBC_01 2 E.E. Zotos 1. Introduction 2013]. In this paper we shall use a generalized form Undoubtedly, one of the most intriguing as well as of the classical Hill problem in order to determine important elds in dynamical astronomy and celes- the equilibrium points and the associated basins of tial mechanics is the few-body problem and espe- attraction. The multivariate version of the Newton- cially the version of the circular restricted three- Raphson iterative scheme will be used for revealing body problem [Szebehely, 1967]. This is true if we the basins of convergence on several types of two- take into account that this problem has numerous dimensional planes. applications in many research elds, such as molecu- The present article has the following structure: lar physics, chaos theory, planetary physics, or even the most important properties of the dynamical sys- stellar and galactic dynamics. This is exactly why tem are presented in Section 2. The parametric evo- this topic remains active and stimulating even to- lution of the position as well as of the stability of day. the equilibrium points is investigated in Section 3. The Hill limiting case is in fact a simpli ed The following Section contains the main numeri- modi cation of the three-body problem which focus cal results, regarding the evolution of the Newton- on the vicinity of the secondary (e.g., [Hill, 1886; Pe- Raphson basins of convergence. In Section 5 we tit & H enon, 1986, 1987]). This allows us to study demonstrate how the basin entropy of the con g- the motion of the test particles in the neighborhood uration (x; y) convergence planes evolves as a func- of the equilibrium points L and L . At this point, 1 2 tion of the perturbation parameter. Our paper ends it should be emphasized that the Hill approxima- with Section 6, where we emphasize the main con- tion is valid only when the mass of the secondary clusions of this work. is much smaller compared with the mass of the pri- mary body (m m ). One can directly obtain the 2 1 2. Description of the dynamical Hill model from the classical three-body problem by translating the origin to the center of the secondary system and also by rescaling the coordinates by a factor The classical Hill problem is derived for the re- 1=3 , where = m =(m + m ) is the mass ratio. 2 1 2 stricted three-body problem when the mass of the Knowing the basins of convergence, associated secondary body is substantially smaller than that with the libration points, is an issue of great im- of the primary (e.g., the Sun-Earth system). In the portance, since the attracting domains re ect some vicinity of the secondary the corresponding poten- of the most intrinsic properties of the dynami- tial, of the planar model, is given by cal system. For obtaining the basins of attrac- 3 1 tion one should use an iterative scheme (e.g., the V (x; y) = x + ; (1) 2 r Newton-Raphson method) and scan a set of ini- 2 2 tial conditions in order to reveal their nal states where of course r = x + y . (attractors). Over the past years a large number According to [Kozlov & Polekhin, 2017], Eq. (1) of studies have been devoted on determining the can be generalized, in a straightforward manner, as Newton-Raphson basins of convergence in many follows types of dynamical systems, such as the Hill's prob- 1 1 y 2 2 2 V (x; y) = x + y + + x : (2) lem [Douskos, 2010], the Sitnikov problem [Douskos 2 r 2 et al., 2012], the restricted three-body problem with We observe that when = 1 the potential (2) is oblateness and radiation pressure [Zotos, 2016], the reduced to the classical form of Eq. (1). electromagnetic Copenhagen problem [Kalvouridis It should be noted that the potential of the gen- & Gousidou-Koutita, 2012; Zotos, 2017b], the pho- eralized Hill problem is composed of very interesting togravitational Copenhagen problem [Kalvouridis, terms. In particular, the rst term is an isotropic 2008], the four-body problem [Baltagiannis & Pa- harmonic oscillator, the second term is a repulsive padakis, 2011; Kumari & Kushvah, 2014; Zotos, Coulomb potential, while the third term breaks the 2017a], the photogravitational four-body problem rotational symmetry of the oscillator. For dierent [Asique et al., 2016], the ring problem of N + 1 bod- from zero, but still smaller than 1, the rst and the ies [Croustalloudi & Kalvouridis, 2007; Gousidou- third parts together form an anisotropic harmonic Koutita & Kalvouridis, 2009], or even the restricted oscillator. We could say that is the perturbation 2+2 body problem [Croustalloudi & Kalvouridis, parameter for the rotational symmetry. This is true January 3, 2018 2:3 IJBC_01 Basins of convergence of equilibrium points in the generalized Hill problem 3 because for approaching 1 from below the system When 2 (0; 1) there are four equilibrium points. changes its qualitative character and it is no longer Two of them, L and L , are located on the x-axis, 1 2 a harmonic oscillator with a repulsive Coulomb po- while the other two, L and L , are located on the 3 4 tential in the middle. On the other hand, for = 1, vertical y-axis. or larger, it becomes unbound, for suciently large When 1 there are only two real equilibrium energy levels. points on the horizontal x-axis. The equations describing the motion of the test In both cases the equilibrium points L and L are particle, in the corotating frame of reference, read 1 2 located at (x ; 0), while the coordinates of L and L 3 x = V + 2y _; L are (0;y ). y = V 2x; _ (3) It would be very interesting to obtain the exact evolution of the coordinates of the libration points where as a function of the perturbation parameter , when @V 1 V = = x 1 + 2 ; > 0. Our numerical analysis is illustrated in Fig. @x r 1(a-b), where the parametric evolution of x and @V 1 y is given as a function of . It is seen that when V = = y 1 : (4) @y r ! 0 both x and y tend to 1. However, as the L L Similarly, the partial derivatives of the second value of the perturbation increases the coordinates order, which will be needed later for the multivari- follow a dierent path. More precisely, x is reduced ate Newton-Raphson iterative scheme, read and tends asymptotically to zero, when ! 1. On 2 2 the other hand, y tends asymptotically to in nity, @ V 1 3x V = = 1 + + 2; xx when ! 1. 2 3 5 @x r r In order to determine the linear stability of an @ V 3xy equilibrium point the origin of the reference frame V = = ; xy @x@y r must be transferred at the exact position (x ; y ) of 0 0 @ V the libration point through the transformation V = = V ; yx xy @y@x 2 2 x = x + ; y = y + : (9) 0 0 @ V 1 3y V = = 1 + : (5) yy 2 3 5 @y r r The next step is to expand the system of the equa- The total orbital energy of the system is pre- tions of motion (3) into rst-order terms, with re- served, according to the Jacobi integral of motion spect to and . 2 2 J (x; y; x; _ y _ ) = 2V (x; y) x _ + y _ = ; (6) _ _ = A; = ; ; ; _ ; (10) where x _ and y _ are the velocities, while is the nu- merical value of the Jacobi constant which is con- where is the state vector of the test particle with served. respect to the equilibrium points, while A is the time-independent coecient matrix of variations 3. Parametric evolution of the 2 3 equilibrium points 0 0 1 0 6 7 0 0 0 1 For the existence of equilibrium points the necessary 6 7 A = ; (11) 0 0 4 5 V V 0 2 and sucient conditions which must be ful lled are xx xy 0 0 V V 2 0 yx yy x _ = y _ = x = y = 0: (7) where the superscript 0, at the partial derivatives For determining the coordinates (x; y) of the copla- of second order, denotes evaluation at the position nar equilibrium points we have to numerically solve of the equilibrium point (x ; y ). The new linearized the following system of equations 0 0 system describes in nitesimal motions near an equi- V (x; y) = 0; V (x; y) = 0: (8) x y librium point. The total number of the equilibrium points in The characteristic equation of the linear system the generalized Hill problem in not constant but it (10) is strongly depends on the value of the perturbation 4 2 parameter . More precisely + b + c = 0; (12) January 3, 2018 2:3 IJBC_01 4 E.E. Zotos Fig. 1. The parametric evolution of (a-left): x (b-right): y , of the equilibrium points in the generalized Hill problem, when L L 2 (0; 100]. Note that the vertical axis in panel (b) displays the common logarithm of y . where This fact ensures that equation (14) has two real negative roots , which consequently implies that 1;2 = 1; there are four pure imaginary roots for . 0 0 b = 4 V V ; xx yy Since we already know the exact positions 0 0 0 0 (x ; y ) of the libration points, we can insert them c = V V V V : (13) 0 0 xx yy xy yx into the characteristic equation (12) and therefore It is seen that equation (12) is quadratic with re- determine the linear stability of the equilibrium spect to = and therefore it can be written points, through the nature of the four roots. In the as interval 2 (0; 10 ] we de ned a uniform sequence + b + c = 0: (14) of 10 values of the perturbation parameter . Then for these values of we numerically solved the sys- The necessary and sucient condition for an tem (8) thus computing the coordinates (x ; y ) of 0 0 equilibrium point to be linearly stable is all four the equilibrium points. The last step was to insert roots of the characteristic equation (12) to be pure the coordinates of the equilibria into the character- imaginary . This means that the following three istic equation (12) and determine the nature of the conditions must be simultaneously ful lled four roots. The above-mentioned numerical analy- b > 0; c > 0; D = b 4ac > 0: (15) sis suggests that for all the equilibrium points the characteristic equation (12) has always, at least, two In Hamiltonian (symplectic) dynamics the monodromy ma- complex roots with non zero (positive or negative) trix is symplectic and this restricts the eigenvalues strongly. real part. Therefore we conclude that all the equi- Then the product of all eigenvalues must be equal to 1. In librium points, L , i = 1; :::; 4, are linearly unstable, addition: if is an eigenvalue, then also - and the com- when > 0. plex conjugate of and the complex conjugate of - must be eigenvalues. For the eigenplanes there are the following possibilities: 2-dimensional hyperbolic planes (with two real 4. The basins of attraction eigenvalues and - ), 2-dimensional elliptic planes (with two imaginary eigenvalues and - ) and 4-dimensional planes Over the years, many methods for solving numeri- of complex spiralling behaviour or complex instability (with cally systems of non-linear equations have been de- four eigenvalues with the properties as mentioned above). veloped. Perhaps the most well-known method of all Here it should be noted that asymptotic stability, in the sense is the Newton-Raphson method. A system of mul- of Lyapunov, is not possible for Hamiltonian systems. There- fore, usually elliptic behaviour of Hamiltonian systems is con- tivariate functions f (x) = 0 can be solved using the sidered stable. However, in the more general sense this stabil- following iterative scheme ity is not asymptotic stability. It is only marginal or neutral stability, which means that generally speaking trajectories do x = x J f (x ); (16) n+1 n n not disappear exponentially. More details, regarding the lin- where f (x ) is the system of equations, while J is ear stability in Hamiltonian systems, can be found in chapter n 3 of [Abraham & Marsden, 1987]. the corresponding inverse Jacobian matrix. In our January 3, 2018 2:3 IJBC_01 Basins of convergence of equilibrium points in the generalized Hill problem 5 case the system of equations is described in Eqs. tions we also keep records of the number N of itera- (8). tions, required for the desired accuracy. Obviously, The iterative formulae for each coordinate the better the desired accuracy, the higher the re- (x; y), derived from scheme (16), are quired iterations. In our calculations the maximum number of iterations is set to N = 500, while max V V V V x yy y xy the iterative procedure stops only when an accu- x = x ; n+1 n V V V yy xx xy racy of 10 is reached, regarding the position of (x ;y ) n n the attractors. V V V V x yx y xx y = y + ; (17) n+1 n In what follows we will try to determine how V V V yy xx xy (x ;y ) n n the perturbation parameter in uences the struc- ture of the Newton-Raphson basins of attraction where x , y are the values of the x and y coordi- n n in the generalized Hill problem, by considering two nates at the n-th step of the iterative process. cases regarding the total number of the equilibrium The numerical algorithm of the Newton- points (attractors). For classifying the initial condi- Raphson method works as follows: The code is acti- tions on the con guration (x; y) plane we will use vated when an initial condition (x ; y ) on the con- 0 0 color-coded diagrams (CCDs), where each pixel is guration plane is inserted, while the iterative pro- assigned a color, according to the nal state (at- cedure continues until an attractor of the system tractor) of the initial condition. is reached, with the desired accuracy. If the itera- tive procedure leads to one of the attractors then we say that the method converges for the partic- 4.1. Case I: Four equilibrium points ular initial condition. However, in general terms, Our numerical exploration begins with the rst case not all initial conditions converge to an attractor of where four equilibrium points are present, that is the system. All the initial conditions that lead to a when 0 < < 1. In Fig. 2 we present the evolu- speci c nal state (attractor) compose the Newton- tion of the basins of convergence for six values of Raphson basins of attraction, which are also known the perturbation parameter . In panel (a) where as basins of convergence or even as attracting re- = 0:001 we observe the existence of several ten- gions/domains. At this point, it should be highly tacles between the dierent basins of attraction. In noticed that the Newton-Raphson basins of attrac- the vicinity of these tentacles we encounter the most tion should not be mistaken, by no means, with the highly fractal areas of the con guration (x; y) plane. classical basins of attraction which exist in the case At this point it should be noted that when we claim of dissipative systems. The Newton-Raphson basins that a region is fractal we simply mean that it has a of attraction are just a numerical artifact produced fractal-like geometry, without conducting, at least by an iterative scheme, while on the other hand for now, any additional quantitative calculations as the basins of attraction in dissipative systems corre- in [Aguirre et al., 2001]. As we proceed to higher spond to a real observed phenomenon (attraction). values of the perturbation parameter three impor- Nevertheless, the determination of the Newton- tant phenomena take place Raphson basins of attraction is very important be- cause they re ect some of the most intrinsic qual- (1) The area on the con guration (x; y) plane, oc- itative properties of the dynamical system. This is cupied by the tentacle-like structures, is re- true because the iterative formulae of Eqs. (17) con- duced. tain both the rst and second order derivatives of (2) Well formed basins of convergence emerge. In the eective potential function V (x; y). particular, the attracting domains correspond- In order to unveil the basins of convergence we ing to the libration points L and L seem to 3 4 have to perform a double scan of the con gura- dominate, while the basins of convergence cor- tion (x; y) plane. For this purpose we de ne uni- responding to equilibrium points L and L are 1 2 form grids of 1024 1024 (x ; y ) nodes which shall 0 0 mainly con ned near the horizontal axis. be used as initial conditions of the numerical algo- (3) The area of the fractal regions on the con gu- rithm. Of course the initial condition (0; 0) is ex- ration space is also reduced, thus increasing the cluded from all grids, because for this initial condi- predictability regarding the nal state (attrac- tion the distance r is equal to zero and consequently tor) of the initial conditions. several terms, entering formulae (17), become sin- gular. During the classi cation of the initial condi- When tends to 1, it is seen in panel (f ) that al- January 3, 2018 2:3 IJBC_01 6 E.E. Zotos Fig. 2. The Newton-Raphson basins of attraction on the con guration (x; y) plane when four equilibrium points are present. (a): = 0:001; (b): = 0:1; (c): = 0:6; (d): = 0:9; (e): = 0:99; (f ): = 0:999. The positions of the four equilibrium points are indicated by black dots. The color code, denoting the four attractors is as follows: L (green); L (red); L (blue); L 1 2 3 4 (magenta); non-converging points (white). January 3, 2018 2:3 IJBC_01 Basins of convergence of equilibrium points in the generalized Hill problem 7 Fig. 3. The distribution of the corresponding number N of required iterations for obtaining the Newton-Raphson basins of attraction shown in Fig. 2(a-f ). The non-converging points are shown in white. January 3, 2018 2:3 IJBC_01 8 E.E. Zotos Fig. 4. The corresponding probability distribution of required iterations for obtaining the Newton-Raphson basins of attrac- tion shown in Fig. 2(a-f ). The vertical dashed red line indicates, in each case, the most probable number N of iterations. The blue line is the best t for the right-hand side (N > N ) of the histograms, using a Laplace probability distribution function. January 3, 2018 2:3 IJBC_01 Basins of convergence of equilibrium points in the generalized Hill problem 9 Fig. 5. The Newton-Raphson basins of attraction on the con guration (x; y) plane when only two equilibrium points are present. (a): = 1:0; (b): = 1:1; (c): = 1:5; (d): = 2:0; (e): = 3:0; (f ): = 5:0. The positions of the two equilibrium points are indicated by black dots. The color code, denoting the four attractors is as follows: L (green); L (red); converging 1 2 points to in nity (yellow); non-converging points (white). January 3, 2018 2:3 IJBC_01 10 E.E. Zotos Fig. 6. The distribution of the corresponding number N of required iterations for obtaining the Newton-Raphson basins of attraction shown in Fig. 5(a-f ). The non-converging points, as well as those which converge to in nity, are shown in white. January 3, 2018 2:3 IJBC_01 Basins of convergence of equilibrium points in the generalized Hill problem 11 Fig. 7. The corresponding probability distribution of required iterations for obtaining the Newton-Raphson basins of attrac- tion shown in Fig. 5(a-f ). The vertical dashed red line indicates, in each case, the most probable number N of iterations. The blue line is the best t for the right-hand side (N > N ) of the histograms, using a Laplace probability distribution function. January 3, 2018 2:3 IJBC_01 12 E.E. Zotos most the entire (x; y) plane is covered by basins of attraction corresponding to libration points L and L . On the other hand, the attracting regions as- sociated with the equilibrium points L and L are 1 2 con ned mainly in the vicinity of the corresponding libration points. The distribution of the corresponding number N of iterations is provided, using tones of blue, in Fig. 3(a-f ). It is observed that initial conditions in- side the attracting regions converge relatively fast, while the slowest converging points are those in the vicinity of the basin boundaries. In particular, the slowest converging points are encountered in the boundaries of either the tentacles or the gure-eight Fig. 8. Evolution of the percentage of non-converging initial conditions, as a function of the perturbation parameter , structures observed when > 0:9. In Fig. 4(a-f ) the when 2 (1; 5]. corresponding probability distribution of iterations is given. The probability P is de ned as follows: if N initial conditions (x ; y ) converge to one of 0 0 0 the attractors, after N iterations, then P = N =N , 0 t where N is the total number of initial conditions in every CCD. With increasing value of the most probable number N of iterations is reduced from 23 when = 0:001 to 6 when = 0:999. The blue lines in the histograms of Fig. 4 indicate the best t to the right-hand side N > N of them (more details are given in subsection 4.3). 4.2. Case II: Two equilibrium points When 1 there are only two real equilib- Fig. 9. Evolution of the vertical coordinate y, as a function of the iterations N , when x = 0, y = 5, and = 5. Note the 0 0 rium points located on the horizontal x-axis. The chaotic uctuations as well as the random peaks throughout Newton-Raphson basins of attraction for six val- the range of iterations which clearly indicate non-convergence ues of the perturbation parameter are presented in of the Newton-Raphson iterative scheme. Fig. 5(a-f ). A very interesting behavior is unveiled in panel (a), where = 1. One may observe that play any sign of convergence even after a substantial the vast majority of the con guration (x; y) plane amount of iterations (N = 10000). is covered by initial conditions which converge to With increasing value of the perturbation parame- extremely large numbers, thus indicating conver- ter the fractal regions of the (x; y) plane are heav- gence to in nity. This phenomenon however is an- ily been reduced, while the areas where the unpre- ticipated. This is true because according to panel dictability is still high are mainly con ned near the (b) of Fig. 1 when = 1, y tends to in nity. On vertical y-axis, around the non-converging initial this basis, we may say that what we see in panel conditions. (a) of Fig. 5 is just a numerical con rmation of the theory. In Fig. 6(a-f ) we illustrate the distribution of As soon as > 1 the convergence properties of the corresponding number N of iterations required the (x; y) plane change drastically. More precisely: for obtaining the desired accuracy, while the cor- A portion of the con guration space is covered by responding probability distribution of iterations is initial conditions which do not converge to any of given in Fig. 7(a-f ). In this case, the most probable the two equilibrium points. Additional computa- number N of iteration starts at 8 for = 1 and tions suggest that for these initial conditions the then it increases up to 19 when = 5. multivariate Newton-Raphson scheme does not dis- Looking at the panels of Fig. 5 it becomes ev- January 3, 2018 2:3 IJBC_01 Basins of convergence of equilibrium points in the generalized Hill problem 13 ident that the amount of non-converging points is a two-dimensional plane in which the x or the y reduced with increasing value of the perturbation coordinate is the abscissa, while the value of is parameter. In Fig. 8 we provide the evolution of the always the ordinate. Panel (a) of Fig. 10 shows the percentage of the non-converging initial conditions attracting domains of the (x = y; ) plane, while as a function of . We clearly see that the reduction in panel (b) of the same gure the distribution of is very smooth and almost linear for 0 < < 3. For the corresponding number N of required iterations larger values of the perturbation parameter ( > 4) for obtaining the Newton-Raphson basins of attrac- the portion of the non-converging initial conditions tion is shown. In panel (a) of Fig. 10 it can be seen remains constant at about 0.22%, while for > 5 very clearly how the convergence properties of the all initial conditions which do not converge to any system change when = 1. At the same time we of the attractors (equilibrium points) lie completely observe how the tentacles emerge and divide the on the vertical y-axis, with x = 0. several basins of attraction. In [Zotos, 2017a], where we investigated the Additional interesting information could be ex- basins of attraction in the planar equilateral re- tracted from the probability distributions of itera- stricted four-body problem, we encountered the tions presented in Figs. 4, and 7. In particular, it phenomenon of slow convergence, that is when for would be very interesting to try to obtain the best an initial condition the Newton-Raphson iterative t of the tails of the distributions. For tting the scheme requires an extremely high number of iter- tails of the histograms, we used the Laplace distri- ations in order to converge to one of the attractors. bution, which is the most natural choice, since this Additional numerical computations strongly sug- type of distribution is very common in systems dis- gest that this is not the case in the generalized Hill playing transient chaos (e.g., [Motter & Lai, 2001; problem. To prove this we set the maximum allowed Seoane & Sanju an, 2008; Seoane et al., 2006]). Our number of iterations equal to 5000 and we repeated calculations strongly indicate that in the vast ma- the classi cation of the initial conditions. However jority of the cases the Laplace distribution is the we found that the percentage of non-converging ini- best t to our data. The only two cases where the tial conditions remains the same. In Fig. 9 we illus- Laplace distribution fails to properly t the corre- trate a characteristic example of the evolution of the sponding numerical data are the cases correspond- y coordinate of an non-converging initial condition ing to = 1 and = 3 (see panels (a) and (e) of with x = 0 and y = 5, when = 5. It is seen that Fig. 7, respectively). 0 0 the vertical coordinate randomly uctuates between The probability density function (PDF) of the negative and positive numbers, while displaying Laplace distribution is given by completely random peaks thus showing no indica- tion of convergence. The same pattern appears even aN after 50000 iterations which automatically leads to 1 exp ; if N < a P (Nja; b) = ; (18) Na the conclusion that the non-converging initial con- 2b exp ; if N a ditions that are present when > 1 are true non- converging points. where a is the location parameter, while b > 0, is the diversity. In our case we are interested only for 4.3. An overview analysis the x a part of the distribution function. The color-coded convergence diagrams on the con- In Table 1 we present the values of the location guration (x; y) space, presented earlier in Figs. 2 parameter a and the diversity b, as they have been and 5 provide sucient information regarding the obtained through the best t, for all cases discussed attracting domains however for only a xed value of in Figs. 4, and 7. One may observe that for most of the perturbation parameter . In order to overcome the cases the location parameter a is very close to this handicap we can de ne a new type of distri- the most probable number N of iterations, while bution of initial conditions which will allow us to in some cases these two quantities coincide. scan a continuous spectrum of values rather than few discrete levels. The most interesting con gura- tion is to set x = y, while the value of the per- turbation parameter will vary in the interval (0; 5]. By the term \tails" of the distributions we refer to the right- This technique allows us to construct, once more, hand side of the histograms, that is, for N > N . January 3, 2018 2:3 IJBC_01 14 E.E. Zotos Fig. 10. (a-left): The Newton-Raphson basins of attraction on the (x = y; ) plane, when 2 (0; 5]. The color code denoting the attractors is the same as in Fig. 2. (b-right): The distribution of the corresponding number N of required iterations for obtaining the basins of convergence shown in panel (a). the basin entropy. We assume that there are N (A) Figure N a b attractors (equilibrium points) in a certain region 4a 0.0001 23 N + 3 9.65 R = [10; 10] [10; 10] of the con guration space 4b 0.1 17 N + 2 6.59 in our dynamical system. Moreover, R can be sub- 4c 0.6 14 N + 2 5.14 divided into a grid composed of N square boxes. 4d 0.9 12 N 2.54 Each box of the square grid can contain between 1 4e 0.99 8 N 1.59 and N (A) attractors. Therefore we can denote P 4f 0.999 6 N + 1 2.04 i;j the probability that inside the box i the resulting 7a 1.0 8 N + 1 2.64 attractor is j. Due to the fact that inside the box 7b 1.1 17 N 1 3.03 the initial conditions are completely independent, 7c 1.5 17 N + 1 5.01 the Gibbs entropy, of every box i, is given by 7d 2.0 18 N + 1 9.04 7e 3.0 18 N + 2 21.81 S = P log ; (19) 7f 5.0 19 N + 1 6.95 i i;j i;j j=1 where m 2 [1; N ] is the number of the attractors i A 5. Parametric evolution of the basin inside the box i. The entropy of the entire region R, on the con- entropy guration (x; y) space, can be computed as the sum In the previous Section we discussed the fractality of the entropies of the resulting N boxes of the of the convergence diagrams using only qualitative square grid as S = S . On this basis, the en- i=1 arguments. However it would be very informative tropy relative to the total number of boxes N , which if we could have quantitative results regarding the is called basin entropy S , is given explicitly by the evolution of the fractality. In a recent paper [Daza following expression et al., 2016] a new tool for measuring the uncer- XX tainty of the basins has been introduced. This new 1 1 S = P log : (20) b i;j tool is called the \basin entropy" and refers to the N P i;j i=1 j=1 topology of the basins, thus describing the notion of fractality and unpredictability in the context of Using the above-mentioned expressions and basins of attraction or basins of escape. also adopting the value " = 0:005, suggested in Let us brie y recall the numerical algorithm of [Daza et al., 2016], we computed the basin entropy January 3, 2018 2:3 IJBC_01 Basins of convergence of equilibrium points in the generalized Hill problem 15 6. Discussion and conclusions The aim of this work was to numerically compute the basins of attraction, associated with the libra- tion points, in the generalized Hill problem. Of paramount importance was the determination of the in uence of the perturbation parameter on the position as well as on the stability of the equilibrium points. Using the multivariate Newton-Raphson it- erative scheme we managed to reveal the beautiful structures of the basins of convergence on several types of two-dimensional planes. The role of the at- tracting domains is very important since they de- scribe how each initial condition is attracted by the equilibrium points of the system, which act as at- tractors. Our numerical investigation allowed us to monitor the evolution of the geometry as well as of the fractality of the basins of convergence as a function of the perturbation parameter. Moreover, the basins of attraction have been successfully re- Fig. 11. Evolution of the basin entropy S , of the con gura- lated with both the corresponding distributions of tion (x; y) space, as a function of the perturbation parameter the number of required iterations, and the proba- . The vertical, dashed, red lines delimit the three intervals, bility distributions. regarding the tendency of the parametric evolution of the As far as we know, there are no previous stud- basin entropy. ies on the Newton-Raphson basins of convergence in the generalized Hill problem. Therefore, all the presented numerical outcomes of the current thor- S of the con guration (x; y) plane for several val- ough and systematic analysis are novel and this is ues of the perturbation parameter . Here it should exactly the importance and the contribution of our be clari ed that in the case where non-converging work. points are present, we count them as an additional The most important outcomes of our numerical basin which coexists with the other basins, corre- analysis can be summarized as follows: sponding to the equilibrium points. In Fig. 11 we present the evolution of the basin entropy as a func- (1) The perturbation parameter strongly in uences tion of the perturbation parameter. At this point, it the dynamical properties of the system. When should be noted that for creating this diagram we 0 < < 1 four equilibrium points exist, while used numerical results not only for the cases pre- for 1 there are only two real libration points. sented earlier in Figs. 2 and 5 but also from addi- (2) Our computations indicate that all the equilib- tional values of . We see that rium points of the system are always unstable. (3) In all examined cases, regarding the numeri- When ! 0 the basins get more complicated, which cal value of the perturbation parameter , the results in an increase of the basin entropy, which basins of attraction corresponding to all equi- displays its maximum value around 0.7. However librium points extend to in nity. as the value of the perturbation increases the value (4) When > 1 we detected a portion of non- of S decreases rapidly and when = 1 the basin converging initial conditions. Additional nu- entropy is almost zero, since the basins look very merical calculation (by setting a much higher smooth (see panel (a) in Fig. 5). limit of allowed iterations) revealed that these When > 1 new fractal structures emerge and the initial conditions are initial conditions for which basin entropy increases almost linearly up to = 4, the iterative scheme fails to converge to one of where S 0:4. the attractors of the system. When > 4 it is seen that the value of the basin (5) The iterative method was found to converge entropy seems to saturate (reaches a plateau), thus very fast (0 N < 15) for initial conditions displaying a constant value at around 0.39. around each equilibrium point, fast (15 N < January 3, 2018 2:3 IJBC_01 16 REFERENCES 25) and slow (25 N < 50) for initial condi- References tions that complement the central regions of the Abraham, R. & Marsden, J.E. [1987] Foundations very fast convergence, and very slow (N 50) of Mechanics, Second Edition, Addison-Wesley for initial conditions of dispersed points lying Publishing Company, Inc. , Redwood City, either in the vicinity of the basin boundaries, CA. or between the dense regions of the equilibrium Aguirre, J, Viana, R.L. & Sanju an, M.A.F. [2001] points. \Wada basins and chaotic invariant sets in the (6) As the value of increases from 0 to 1 the most H enon-Heiles system", Phys. Rev. E 64, pp. probable number of required iterations, N , was found to decrease, while for > 1 the tendency Asique, Md.Ch., Prasad, U., Hassan, M.R. & is reversed. Suraj, Md.S. [2016] \On the photogravita- (7) It was observed that the basin entropy of the tional R4BP when the third primary is a triax- con guration (x; y) plane is highly in uenced ial rigid body", Astrophys. Space Science, 361, by the perturbation parameter. More precisely, pp. 379. the highest value of S is exhibited when ! 0, Baltagiannis, A.N. & Papadakis, K.E. [2011] \Equi- while on the other hand the basin entropy tends librium points and their stability in the re- to zero when ! 1. stricted four-body problem", Int. J. Bifurc. Chaos, 21, pp. 2179-2193. For all the calculation, regarding the deter- Croustalloudi, M.N. & Kalvouridis, T.J. [2007] \At- mination of the basins of attraction, we used a tracting domains in ring-type N-body forma- double precision numerical code, written in stan- tions", Planet. Space Science, 55, pp. 53-69. dard FORTRAN 77 [Press et al., 1992]. Furthermore, Croustalloudi, M.N. & Kalvouridis, T.J. [2013] the latest version 11.2 of Mathematica [Wolfram, \The Restricted 2+2 body problem: Paramet- 2003] was used for creating all the graphical illustra- ric variation of the equilibrium states of the tion of the paper. For the classi cation of each set minor bodies and their attracting regions", of initial conditions, in all types of two-dimensional ISRN Astronomy and Astrophysics, Article ID planes, we needed about 5 minutes of CPU time, r TM using an Intel Quad-Core i7 2.4 GHz PC. Daza, A., Wagemakers, A., Sanju an, M.A.F. & We hope that the present numerical outcomes Yorke, J.A. [2015] \Testing for basins of to be useful in the active eld of basins of conver- Wada", Scienti c Reports, 5, 16579. gence in dynamical systems. Since our present ex- Daza, A., Wagemakers, A., Georgeot, B., Gu ery- ploration, regarding the attracting domains in the Odelin, D. & Sanju an, M.A.F. [2016] \Basin generalized Hill problem, was encouraging it is in entropy: a new tool to analyze uncertainty our future plans to expand our investigation. In in dynamical systems", Scienti c Reports, 6, particular, it would be of great interest to try other types of iterative formulae (i.e., of higher order, with Douskos, C.N. [2010] \Collinear equilibrium points respect to the classical iterative method of Newton- of Hill's problem with radiation and oblateness Raphson) and determine how they in uence the ge- and their fractal basins of attraction", Astro- ometry of the basins of convergence. Additionally, phys. Space Science, 326, pp. 263-271. the disconnected Wada property (e.g., [Kennedy & Douskos, C.N., Kalantonis, V., Markellos, P. & Per- Yorke, 1991; Daza et al., 2015]) is a very common dios, E. [2012] \On Sitnikov-like motions gen- feature in Newton-Raphson schemes. Therefore, we erating new kinds of 3D periodic orbits in could certainly examine if the basins of attraction in the R3BP with prolate primaries", Astrophys. the generalized Hill problem have also this striking Space Science, 337, pp. 99-106. topological property. Gousidou-Koutita, M. & Kalvouridis, T.J. [2009] \On the eciency of Newton and Broyden nu- Acknowledgments merical methods in the investigation of the reg- I would like to express my warmest thanks to the ular polygon problem of (N + 1) bodies", Appl. anonymous referees for the careful reading of the Math. Comput, 212, pp. 100-112. manuscript and for all the apt suggestions and com- Hill, G.W. [1886] \On the part of the motion of ments which allowed us to improve both the quality lunar perigee which is a function of themean and the clarity of the paper. January 3, 2018 2:3 IJBC_01 REFERENCES 17 motions of the Sun and Moon", Acta Math, 8, magnetic Copenhagen problem", International pp. 136. Journal of Non-Linear Mechanics, 90, pp. 111- Kalvouridis, T.J. [2008] \On some new aspects 123. of the photo-gravitational Copenhagen prob- lem", Astrophys. Space Science, 317, pp. 107- Kalvouridis, T.J. & Gousidou-Koutita, M.Ch. [2012] \Basins of attraction in the Copen- hagen problem where the primaries are mag- netic dipoles", Applied Mathematics, 3, pp. 541-548. Kennedy, J. & Yorke, J.A. [1991] \Basins of Wada", Physica D, 51, pp. 213-225. Kozlov, V. & Polekhin, I. [2017] \On the covering of a Hill's region by solutions in the restricted three-body problem", Celest. Mech. Dyn. As- tron, 127, pp. 331-341. Kumari, R. & Kushvah, B.S. [2014] \Stability re- gions of equilibrium points in restricted four- body problem with oblateness eects", Astro- phys. Space Science, 349, pp. 693-704. Motter, A.E. & Lai, Y.C. [2001] \Dissipative chaotic scattering", Phys. Rev. E, 65, 015205. Petit, J.M. & H enon, M. [1986] \Satellite encoun- ters", Icarus, 66, pp. 536555. Petit, J.M. & H enon, M. [1987] \A numerical simu- lation of planetary rings. I-Binary encounters", Astron. Astrophys, 173, pp. 389404. Press, H.P., Teukolsky, S.A., Vetterling, W.T. & Flannery, B.P. [1992] Numerical Recipes in FORTRAN 77, 2nd edn. Cambridge Univer- sity Press, Cambridge, USA. Seoane, J.M. & Sanju an, M.A.F. [2008] \Exponen- tial decay and scaling laws in noisy chaotic scattering", Phys. Let. A, 372, pp. 110-116. Seoane, J.M., Aguirre, J., Sanju an, M.A.F. & Lai, Y.C. [2006] \Basin topology in disipattive chaotic scattering", Chaos, 16, 023101. Szebehely. V. [1967] Theory of Orbits, Academic Press, New York. Wolfram, S. [2003] The Mathematica Book, Wolfram Media, Champaign. Zotos, E.E. [2016] \Fractal basins of attraction in the planar circular restricted three-body prob- lem with oblateness and radiation pressure", Astrophys. Space Science, 361, pp. 181. Zotos, E.E. [2017] \Revealing the basins of con- vergence in the planar equilateral restricted four-body problem", Astrophys. Space Science, 362, 2. Zotos, E.E. [2017] \Determining the Newton- Raphson basins of attraction in the electro-

Nonlinear Sciences – arXiv (Cornell University)

**Published: ** Jan 2, 2018

Loading...

You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!

Read and print from thousands of top scholarly journals.

System error. Please try again!

Already have an account? Log in

Bookmark this article. You can see your Bookmarks on your DeepDyve Library.

To save an article, **log in** first, or **sign up** for a DeepDyve account if you don’t already have one.

Copy and paste the desired citation format or use the link below to download a file formatted for EndNote

Access the full text.

Sign up today, get DeepDyve free for 14 days.

All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.