Universal Gravity without Shell Theorem

  • #1
physics4all
7
1
TL;DR Summary
I wondered what would happen if one tried to calculate Newtonian Gravity between two bodies but in many-to-many style (like all atoms to all atoms), i.e. completely ignoring
Newton’s Shell theorem. Did not find anything online about anybody trying that.
So, I tried it myself and got curious results
I couldn't get Latex text to render. Sorry if text looks unreadable...

[Mentor Update -- OP has updated the below text using LaTeX]

Thus, for two perfect spheres with radiuses r and s, densities Pr and Ps, which are positioned in space at distance d between their respective centers, expression for Universal Gravity I got looks this way
(1) ##F = GPsPrπ^2Gg ##
Where Gg is a function of r,s and d and looks like this:

(2) \begin{align}
Gg = \frac{1}{30d}(&\nonumber \\
&(4d^3sr+44dsr^3+44ds^3r)\nonumber \\
&+(-20d^2r^3+20s^2r^3-4r^5)(ln(d+s+r)+ln(d+s-r)-ln(d-s+r)-ln(d-s-r))\nonumber \\
&+(-20d^2s^3-4s^5+20s^3r^2)(ln(d+s+r)-ln(d+s-r)+ln(d-s+r)-ln(d-s-r))\nonumber \\
&+(d^5-10d^3s^2-10d^3r^2-15ds^4-15dr^4+30ds^2r^2)(ln(d+s+r)-ln(d+s-r)-ln(d-s+r)+ln(d-s-r)))
\end{align}
Took many steps to find the integral. Probably need to post somewhere separately…

Took me some time to find a precise enough calculator for natural logs. As expected, once plugging in numbers for all planets in Solar system, plus spheres Cavendish used in his experiment, Gg is always equal to
(3) ##\frac{16s^3r^3}{9d^2}##
Plugging into (3) into (1) results in standard looking
(4) ## F = \frac{GMm}{d^2} ##
Indeed, using Taylor series to model each natural log, and simplifying (2), yields a very dominant (3) with extra ignorable branches which get smaller and smaller with every Taylor iteration.

However, interesting is what happens when d is so big compared to s and r, that it can be considered ## \infty ## . Lets refactor logs as following:
\begin{align}
(5) &(ln(d+s+r)+ln(d+s-r)-ln(d-s+r)-ln(d-s-r)) =\nonumber \\
&= ln(d^2+s^2-r^2+2ds) - ln(d^2+s^2-r^2-2ds) =\nonumber \\
&= ln(d^2+s^2-r^2+2ds) - ln(d^2+s^2-r^2-2ds) + ln(d^2+s^2-r^2) - ln(d^2+s^2-r^2) =\nonumber \\
&= ln(\frac{d^2+s^2-r^2+2ds}{d^2+s^2-r^2}) - ln(\frac{d^2+s^2-r^2-2ds}{d^2+s^2-r^2}) =\nonumber \\
&= ln(1 + \frac{2ds}{d^2+s^2-r^2}) - ln(1 - \frac{2ds}{d^2+s^2-r^2})
\end{align}

Now, applying limit
\begin{align}
(6.1) lim_{d \rightarrow +\infty}{ ( ln(1 + \frac{2ds}{d^2+s^2-r^2}) - ln(1 - \frac{2ds}{d^2+s^2-r^2})) }= ln(e^\frac{2ds}{d^2+s^2-r^2}) - ln(e^\frac{-2ds}{d^2+s^2-r^2}) = \frac{4ds}{d^2+s^2-r^2} \nonumber
\end{align}

The same transformations for other log expressions
(6.2) ## lim_{d \rightarrow +\infty}{ (ln(d+s+r)-ln(d+s-r)+ln(d-s+r)-ln(d-s-r)) }= \frac{4dr}{d^2-s^2+r^2} ##
(6.3) ## lim_{d \rightarrow +\infty}{ (ln(d+s+r)-ln(d+s-r)-ln(d-s+r)+ln(d-s-r)) } = \frac{4sr}{d^2-s^2-r^2}##

Doing the same transformations to other logs and plugging everything into (1) and simplifying, produces (I skip intermediate steps, otherwise text seems to be too busy)
\begin{align}
(7) Gg = &\frac{ \frac{160ds^3r^3}{d^2 +s^2 -r^2}+\frac{160ds^3r^3}{d^2 -s^2 +r^2}- \frac{48ds^3r^3}{d^2 -s^2 -r^2}+\frac{96dsr^5}{d^2 -s^2 -r^2}-\frac{96dsr^5}{d^2 +s^2 -r^2}+\frac{96ds^5r}{d^2 -s^2 -r^2}-\frac{96ds^5r}{d^2 -s^2 +r^2} } {30d}
\end{align}

Whichin case of d >> s > r is roughly
(8) ##\frac{272s^3r^3}{30d^2} ##

(8) is 5.1 times bigger than (3), making Gravitational pull 5+ times stronger at a very long distance, even though still very small. So matter seems 5 times heavier. Pretty much the same ratio as Dark matter to baryonic matter that is needed to explain higher velocity.

Now, orbital velocity based on (1) I believe would be (need to remove density of orbiting body and its density)
(9) ##V = \sqrt{\frac{GPsπGgd}{\frac{4}{3}πr^3}} ##
Lets take derivative of (9) (ignoring G, Ps, and π for simplicity) from above:
\begin{align}
(10)
V’=&\frac{((5d^4+(-30s^2-30r^2)d^2-15s^4+30r^2s^2-15r^4)ln(\frac{(d-s-r)(d+s+r)}{(d-s+r)(d+s-r)})\nonumber \\
-40s^3dln(\frac{(d-s+r)(d+s+r)}{(d-s-r)(d+s-r)})\nonumber \\
-40r^3dln(\frac{(d+s-r)(d+s+r)}{(d-s-r)(d-s+r)})\nonumber \\
+20rsd^2+60rs^3+60r^3s)}\nonumber \\
&{4\sqrt{10}r^\frac{3}{2}\sqrt{(d^5-10s^2d^3-10r^2d^3-15s^4d+30r^2s^2d-15r^4d)ln(\frac{(d-s-r)(d+s+r)}{(d-s+r)(d+s-r)})\nonumber \\
+(-20s^3d^2-4s^5+20r^2s^3)ln(\frac{(d-s+r)(d+s+r)}{(d-s-r)(d+s-r)})\nonumber \\
+(-20r^3d^2+20r^3s^2-4r^5)ln(\frac{(d+s-r)(d+s+r)}{(d-s-r)(d-s+r)})\nonumber \\
+(4rsd^3+44rs^3d+44r^3sd)}}
\end{align}

Using Taylor series, produces exactly the same negative acceleration as 1\nonumber \\sqrt(d) would do (plus some small ignorable branches):
(11) ## V’ = -\frac{0.6\sqrt(s^3)}{d\sqrt(d)} ##
However, using the same log transformations (5), at very large distance d, acceleration flips to positive and becomes a very small
(12) ##V’ = \frac{5\sqrt(s^3)}{3d\sqrt(d)} ##
In other words, at very large distances, Gravitational pull becomes more efficient and makes a smaller body to rotate around a much bigger body with pretty much constant velocity.

Granted, Universal Gravity is largely abandoned in favor of General Relativity, but still interesting.

P.S. After I wrote this text,, an idea came to see what happens at distances very close to a massive body, i.e. d~(s+r). Turns out in this case (1) is bigger than (4) too. Exactly as Einstein predicted. Another coincidence…
 
Last edited by a moderator:
  • Sad
Likes Motore and PeroK
Physics news on Phys.org
  • #2
You can simplify the first expression analytically and it's done by students all over the world as part of their coursework. It produces the expected result. If your approximation produces 5 times the result of the exact calculation then you must have made a calculation error in your approximation. It's that simple.

You took a needlessly complicated approach to the whole problem, so I don't know where the error was introduced.
 
  • Like
Likes ohwilleke, Vanadium 50 and PeroK
  • #3
This is exactly what I did. 6 iterations of Taylor series for natural logs is enough to confirm this long expression converges on form (3) as expected.
Then I noticed at distances d very super large comparing to both s and r, all 3 log combinations begin to grow as power of e with ultimate limit at 5 times greater than (3).
I tried to build a custom calculator based on __float128. Whatever log implementation C++’s stl got, shows exponential growth for Milky Way starting somewhere in vicinity of kilo parsecs. Beyond that even float128 cannot handle. So I decided to ask here…
 
  • #4
The lesson here is that if your work is needlessly messy and complicated, you are more likely to make mistakes. So don't do that.
 
  • Like
Likes ohwilleke, mfb and pinball1970
  • #5
Not to be rude… this is not the lesson. At least not until an error is pointed out. If my log transformations in 6.1-6.3 correct, would not take more then 15 mins to simplify. Or even plug in some numbers straight up and do simple math to confirm.
Just don’t know personally anybody good in math to ask for a favor and check:(
I understand though, not seeing entire chain of simplifications does not help. I tried to post everything, but then text becomes many pages long…
 
Last edited:
  • #6
Um...yes it is.

Fitst, a theorem is a theorem.

If you want people to help find your error, you need to be clear and methodical. For example, LaTex. You probably have a 50-50 chance that the person who spots your mistake first will be you.

If someone told you they were trying to multiply two 5 digit numbers but kept getting the wrong answer when using Roman Numerals, what would you tell them?
 
  • #7
physics4all said:
Not to be rude… this is not the lesson. At least not until an error is pointed out.
That's a lot of work to little purpose. Finding errors is hard.
 
  • Like
Likes Vanadium 50
  • #8
Vanadium50 makes good point that putting this in LaTex would certainly facilitate others (not to mention yourself) reviewing your work. LaTex is de regueur on PF.
 
Last edited:
  • #9
Well, fair is fair… will work on it.

[Mentor Update -- OP has edited the first post above to use LaTeX now.]
 
Last edited by a moderator:
  • Like
Likes berkeman
  • #10
Vanadium 50 said:
The lesson here is that if your work is needlessly messy and complicated, you are more likely to make mistakes. So don't do that.
true th
PeroK said:
That's a lot of work to little purpose. Finding errors is hard.
I'm not looking for somebody to review arithmetic:)

I was doing something else, and certain things began to remind something. Somehow that expression, when Plugged into Newtonian Gravity matches (almost to tenth) "missing" matter and constant speed (or rather miniscule positive acceleration) at the edge of a galaxy.
Curious. Was hoping somebody saw something similar somewhere for an additional reading
 
  • #11
Vanadium 50 said:
Um...yes it is.

Fitst, a theorem is a theorem.

If you want people to help find your error, you need to be clear and methodical. For example, LaTex. You probably have a 50-50 chance that the person who spots your mistake first will be you.

If someone told you they were trying to multiply two 5 digit numbers but kept getting the wrong answer when using Roman Numerals, what would you tell the
physics4all said:
true th

I'm not looking for somebody to review arithmetic:)

I was doing something else, and certain things began to remind something. Somehow that expression, when Plugged into Newtonian Gravity matches (almost to tenth) "missing" matter and constant speed (or rather miniscule positive acceleration) at the edge of a galaxy.
Curious. Was hoping somebody saw something similar somewhere for an additional reading

Vanadium 50 said:
Um...yes it is.

Fitst, a theorem is a theorem.

If you want people to help find your error, you need to be clear and methodical. For example, LaTex. You probably have a 50-50 chance that the person who spots your mistake first will be you.

If someone told you they were trying to multiply two 5 digit numbers but kept getting the wrong answer when using Roman Numerals, what would you tell them?

fair point. Converted to LaTeX for easier reading I hope... Would not think much of it, but curiously matching results...
 
  • #12
physics4all said:
I was doing something else, and certain things began to remind something.
There's a derivation of Newton's Shell Theorem here, for example:

https://www.math.ksu.edu/~dbski/writings/shell.pdf

That proves the result. Any derivation of an alternative result (starting with Newton's Law of Gravity) must be in error.
 
  • Like
Likes Vanadium 50
  • #13
PeroK said:
That proves the result. Any derivation of an alternative result (starting with Newton's Law of Gravity) must be in error.
Vanadium 50 said:
a theorem is a theorem.
 
  • #14
PeroK said:
There's a derivation of Newton's Shell Theorem here, for example:

https://www.math.ksu.edu/~dbski/writings/shell.pdf

That proves the result. Any derivation of an alternative result (starting with Newton's Law of Gravity) must be in error.
Thank you!

I was looking for something similar to this essay, although not that.
Author(s) matched Newton’s pre-calculus geometry with Energy rather than. Force. There are some steps skipped and even though I think I understand where derivations are coming from, would not want to presume. Is there a follow up to this paper to Universal Gravity?

Still, I got results matching UG without shell theorem. Surprising result I got is at very large distances, orbital acceleration very rapidly switches from negative to positive (with limit -> 0). That’s what got my attention to look further and that’s how (again surprise) found orbital speed is at level as is mass is 5+ times larger matching estimates given for Dark matter…
Might be an error indeed. But too many coincidences keep me glued to this
 
  • Skeptical
Likes berkeman

Similar threads

Back
Top