Skip to main content
Aug 16, 2010 4:28 PM UTC
Question

"Dark Star" Steals Planet

In this simulation of three gravitating Newtonian particles, a blue "planet" orbits a red "star" until they meet the "dark star." The physics is simple, yet the numerical behavior is extreme. Indeed, Mathcad's numerical integrators have difficulty modeling the motions correctly. See http://twt.mpei.ac.ru/TTHB/1/3-Planets.xmcdz for the Mathcad 14 worksheet.

 

Unable to play video. Please try again later.
(view in My Videos)

    4 replies

    Aug 16, 2010 7:29 PM UTC

    The "black star" appears to have a discontinuity in its trajectory.  Is that expected or even plausible?  I had thought that the black star was relatively massive, and that the blue star was relatively light, so the blue star shouldn't be capable of dragging black star around like that.

    Aug 17, 2010 7:09 AM UTC

    >The "black star" appears to have a discontinuity in its trajectory

    No, see http://twt.mpei.ac.ru/ochkov/Mathcad_12/Planet/03_Intercept_Sputnik.png

    And try please create new cases - http://twt.mpei.ac.ru/MCS/Worksheets/3-Planets.xmcd

    Aug 17, 2010 9:15 PM UTC

    Eden,

    The masses are m1=20 (black star), m2=2 (red star), and m3=0.01 (blue planet) for Valery's dark star video, with G=1.

    Based upon my independent validation of Valery's worksheet, the physics is correctly modeled. I, too, noticed the apparent "kink" in the dark star's trajectory after it stole the blue planet. I also supspected a discontinuous derivative.

    Also, if you run the animation with more frames, you can see that the dark star moves in an apparently straight line after the "kink."

    1. Instead of Odesolve numerical integration, I attempted Rkadapt numerical integration of the 18 simultaneous first-order ODEs associated with this problem (here I modeled the motions in the x, y, and z directions, but set z=0 and vz=0 for all three bodies). Now the black star does not capture the blue planet. In other words, Rkadapt does not give the same results as Odesolve.

    2. When I try m1=20, m2=2, m3=0.1, and FRAME=299, the black star barely captures the blue planet, in that the blue planet's orbit is now highly eccentric about the black star. My results here agree perfectly with Valery's for these three masses, so I am confident that I have set up my system of 18 first-order ODEs correctly.

    3. When I try m1=20, m2=10, m3=0.1, and FRAME=799, the black star and red star exhibit three close encounters. They eject the blue planet after the second close encounter. This was not expected. The numerical integration process becomes very challenging, indeed, when the bodies behave in this manner.

    This extraordinary video and the associated Mathcad 14 worksheet are quite important for two reasons.

        a. They show that when you go from two Newtonian particles to three or more, almost anything can happen. E.g., in the case of two stars and a planet, the planet can switch stars, or it can be ejected from the system. In the much more well-understood two-body problem (e.g., my "Modeling Extrasolar Planets" video), one only has to deal with elliptical, parabolic, and hyperbolic relative motions.

        b. They provide a framework for demonstrating that Mathcad's integrators (Odesolve, Rkadapt, AdamsBDF, etc.) can give differing results for the same problem.

    Does reason (b) imply that the Mathcad integrators are incorrectly implemented? I think not. By displaying the elapsed time in the animation window, I think that we could pinpoint the time of the "kink" and investigate what is going on there. If there were a collision, then perhaps we could segment the time interval into piecewise-continuous velocity intervals and re-initialize as necessary. (But according to Valery, there is no such velocity discontinuity.)

    Valery,

    I have tried out test case #3 at your Mathcad Calculation Server, but the Server does not permit enough integration points to see if my results match yours.

    Aug 24, 2010 6:51 AM UTC

    One case of the moving of 4 "planets":

    We can see a break of the numerical solution at the end of the moving!