A Simple Explicit Model: Short term planetary orbits

Abstract A transit of planets occurs when the planet passes between the Earth and the Sun in the plane of the Earths orbit. Transits of Venus occur approximately twice every 125 years and are generally events of great interest to the astronomy community and the general public. Calculating the times at which transits of Venus will occur is a rather complicated problem in its Newtonian formulation, however a number of approximations can be made that allow vastly simplied treatments. This report details a number of methods of numerically calculating the times at which transits of Venus occur, and presents the results of some calculations produced by these methods.

Introduction

A transit of a planet occurs when the planet passes between the Earth and the Sun in the plane of the Earths orbit (the plane of the ecliptic). Transits of Venus are relatively rare events; although Venus passes between the Earth and the sun an average of once every eight years, the vast majority of these alignments are outside the plane of the ecliptic. Alignments within the plane of the ecliptic occur approximately twice every 125 years [1]. For exact calculations of planetary orbits that are accurate over arbitrarily large time frames, complicated models that take into account gravitational perturbations (in particular from Jupiter) and relativistic eects must be implemented. However, reasonable approximations over a few hundred years can be obtained by using Keplers laws of planetary motion and evolving the orbital parameters with time based on observed data. This is the approach that will be outlined in the remainder of this report.

Keplers laws of planetary motion are as follows [2]: 1. Each planet moves along an ellipse with the Sun at a focus. 2. The line between a planet and the Sun sweeps out equal areas in equal times. 3. The square of a revolutions duration (orbital period), divided by the cube of the orbits greatest width, is the same for all planets. From these laws it is possible to derive expressions for the angular position of each planet in terms of a number of observable parameters as shown in equation 1. These parameters the longitude of a planet in the plane of the ecliptic (true ), the orbital inclination (i), the longitude of the perihelion () and the ascending node () all have observed values (or are derived from observed values with relative ease) that are catalogued by NASA [4]. tan true = cos sin + sin cos cos(i) = p + tan1 ( sin( ) ) cos(i) cos( ) (1)

Although equation 1 can be used to numerically calculate the approximate times at which transits occur, it does not take into account the angular diameter of the Sun as seen from Earth. A consequence of this treatment is that transits are calculated as instant events that occur only at one discrete value of time, whereas in reality a transit of Venus will take a number of hours as the planet moves across the face of the Sun. To alleviate this issue, it becomes useful to dene vectors for each planet between the planet and the Sun (r). Dening these vectors with coordinates relative to the plane of the ecliptic, it can be shown geometrically that for a transit to occur the following condition must be met. rE (rE rV ) >| rE || rE rV | cos Where is the angular radius of the Sun. It is relatively straightforward to calculate the vectors from equation 2 for values of time over some interval surrounding an approximate transit time (identied from equation 1). Thus it is possible to see in the range of times over which a transit occurs in much greater detail. It may be possible to perform transit checks for arbitrarily large time periods, negating the need for the initial determination of transit times. However in the interests of minimising the required computing power, it was decided to rst determine the times of approximate transits and check in detail only for times in the order of days before and after. 2 (2)

2
2.1

Methods
Circular orbits

As a rst approximation to modelling the planetary orbits using Keplers laws, it can be useful to assume that the planets are in circular orbits in a single plane with the Sun at its centre. In this case the angular positions of each planet are given by: 2 (t t10 ) T1 2 2 (t) = 20 + (t t20 ) T2

1 (t) = 10 +

(3)

Where T1 and T2 are the periods of orbit of the respective planets. 10 and 20 are the initial angular positions of each planet relative to some conveniently dened axis. t10 and t20 are the times at which the intial angular positions are dened and 1 and 2 are the angular positions of the planets. From equation 3 it follows that the times at which the planets are aligned can be easily found by numerically solving equation 4 for tn . It is worth noting that virtually all orbital parameters used in this report are obtained from [4]. n + ( t10 T1
t20 T2

tn =

20 10 ) 2

1 ( T1

1 T2 )

; n = . . . , 2, 1, , 1, 2, 3, . . .

(4)

2.2

Elliptical orbits

While the assumption of circular orbits may seem rather haphazard, it will be shown that the results yielded are indeed a good approximation to those obtained from more complicated models.1 Complicating the model by introducing orbits that are elliptical from Keplers rst law and in the same plane; it is possible to derive an expression for the true anomaly ( p ) in terms of the eccentric anomaly ( a time dependant parameter of the elliptical orbit) and the eccentricity of the orbit (e) using Keplers other laws. This expression is as follows: ( 1 e2 ) sin tan( p ) = (e + cos )

(5)

1 It is worth noting that circular orbits may be a reasonable assumption in the case of Earth and Venus, but that it will not provide realistic results in the case of Mercury

The eccentric anomaly () can be calculated iteratively until it converges (usually < 10 iterations) from the relations: (n+1) M A + e sin (n) (1) M A + (6) (7)

In order to determine if a transit occurs at a time tn , the true anomaly can be calculated from equation 5 for each planet at tn . If the true anomaly is the same for each planet at tn then that time (t) can be considered an alignment. If however, the true anomaly is dierent for each planet (highly likely in the rst instance) then a new time must be chosen in order to nd an alignment. A convenient approach to selecting a new time is to use a bisection method. The bisection method in this case relies on it being possible to test whether the current time is before or after an alignment. This test can be implemented simply by taking the dot product of the true angular positions of each planet e v ( is easily calculated for each planet once the true anomaly is known). If the dot product is <1 then t is before alignment and if the dot product is >1 then t is after alignment. If tn is dened as tn = ta +tb where tb is some time 2 before alignment and ta is some time after alignment the bisection method redenes ta and tb as follows: tb = tn if the dot product is < 1 and ta = tn if the dot product is < 1. Resultantly, tn becomes closer to t (the alignment time) with each iteration. With the introduction of some new parameters the longitude of a planet in the plane of the ecliptic (true ), the orbital inclination (i), the longitude of the perihelion () and the ascending node () the above method can be adapted to take the relative tilt of each orbit into account, and hence to determine if alignments occur near the nodes (i.e. the intersection between the plane of Venus orbit and the plane of the ecliptic) using the following relations: tan true = cos sin + sin cos cos(i) = p + tan1 ( sin( ) ) cos(i) cos( ) (8)

Alignments that occur at or suciently close to one of the nodes can be assumed to be transits. Similarly, alignments that do not occur close to the nodes can be assumed to be out of plane alignments. One further complication takes into account the perturbations of each orbit due to interaction with other massive bodies in the Solar System (particularly Jupiter). This can be done by dening a time varying version of each parameter in the relevant equations. While the perturbations are not strictly constant, they vary very slowly relative to the orbits themselves and can be approximated for time periods in the order of a few hundred years by a constant. As a result of this, if the initial value of some orbital parameter (X) is known and the rate 4

 of change of X (X) is known, then the time varying version of X (X(t)) can be approximated by: X(t) = X + Xt. The rate of change of all relevant constants is available from the same source as the constants [cite NASA], and as such it is relatively straightforward to adapt the method for calculating elliptical orbits to use the time varying parameters instead of the time constant parameters.

2.3

Identifying true transits

The elliptical orbits approximation determines that transits of Venus happen at a single instant t. In reality, transits take a number of hours over which Venus can clearly be seen moving across the face of the Sun. To calculate the time range over which the transit occurs it is useful to use the vector formulation (as outlined in section 1). The condition that is true while a transit is occurring is given by equation 2.2 In order to calculate the r vector for Earth, its parameters must be substituted into the following equation: r = {u(cos u sin v) rp + v(sin u + cos v) rp } Where = p : the true anomaly 1+e = 1 + e cos a(1 e) rp = { cos + y sin z tan(i) sin( )} x 2 1 + tan (i) sin2 ( ) Calculation of r for Venus is possible, but rst it must undergo a coordinate transform to nd its orbital parameters relative to the axes dened by the Earths orbit. The coordinate transform can be calculated relatively straightforwardly through the use of the following relations: u = x cos + y sin v = sin cos(i) + y cos cos(i) + z sin(i) x n = x sin sin(i) y cos sin(i) + z cos(i) (10) (9)

Where u, v and n are the equivalent of x, y and z in Venus orbital plane. Once the coordinates have been correctly transformed, it is straightforward to calculate r for Venus in the same way as it is done for Earth. r vectors can then be calculated for a range of times surrounding suspected transits and compared with the condition from equation 2 to determine over what range of times the transit occurs.
2r E

(rE rV ) >| rE || rE rV | cos ; Where is the angular radius of the Sun.

Results

Numerous models of gradually increasing complexity were implemented in order to numerically calculate the times at which transits of Venus occur. To gain some insight into the accuracy of the various models, it is useful to compare the calculated transit times with known transit times i.e. transits that have been calculated using the full relativistic treatment or that have already occured. Figure 1 compares computed transit times and dates to times obtained from a publicly available NASA database [3]. The rst ve transits occurring after January 1 2000 are shown.

Figure 1: Numerical results Interestingly, the circular orbit model appears to produce results that are of similar accuracy to the elliptical orbit model in most cases; and achieves signicantly greater accuracy for transit 4 (08/12/2125). The vector model is consistantly accurate in its predictions, although it seems to be aected by some constant oset (starting approximately one hour early and ending approximately two hours early). From gure 2 it can be seen that this oset is not negated by reducing the size of the time steps and therefore it seems plausible that the oset is an indication of the best possible accuracy of the model. From these results it is possible to say that the vector model can calculate the times at which transits of Venus occur to with accuracy in the order of a few hours. The circular and elliptical orbit models are somewhat less reliable, but are still accurate to within a few days.

Figure 2: Eect of dierent timesteps on the accuracy of the vector model

Conclusion

Various numerical methods were used to calculate the times at which transits of Venus occur. The most accurate of these methods was the vector model outlined in section 2.3, which was able to calculate transit times to within a few hours. When it is taken into consideration how minimal the required computational power is for these methods in comparison to full relativistic or Newtonian treatments the accuracy of the results is remarkable.

References
[1] Steven J. Dick. The transit of venus. Scientic American, 290(5):98 105, 2004. [2] A T Hyman. A simple cartesian treatment of planetary motion. European Journal of Physics, 14(4):145147, 1993. [3] NASA. http://eclipse.gsfc.nasa.gov/transit/catalog/VenusCatalog.html. [4] NASA. http://ssd.jpl.nasa.gov/?planet pos.