作者君在作品相關中其實已經解釋過這個問題。
不過仍然有人質疑——“你說得太含糊了”，“火星軌道的變化比你想象要大得多！”
那好吧，既然作者君的簡單解釋不夠有力，那咱們就看看嚴肅的東西，反正這本書寫到現在，嚷嚷著本書BUG一大堆，用初高中物理在書中挑刺的人也不少。
以下是文章內容：
&n integrations and stability of planetary orbits in our&n
Abstract
We present tts of&n&nerical integrations of planetary¬ions 109&nespans including all nine planets. A quick inspection of&nerical data¬ion， at least in&nple&nical&nodel，&ns to be quite stable even &nespan. A closer look at towestfrequency oscillations using a lowpass filter sly diffusive c¬ion， especially that of Mercury. The behaviour of the eccentricity of Mercury in our integrations is qualitatively&nilar to&n Jacques Laskar's secular perturbation theory&nax∼ 0.35 ∼± 4 Gyr. However， tar increases of eccentricity or inclination in any&nents of tanets， wed by still&n&nerical integrations. We&ned a couple of trial integrations¬ions of tanets the duration of ± 5 × 1010 yr. Tt indicates that the&najor resonances in&n have&naintained the&nespan.
1 Introduction
1.1Definition&n
Tity of our&n hundred years，&n has&nany&nous&nathematicians tayed a central role&nent of nonlinear&nics and chaos theory. However， we do not yet have a definite answer to the question of&n is stable or not. Ty a result of the fact that the definition of tity’ is vague wation&n of¬ion in&n. Actually it is not easy to give a clear， rirous and&neaningful definition of tity of our&n.
&nong&nany definitions of stability， l definition&nan 1993: actually this is not a definition of stability， but of instability. We define&n&ning unstable wose encounter&newhere in&n，&n a certain initial configuration&np;amp; Boss 1996;&np;amp; Tanikawa 1999.&n is defined as experiencing a close encounter when two bodies approach one another witarger Hill radius. Otherwise&n is defined as being stable. Henceforward we state tanetary&n&nically stable if no close encounter happens during&n， about ±5 Gyr. Incidentally， taced by one in which an occurrence of any orbital crossing between eitanets takes place. This is because we&n experience t crossing is very likely to lead to a close encounter in planetary and protoplanetary&ns Yoshinaga，&np;amp; Makino 1999. Of course ty applied&ns wite orbital resonances&n.
1.2Previous studies&ns of this research
In addition to the vagueness of tity， tanets in our&n s&nical&nan&np;amp;&n 1988， 1992. The cause of this cy understood as being a result of resonance lapping&np;amp;&nan 1999; Lecar，&np;amp;&nan 2001. However， it rating &nble of planetary&ns including all nine planets for a period cing several 10 Gyr to ty understand&n evolution of planetary orbits， since&ns are characterized by their strong dependence on initial conditions.
&n that point of&nany&n&nerical integrations included only&nan&np;amp;&n 1988;&np;amp; Nakai 1996. T periods of tanets&nuconger than tanets that&nuclow&n for a given integration period. At present，&nerical integrations publiss are those of&np;amp; Lissauer 1998. Although&nain target was the effect&nainsequence&nass loss on tity of planetary orbits，&ned&nany integrations cing up to ∼1011 yr¬ions of tanets. T&nents&nasses of planets are&ne as&n in&np;amp; Lissauer's paper， but they decrease tly in&nents. This is because they consider the effect&nainsequence&nass loss in ty， they found te of planetary orbits， w indicator of&nescale， is quite sensitive to the rate&nass decrease of the Sun. When&nass of tose to its present value，&nain stable 1010 yr， or peronger.&np;amp; Lissauer&ned&nilar&nents¬ion of seven planets Venus to Neptune， which c a span of ∼109 yr.&nents on tanets are not&nprehensive， s t planets&nain stable during the integration&naintaining&nost regular oscillations.
On the other ytical secular perturbation theory Laskar 1988， Laskar finds targe and irregular variations can appear in tinations of t planets， especially of Mercury and Mars&nescale of several 109 yr Laskar 1996. Tts of Laskar's secular perturbation td&ned and investigated erical integrations.
&ninary results of&n&nerical integrations on all nine planetary orbits， cing a span of several 109 yr， and of trations cing a span of ± 5 × 1010 yr. T&ne for all integrations&nore t dedicated PCs and workstations. One of t conclusions of&n integrations is&n¬ion&ns to be stas of tl&nentioned above， at least&nespan of ± 4 Gyr. Actually， in&nerical integrations&n was&nore stable than wl stability criterion: not only did no close encounter happen during tso all tanetary&nents have been confined in a narroion e and&nain，¬ions are stochastic. Since the purpose of this paper is to exts of&n&nerical integrations， we&nple figures as evidence&n stability of&n¬ion. For readers who&nore specific and deeper interests in&nerical results， we have prepared a e access&nents， towpass filtered results， variation of&nents and&nomentum deficit， and results of&nple&ne–frequency analysis on all of our integrations.
In Section 2 we briefly explain&nical&nodel，&nerical&net conditions used in our integrations. Section 3 is devoted to a description of tts of t integrations.&n stability of&n¬ion is apparent botanetary positions and&nents. A roug errors is also given. Section 4 es on to a discussion of&n variation of planetary orbits using a lowpass filter and includes a discussion of&nomentum deficit. In Section 5， we present a set&nerical integrations for tanets that spans ± 5 × 1010 yr. In Section 6 we also discuss&n stability of¬ion and its possible cause.
2 Description of t integrations
（本部分涉及比較複雜的積分計算，作者君就不貼上來了，貼上來了起點也不一定能成功顯示。）
&nerical&nethod
We utilize a secondorder&n–Holman&nplectic&nap&nain&netman 1991; Kinoshita，&np;amp; Nakai 1991 wit startup procedure to reduce te&n start’(Saha&np;amp;&naine 1992， 1994.
T integrations is 8 d tl integrations of tanets N±1，2，3， which is about 1/11 of t period of tanet Mercury. As for&nination of stepsize， we partly follow t integration of all nine planets&nan&np;amp;&n 1988， 7.2 d and&np;amp;&naine 1994， 225/32 d. We rounded t part of the their stepsizes to&nake ttiple of 2 in order to reduce tation of roundoff error in tation&nan&ned&nerical integrations of tanetary orbits using&nap with a stepsize of 400 d， 1/10.83 of t period of Jupiter.&ns to be accurate enougy justifies&nethod&nining the stepsize. However， since the eccentricity of Jupiter ∼0.05&nucler than that of Mercury ∼0.2， we&ne care when&npare ty&ns of stepsizes.
In the integration of tanets F±， we fixed the stepsize at 400 d.
We adopt Gauss' f and g functions in&nap&netver for Kepler equations.&naximum iterations we set in&nethod is 15， but they never reached&naximum in any of our integrations.
T of the data output is 200 000 d ∼547 yr for tculations of all nine planets N±1，2，3， and about 8000 000 d ∼21 903 yr for the integration of tanets F±.
Alttering was done w integrations were in process， we applied a lowpass filter to t data after we eted all tculations. See Section 4.1&nore detail.
2.4&nation
2.4.1 Relative errors in total energy and&nomentum
According to one of tectic integrators， wly conservative quantities well total orbital energy and&nomentum，&n&nerical integrations&n to have&ned witl errors. Tative errors of total energy ∼10−9 and of total&nomentum ∼10−11 y constant throughout the integration period Fig. 1. T startup procedure，&n start， would ative error in total energy by about one order&nagnitude&nore.
&nerical error of t&nomentum δA/A0 and t energy δE/E0 in&nerical integrationsN± 1，2，3， where δE and δA are tute c energy and total&nomentum， respectively， andE0andA0are t values. T unit is Gyr.
Note that different operating&ns，&nat libraries， and different t in&nerical errors， through the variations in roundoff error ing&nerical alrit of Fig. 1， nize&nerical error in t&nomentum， wd be rirously preserved&nachineε precision.
2.4.2 Error in planetary longitudes
Since&naps preserve total energy and total&nomentum of&nical&ns iny ree of their preservation&nay not be&neasure of t integrations， especially&neasure of t error of planets， i.e. tanetary longitudes.&nate t error in tanetary longitudes，&ned tlo procedures.&npared tt&nain&n integrations&ne test integrations， which&nuch shorter periods but her accuracy than&nain integrations. For this purpose，&ned&nuch&nore accurate integration with a stepsize of 0.125 d 1/64&nain integrations spanning 3 × 105 yr， starting wit conditions as in the N−1 integration. We consider that this test integration provides us witution of planetary orbital evolution. Next，&npare the test integration with&nain integration， N−1. For the period of 3 × 105 yr， we see a difference&nean&nalies of the Earth between the trations of ∼0.52°(in the case of the N−1 integration. Tated to tue ∼8700°， about 25 rotations of Earth after 5 Gyr， since tongitudes increases linearly&nap.&nilarly， tongitude error of Pluto can&nated as ∼12°. Tue for Pluto&nuct in&np;amp; Nakai 1996 where the difference&nated as ∼60°.
&nerical results – I. Glance at the raw data
In&n stability of planetary¬ion&ne snaps data.¬ion of planets indicates&n stability in all of&nerical integrations: no orbital crossings nor close encounters between any pair of planets took place.
3.1 General description of tity of planetary orbits
First， we briefly look at&n stability of planetary orbits. Our interest arly on t planets for&nescales&nuch shorter than tanets. As we can see&n tanar orbital configurations s positions of t planets differ little between t and final part of eac integration， w Gyr. Tid lines denoting tanets&nost within&n of dots even in t part of integrations b and d. This indicates that throughout the entire integration&nost regular variations of planetary¬ion&nain nearly&ne as they are at present.
Vertical view of tanetary&n the z axis direction at t and final parts of the integrationsN±1. The axes units are au. Tane is set to tane of&n total&nomentum.(a T part ofN+1  t = 0 to 0.0547 × 10 9 yr.(b T part ofN+1  t = 4.9339 × 10 8 to 4.9886 × 10 9 yr.(c T part of N−1 t= 0 to −0.0547 × 109 yr.(d T part ofN−1  t =−3.9180 × 10 9 to −3.9727 × 10 9 yr. In eac， a total of 23 684 points are plotted wit of about 2190 yr 5.47 × 107 yr . Solid lines in eac denote the present orbits of t planets&n DE245.