Sei sulla pagina 1di 23

The Gravitational Three-Body Problem:

Investigating Periodic Orbits with Ode113

Brent Mason

Advisor Dr. Miroslaw Mystkowski

Gardner-Webb University

Spring 2016

The Gravitational Three-Body Problem:

Investigating Periodic Orbits with Ode113


From the future chemist to the future car salesman, all young students begin

their unique exploration of the sciences with familiar introductions. These will, without

exception, include some basic explanation of gravity, often accompanied by an

entertaining fable about Sir Isaac Newton and an apple. From a combination of

classroom lectures and perhaps unfortunate first-hand experiences, everyone has

acquired at least a rudimentary understanding of the gravitational force. But

generations of ingenious scientists have taken our understanding of gravitational

mechanics far beyond Newton’s Principia.

The work of Leonhard Euler, Joseph-Louis Lagrange, and Carl Gustav Jacob

Jacobi among others, specifically laid the groundwork for centuries of exploration into

one of the most persistently difficult problems in classical mechanics, the gravitational

three-body problem. More recently, mathematicians have utilized modern computing

power to numerically integrate solutions quickly and with previously unobtainable

levels of precision. However, even with 21st century innovations and nearly 350 years

since Newton first pondered the cause of lunar perturbations (Fitzpatrick 5), only a

handful of periodic solutions to the unrestricted problem of three bodies have been

identified.

With this paper I hope to provide an outline of the successive characterizations

of these systems made over the past three centuries through to modern day analysis and

the search for periodic orbits.

History
Sir Isaac Newton, after both formulating his groundbreaking equations of

motion and defining the inverse-square relation of the force of gravity, went on to, in

1687 and for the first recorded time, pose the “problem of the movements of three

massive bodies subject to their mutually perturbing gravitational attractions”. The

Father of Classical Mechanics dedicated his 66th proposition of Principia to the question

and would apply the formulas he developed to the problem as it pertained to lunar

theory of the time (Three-Body).

Following Newton, in 1767 Swiss mathematician Leonhard Euler discovered the

first three families of periodic solutions, the collinear solutions corresponding the three

libration points L1, L2, and L3, which I will describe in greater detail. He resolved these

solutions for a reduced system that he defined, the planar circular restricted three-body

problem (or PCR3BP), though his solutions can be generalized to the “honest”, or

unrestricted, problem. This reduced formulation has been the subject of as much

research as the unrestricted problem because of both its close resemblance to real-life

orbital systems and the comparatively low complexity of its solutions.

In 1788, Joseph-Louis Lagrange put forth his reformulation of Newton’s

equations, which greatly simplified analysis by allowing for the generalization of

coordinate systems (Lagrangian). Lagrangian mechanics permitted future researchers to

reduce the dimension of the solution space by factoring out symmetries and treating

conserved values as constraints. Lagrange himself also investigated the PCR3BP and

identified the other two libration points, L 4 and L5, only 5 years after Euler in 1772

(McInnes 2).
Many others have offered notable insights. Carl Gustav Jacob Jacobi identified

the singular conserved quantity of the PCR3BP, the Jacobi integral, which restricts the

third body to a region in the plane that satisfies the constraint (Collins 114). French

physicist Henri Poincaré accurately described the chaotic nature of the infinitely many

solutions to the PCR3BP not corresponding to a libration point in a paper published in

1892, nearly a century before the technology to model such orbits existed (Moeckel).

Most recently, in 1988, American mathematicians Rick Moeckel and Richard

Montgomery defined a mapping of the phase space of the honest three-body problem on

to a shape sphere in , enabling the discovery of up to 13 new families of periodic

solutions to the unrestricted problem in just 3 decades (Chenciner).

Initial Thoughts

I was initially drawn to the topic because of a long-held fascination with

astronomy. I have made observations of the movements of celestial bodies with an 8”

Newtonian reflector since I was a teenager, and the fact that such massive systems are,

at their core, deterministic and can be modeled by well-known equations appeals to me.

While searching for a specific subject, I came across the three-body simulations created

by Volker Dörr, which featured systems with masses and initial conditions that were

completely theoretical with no resemblance to real-world scenarios. This is when I

realized the potential of computers to numerically integrate even the most exotic of

these systems and decided on this problem as my research topic.

I understood relatively little about the problem prior to beginning my research.

Having taken a year of college physics, I understood the inverse-square, attractive


nature of the gravitational force as well as the relationship between the kinetic and

potential energy of a dynamic system. I knew how to integrate numerically using Euler

methods, though I was inexperienced putting these methods into practice with computer

programs.

The Problem: Basics

The gravitational three-body problem is the problem of calculating the position

and velocity of three massive bodies under the influence of their mutual gravitational

attraction at a time t from specified initial conditions. Systems are parameterized by the

three mass quantities, M1, M2, and M3, and Newton’s Shell Theorem allows bodies to be

treated as point masses (Moulton 141-145). Because gravitational force vectors

between three bodies are always planar, the problem can be modeled in two dimensions

as long as the initial velocities are planar. Therefore the problem is one of 12 degrees of

freedom, as 2 position coordinates, , and 2 velocity coordinates, , for each

of 3 bodies ( ) is required to specify a unique configuration, and these can by

modeled by 6 simultaneous, 2nd order differential equations:

.
Therefore, the solution space over the lifetime of an initialized system is a

parametric curve dependent on in the phase space of all possible position and

velocity configurations. Liouville’s Theorem describes the deterministic nature of these

systems as a product of the uniform density of points within the phase space and implies

that the region of phase space accessible to a system relies solely on its initial

conditions (Collins 124-125). It follows then, that phase space curves will never

intersect. A solution is said to be periodic if and only if its phase space curve is closed

such that each mass returns to its precise initial state:

As our understanding of three-body systems has evolved and our ability to

integrate them numerically has increased with technology, research has become more

and more focused on identifying and classifying this periodicity. Today, the frontier of

research into the gravitational three-body problem exists in quantifying the density of

periodic solution curves within the entire phase space.

The Problem: Two Bodies and Keplerian Orbits

It is useful to first understand the dynamics of a gravitational system in two

bodies so that certain properties may be generalized to the more complex problem.

Newton first provided the now-familiar analytical description of the mutual

gravitational forces that exist along the line between two bodies:
When coupled, they describe the motion of the center of mass (or barycenter), of

the system.

Since , the barycenter can be shifted to the origin and the motion modeled by

paired one-body problems following Kepler orbits, , in polar coordinates where

is the distance to mass from and is the true anomaly, or angle from

perihelion. As previously stated, because gravity is a central force and angular

momentum is conserved, each orbit can be modeled in two dimensions. The total

energy, , of the two-body system is also a constant quantity,

The magnitude of the eccentricity vector, , exists as the final constant of motion of the

two body system, allowing the integration of distinct orbits:

So the length of the radius between each body and the barycenter, or the norm of the

orbital position vector , is a function of only the true anomaly, :

,
.

Then, from polar equations in this form, the shapes of orbits are conic sections

whose type corresponds to the value of , such that orbits are

The Problem: Matlab and Integration with ode113

I chose to use the integration tools found in the program Matlab to model the

three-body problem. Matlab offers dozens of integrating “functions” which perform

various Runge-Kutta methods and take up to 4 arguments. The default choice that is

most efficient for the widest range of problems is ode45. Matlab documentation states

this script, “should be the first solver you try” and will be adequate for “nonstiff”

ODE’s, or ones which do not require widely varying time steps. However, ode113 is

recommended specifically for the problem of modeling the orbits in our problem, as

solution curves are relatively smooth but strict error tolerances are essential (Matlab).

Ode113 is called as follows:

[t,y] = ode113(odefun,tspan,y0,options).

The primary argument required by ode113 is the “odefun” input. This is the .m file

containing the system of up to 12 ODE’s to be integrated. The function also requires

vectors corresponding to the time span to be integrated over and the initial value of each

ODE. Finally, a series of options can be fine-tuned. I chose to take advantage of

ode113’s expressed ability to minimize error by setting both absolute and relative

tolerances to 1e-14.
The Problem: The Planar Circular Restricted Three-Body Problem

Before the existence of modern computing power, the three-body problem was

analyzed in a form that was more manageable for pen-and-paper calculations. The

PCR3BP restricts the size of the 3rd mass such that the problem can be calculated in

terms of a pair of two-body systems such that . Also, motions are

confined to a plane and the orbits of the two primaries are circular.

This reductive model has many advantages. Firstly, because the length of the

radius between the primaries is invariant over time, a transformation can be made from

the inertial to a synodic, or co-rotating reference frame such that, for any vector in

the initial frame of reference,

where is the true anomaly of the primaries . As a result, the positions of the

primaries remain constant and collinear along the x’-axis with at the origin. Also,

because of the relative stability inherent to configurations of this form, they bare

striking resemblance to systems most commonly observed in nature, such as the sun-

earth-moon system (McInnes 3-5).

Again, the system is one of 12 degrees of freedom and as such, its solution

requires 12 constants of motion to be integrated analytically. But, since the

gravitational force is no longer central, the angular momentum is not constant in this

frame of reference. The constant positions and zero velocities of the primaries reduce
the order of the problem by 8, however the required angular coordinate, , of axis

rotation returns the dimension of our solution space to 5 ( for circular orbits).

Total energy is still conserved in the form of Jacobi’s constant, as long as

centrifugal and Coriolis forces present in the synodic reference

frame are accounted for:

where is equal to the constant mean motion and accounts for

the now variable angular component of momentum. The Jacobi

constant is the only quantity conserved in all forms of the three-

body problem and is equivalent to . In the case of the PCR3BP,

it provides the only restriction on the motion of the 3 rd mass by defining zero velocity

surfaces (Fitzpatrick 124-127). By solving for and setting the resultant equation

equal to zero, it can be seen that for the 3 rd mass to pass through the surface defined,

must change sign. Therefore, as decreases, the 3rd mass is dynamically included in a

larger and larger space, or Hill Region (see figure 1).

In stating the problem in this way, we can deduce the 5 libration, or Lagrange,

points for which the centrifugal and gravitational potentials are balanced and the 3 rd

mass remains stationary relative to the primaries.


FIGURE 1: Views of zero
Euler proved that the first 3 of these are collinear, velocity surfaces from the
orbital plane (top) and from
residing on the x’ axis such that y’ = z’ = 0. Then, setting the above (bottom)

component of acceleration in the x’ direction equal to zero,


The resulting quintic equation has 5 roots that can be calculated numerically with

Newton’s method (McInnes 7). Two of these roots correspond to the positions of the

primaries themselves. The other three represent saddle points of the composite

potentials existing at the L1, L2, and L3 libration points. Consequently, while masses at

precisely these points will exhibit periodic orbits, the motion of objects in their

neighborhoods will demonstrate the same chaos characteristic of these systems.

On the other hand, the 2 libration points discovered by Lagrange himself have

been proven stable given a certain mass ratio between the primaries. The L 4 and L5

points exist at and such that

At any time, the three bodies will form an equilateral triangle. Given ,

perturbations of an object in the vicinity of the L4 and L5 points will be corrected for,

and because of this quality, these regions are often home to cosmic debris that is trapped

naturally by the surrounding potentials (Fitzpatrick 127-131).

Other constrained forms of the three-body problem have been examined,

however such efforts have not yet revealed much more regarding periodicity that can be

meaningfully generalized to the honest problem. For instance, study of the planar

elliptical restricted system has yielded information vital to the launching and guiding of
spacecraft and satellites such as how Hill regions and stability near libration points

manifest in orbits with nonzero eccentricity (Chenciner). Also, similar non-planar

periodic solutions were proven for circularly-constrained orbits and an infinitesimal 3 rd

mass by American physicist William Duncan MacMillan and proposed in the

elliptically-constrained case by Russian mathematician Kirill Alexandrovitch Sitnikov

in 1911 and 1961 respectively (see figure 2), though neither has been related to the

unrestricted problem (Chenciner).

FIGURE 2: The restricted 3rd masses in the periodic orbits described by


MacMillan (left) and Sitnikov (right) oscillate perpendicularly through the orbital
plane of the primaries

The

Problem: The Honest Three Body Problem

These qualities of the solutions to the PCR3BP are useful in as much as they are

shared by solutions to the problem in which the mass parameters are unrestricted. As it

turns out, the 5 solutions periodic in the PCR3BP identified by Euler and Lagrange each

represent a class of periodic solutions to the true problem of three bodies whose mass

ratios, angular momentum, and energy is allowed to vary (see figure 3).

U n f o r t u n a t e l

unconstrained problem, these

remained the only 5 classes of

periodic orbits identified for over

FIGURE 3: Six examples of the L4 class of periodic


orbits varying by mass ratio (rows) and by angular
momentum (columns)
150 years, until the dawn of the computer age allowed Roger Broucke and Michel

Henon to resolve a family of new periodic orbits (Three-Body). However, even then

progress stalled for a few decades until researchers Moeckel and Montgomery proposed

a new method of analysis. In 1988, the two mathematicians from the University of

Minnesota proposed a mapping of composite projections from the space of possible

coplanar positionings of the 3 equal masses with 0 total angular momentum equivalent

to on to the surface of a non-Euclidean “shape sphere” in of corresponding

oriented congruent triangles such that each mass is located at a vertex (Chenciner).

The Problem: The Shape Sphere

Moeckel and Montgomery begin the construction of their mapping

by assigning the 3 masses position vectors in the complex plane rather than , such

that the configuration space becomes , in order to simplify the operation of rotation

by angle . Then the located triangle formed by the masses is a vector with complex

components,

A mass inner product is defined to induce a Hermitian mass metric on the space,

which will later account for varying potentials in the final shape space,

This initial 6-dimensional metric space is then reduced by a series of surjective

projections performed by first identifying oriented congruence and similarity classes of

triangles and then factoring out corresponding regions of the solution space.
The oriented congruence classes are specified by identifying triangles that are

invariant under the group of rigid-body transformations of translation and rotation. The

image of the corresponding component mappings of are the quotient spaces formed

by dividing the preimage by the equivalence relation:

Equating the triangles further by similarity takes the solution space from the

shape space to the shape sphere, and because each of these spaces exhibits radial

symmetry, the necessary mapping, , remains isometric. The partition of oriented

similarity classes groups together triangles that are similar when their coordinates are

scaled.

The resulting space of all oriented congruence and

similarity classes is the surface of a 3-dimensional unit

sphere centered at the triple collision point with an

inherited non-Euclidean geometry.

FIGURE 4: Notable points on the


shape sphere of oriented triangles
The -coordinate indicates the scaled, oriented area of the triangle formed by

the 3 masses such that triangles with inverted vertices are located at . Then the

two equilateral cases including each of the families of solutions discovered by Lagrange

are mapped to the poles . Alternately, degenerate triangles end up along the

unit circle with systems

in which 2 masses are colliding

exist at the third roots of unity.

Euler’s 3 families of solutions in

which the masses remain FIGURE 5: Contours of the point density showing peaks
centered at the double-collision points of 3-body systems with
equal masses (left) and mass ratios 1:2:10 (right)
collinear exist at points along

the hemisphere equidistant

from two double collision points with respect to the inherited metric.

This mass metric bestows a useful quality on the shape space. The density of

points equidistant in Euclidean space reaches local maxima on the shape sphere as we

approach instances where two masses coincide. Therefore when they are visible, it is

apparent that the contours of the density of space induced by the mass metric relate to

the gravitational force field for the given mass ratios, with strongest forces existing at

each of the double collision points, saddle points between them, and local minima of the

potential located at the poles (see figure 5). In fact, this density is proven to vary

precisely as the inverse-square of the distance in shape space from the double collision

point, such that energy required to traverse a curve, , in phase space depends entirely

on the distance traveled, . This greatly simplifies the task of identifying periodic
solutions by the process of minimizing the action function over all curves between

strategically chosen points in a closed orbit.

Unfortunately, the methods of choosing appropriate points and minimizing the

total energy of the paths between them exceed the scope of this paper, though it can be

inferred from the Lagrangian’s dependence upon that minimizing curves will follow

geodesics in the phase space (Montgomery). And it was by exactly this process that

American computer scientist Cristopher Moore identified the figure eight family of

periodic solutions in 1994, and then most recently, Milovan Šuvakov and Veljko

Dmitrašinović described 13 more likely families of periodic solutions, representing

arguably the most significant breakthrough in this field in centuries (Chenciner).

Perturbations, Chaos, and Error

It is important to note that the periodic nature of the orbits mentioned is entirely

theoretical, as it depends on the complete absence of perturbing forces. These

conditions do not exist in any observable system, so even with an understanding of an

orbits stability, calculations require the inclusion of a mathematics of perturbations to

make them at all meaningful to real-world endeavors.

Equations in this field have been developed to account for forces resulting from

additional masses and non-spherical bodies, frictional forces like air resistance, as well

as the influence of special relativity (Gidea, Canalias, Masdemont).


Renowned physicist Henri Poincaré gave a description of the chaos exhibited by

the 3rd body in a PCR system that was remarkably intuitive for its time:

“If one tries to imagine the figure formed by these two curves with an infinite number

of intersections… these intersections form a kind of trellis, a fabric, a network of infinitely tight

mesh; each of the two curves must not cross itself but it must fold on itself in a very complicated

way to intersect all of the meshes of the fabric infinitely many times.” ( Poincaré)

These curves Poincaré describes are the possible phase space curves of an

initiated PCR system along stable and unstable manifolds in 2 dimensions. Homoclinic

orbits are a result of these two manifolds’ infinite transverse intersections,

corresponding to a Smale horseshoe map, and lead to incredibly complicated and

unpredictable dynamics (Moeckel). This fact is apparent when analyzing the variations

produced by even minute changes to initial conditions.

This volatility underscores the need to minimize error when integrating solutions

of this problem numerically. The ode113 function accomplishes this by progressively

decreasing the integration step over the periods of a solution that are particularly

energetic, near double collision points for instance, until error bounds are met (Matlab).

Relation to Branches of Mathematics

While solutions to the gravitational three-body problem are most directly

resolved using the tools of physics and differential calculus, analysis from the viewpoint

of other mathematical disciplines can be informative.

Fundamental to plotting and classifying the behavior of orbits is a firm grasp of

the geometry of conic sections. Familiarity with the equations of ellipses and
hyperbolas in both polar and Cartesian coordinates helps in quickly understanding how

a system will develop in terms of conserved or variable quantities.

Designing a composition of mappings like the one created by Moeckel and

Montgomery requires a firm grasp of linear and abstract algebra. Even utilizing such a

function involves identifying the codomain as well as the properties and operations

inherited through quotient maps.

Finally, a knowledge of certain concepts of topology, especially those pertaining

to the intersection and stability of manifolds, would allow for a deeper investigation into

the chaotic nature of three-body systems.

Modern Application: Planet 9

Though perfectly periodic orbits in three-body systems exist only in theory, the

quest to discover and understand them has produced invaluable explanations of how and

why real-world orbits tend to develop in the ways they do.

A recent application of this knowledge this past January has led researchers at

Caltech to hypothesize the existence of a 9th planet in our solar system at a orbiting

distance 20 times that of Neptune. Astronomers Konstantin Batygin and Mike Brown

made this claim solely based on the necessity of a source of the gravitational forces that

appear to have manipulated the orbits of up to 6 smaller bodies. The Kuiper Belt

objects, which include the minor planet Sedna, all exhibit highly eccentric orbits aligned

to share certain region of space as a common foci. In order to perturb 6 orbits in this

manner, the proposed addition to the solar system would need to be 5,000 times more

massive than former 9th planet Pluto.


Given that its proposed orbit would put it at such a lengthy distance from our

telescopes, it is understandable that Planet Nine has yet to be observed directly.

However Batygin and Brown estimate the probability that the 6-body system of orbits

would occur naturally without a complementary source of gravity at 0.007. Additional

support for the existence of Planet Nine resides in a now confirmed prediction of Kuiper

Belt objects orbiting in the same region in a plane perpendicular to the known orbits.

Four bodies orbiting in the prescribed manner have since been observed. This mounting

level of evidence means that completions of the next generation of massive telescopes

in the coming years will almost certainly coincide with a correction to astronomy

textbooks (Fesenmaier).

In Conclusion

For centuries, resolving precise solutions to the gravitational three-body problem

was such an unmanageable task that the most celebrated minds in history were forced to

formulate a much simpler, constrained question in order to say anything at all about the

way such solutions must behave. Motion of this simpler variety was well understood in

comparison and could be described by established equations of classical mechanics.

Also, at the other extreme, motion of systems of large numbers of particles governed

essentially by randomness could be predicted with unprecedented accuracy using the

probabilistic tools of thermodynamics and quantum mechanics. However, the chaotic

nature of systems of intermediate complexity, like those of three bodies, resisted human
comprehension right up until the development of computing power adequate for

numerical integration within the last 50 years.

With the aide of advanced processors and intuitive integrating algorithms, the

limits of our understanding of the gravitational three-body problem are being extended

each year. The frontier of research in this field exists today in identifying families of

periodic orbits in order to one day quantify the density of such orbits within the greater

phase space. And given the brilliant deductions made fundamentally by Newton, Euler,

Lagrange, and Jacobi, today’s researchers can view this frontier from the shoulders of

giants.

Works Cited

Attaway, Stormy. Matlab: A Practical Introduction to Programming and Problem

Solving. 3rd ed. S.L.: Butterworth-Heinemann, 2013. Print.

Chenciner, Alain. "Three Body Problem." Scholarpedia. 3 Oct. 2007. Web.


<http://www.scholarpedia.org/article/Three_body_problem>.

Collins, George W. "Chapter 8: The Dynamics of More Than Two Bodies." The

Foundations of Celestial Mechanics. Tucson: Pachart House, 1989. 111-28.

NASA Astrophysics Data System. Web.

<http://ads.harvard.edu/books/1989fcm..book/Chapter8.pdf>.

Fesenmaier, Kimm. "Caltech Researchers Find Evidence of a Real Ninth Planet |

Caltech." The California Institute of Technology. 20 Jan. 2016. Web.

<https://www.caltech.edu/news/caltech-researchers-find-evidence- real-ninth-

planet-49523>.

Fitzpatrick, Richard. An Introduction to Celestial Mechanics. Cambridge: Cambridge

UP, 2012. Afhalifax.ca. Alliance Française. Web.

<http://www.afhalifax.ca/bete/DALEMBERTIMAGES/CoursMecanique

Celeste/CelestialFitzpatrick.pdf>.

Gidea, Marian, Elisabet Canalias, and Josep Masdemont. Chaotic Dynamics in the

Three-body Problem: With Applications to Space Mission Design. Rep. Northeastern

Illinois University, 28 Mar. 2005. Web. <http://www.ma1.upc.edu/recerca/reports-

recerca/repre- 2005/Fitxers/rep040502masde.pdf>.

"Lagrangian Mechanics." Wikipedia. Wikimedia Foundation. Web.

<https://en.wikipedia.org/wiki/Lagrangian_mechanics>.

"Matlab Documentation." Mathworks.com. The MathWorks, Inc. Web.

<http://www.mathworks.com/help/index.html>.

McInnes, Allan. "An Introduction To Libration Point Orbits." ECE Cowebs.

University of Canterbury, 5 July 2009. Web.


<http://coweb.elec.canterbury.ac.nz/cda/uploads/mcinnes-

intro_to_libration_point_orbits.pdf>.

Moeckel, Richard, and Richard Montgomery. "Symmetric Regularization, Reduction

and Blow-up of the Planar Three-body Problem." Pacific Journal of Mathematics

Pacific J. Math. 262.1 (2013): 129-89. ArXiv.org. Web.

<http://arxiv.org/pdf/1202.0972v1.pdf>.

Moeckel, Rick. "Chaos in the Three-Body Problem." Poincaré 100 Conference.

Institut Henri Poincare, Paris, France. Nov. 2012. Presentation. Web.

<http://www.math.umn.edu/~rmoeckel/presentations/PoincareTalk .pdf>

Montgomery, Richard. "The Three-Body Problem and the Shape Sphere." The

American Mathematical Monthly 122.04 (2015): 299-321. ArXiv.org. Web.

<http://arxiv.org/pdf/1402.0841v1.pdf>.

Moulton, Forest Ray. "Chapter V: The Problem of Two Bodies", "Chapter VIII: The

Problem of Three Bodies." An Introduction to Celestial Mechanics. New York:

Macmillan, 1914. 140,190; 277,320. Internet Archive. Web.

<https://ia801407.us.archive.org/28/items/introcelestial00moulrich/intr

ocelestial00moulrich.pdf>.

Poincaré, Henri. Les Méthodes Nouvelles De La Mécanique Céleste. New York: Dover

Publication, 1957. Print.

Šuvakov, Milovan, and V. Dmitrašinović. "Three Classes of Newtonian Three- Body

Planar Periodic Orbits." Phys. Rev. Lett. Physical Review Letters 110.11 (2013).

ArXiv.org. Web. <http://arxiv.org/pdf/1303.0181v1.pdf>.

"Three-body Problem." Wikipedia. Wikimedia Foundation. Web.


<https://en.wikipedia.org/wiki/Three-body_problem>.

Potrebbero piacerti anche