Abstract
Paper 1116 shrank a uniform ellipsoid within its confocal family without changing gravity seen from outside, and showed numerically that the end point, the core, is a point for a sphere, a segment for a prolate spheroid and a flat disc for an ellipsoid. There is another way to shrink a body: shrink its inside continuously without changing whether loops drawn inside it can shrink (its connectivity). A solid ellipsoid then shrinks to a point, and a solid torus shrinks to the circle at the centre of its tube and stops there. The question of this paper is whether three deformations, the ellipsoid obtained by stretching a sphere, the torus obtained by revolving a circle and the flattened shape obtained by pressing a rice-flour dumpling with a dimple, can be lined up with Dive’s theorem (the force vanishes in the cavity enclosed by a uniform similar shell only for ellipsoidal shells) and the Poincaré conjecture (the only closed three-dimensional space in which every loop shrinks is the three-sphere) as one story about relatives of the sphere. This paper counts the two cores on the same bodies and separates them by the criterion of Paper 300. It claims no new theorem or law, and it is neither a proof nor another reading of the Poincaré conjecture. The gravitational constant and the masses are set to 1, and the fields outside and the forces inside shells were found by adding up the bodies or skins directly. For the flux through the skin in Section 4, the formula for the force inside a uniform ellipsoid was used and compared with the direct sum.
The gravitational core of the ellipsoid (3, 2, 1) is a disc with semi-axes √8 and √3 (dimension 2), while its connectivity core is a point (dimension 0). For a prolate spheroid they are a segment and a point, for a sphere a point and a point; among ellipsoids, only for the sphere can the two cores be placed on one and the same point (this follows directly from the dimension of the core in Paper 1116). The connectivity core of a solid torus (core circle of radius 2, tube radius 1) is a circle, but gathering all of its mass onto that circle changes the force at three outer points (one of them in the hole) by a relative amount of 1.51×10⁻² to 1.09×10⁻¹. With the radius of the circle changed to √(R²+r²/4) = 2.0615528, the number for ℓ = 2 that can be read from outside matches and the difference falls to 4.84×10⁻⁴ to 1.33×10⁻², but the number for ℓ = 4 stays off by −5r⁴/16 (checked symbolically).
For bodies that are star-shaped as seen from the centre, a thin shell was placed between the surface and a similar copy enlarged slightly about the centre (thickness proportional to the distance p from the centre to the tangent plane), and the forces at three inner points were compared. For two ellipsoids they are of order 10⁻¹⁶ to 10⁻¹⁵ (the size of the numerical error). For shapes that are not ellipsoids they do not vanish: 5.63×10⁻⁴ to 8.70×10⁻³ for the convex rounded cube x⁴+y⁴+z⁴ = 1, 2.85×10⁻⁴ to 3.05×10⁻³ for a flattened shape with dimpled poles and 1.38×10⁻³ to 2.93×10⁻² for a peanut shape. Moving from an ellipsoid toward the dimpled shape by a fraction t, the force vanished, among the values of t tried, only at t = 0, and it is nearly proportional to t while t is small. For a torus, p runs from −1 to 3 as seen from the centre and is negative on 0.195501 of the skin, so a similar shell cannot even be formed (with the centre moved to four other places, negative parts remained). A star-shaped body shrinks to a point when its inside is moved straight toward the centre, so the condition on similar shells (Dive’s condition) is stronger than "the connectivity core is a point", and the dimpled shape separates the two.
There are two 4π on a skin. The total curvature is fixed by the connectivity of the skin (Gauss–Bonnet): 4π for relatives of the sphere (bodies without holes), and for a torus 0, the sum of +4π on the outer side and −4π on the inner side. Even for the dimpled shape it is 4π, the sum of a positive part 12.660171 and a negative part −0.093801. The gravitational flux inward through the skin is 4π times the mass inside (Gauss’s law) and does not depend on the connectivity (on a skin enclosing a solid torus it differs from 4π by 5.15×10⁻¹⁴). Paper 128 already wrote that the two 4π part on a torus. This paper counts how, even on ellipsoids where the two totals agree, their places move in opposite directions. When the confocal family of (3, 2, 1) is squeezed toward its core disc, the share of the total curvature in the rim band that occupies 1/10 of the area of the skin rises from 0.359423 to 0.999908, while the share of the flux falls from 0.097399 to 0.034717 (both are 0.1 for a sphere). The 4π of curvature goes to the rim of the core disc, and the 4π of flux goes to the faces, following the mass of the core disc. When a prolate spheroid is thinned from thickness 1 to 0.01, the share of the total curvature at the two ends occupying 1/10 of the area rises from 0.362834 to 0.992155, and the share of the flux falls from 0.089406 to 0.053189.
The three uniqueness statements, the density exponent n = 2 of Paper 01, Dive’s ellipsoid and the three-sphere of the Poincaré conjecture, have the same form, "only this one", but they choose different things (an exponent, the shape of a body in flat space, the space itself) and rest on different things (the area 4πr² of a shell, the Laplace equation of a field of zero mass, connectivity alone), so their roots are distinct. The separator is whether shrinking keeps the gravity outside or keeps the loops that cannot shrink.
Keywords: gravitational core, connectivity core, Dive’s theorem, similar shell, total curvature, Poincaré conjecture, solid torus, separator.
Scope (what is not claimed)
This paper claims no new theorem or law. It is neither a proof, nor a partial proof, nor another reading of the Poincaré conjecture (the three-dimensional case is known to have been settled by the work of Perelman 2002 and 2003). The core of confocal ellipsoids (Paper 1116), the vanishing of the force in the cavity enclosed by a uniform shell between two similar ellipsoids (Newton 1687), that the force vanishes in the cavity enclosed by a uniform similar shell only for ellipsoidal shells (often cited under the names of Dive 1931 and Nikliborc; the original has not been read), that the total curvature is 2π times the Euler characteristic of the skin (the Gauss–Bonnet theorem), that the gravitational flux through a skin is 4π times the mass inside (Gauss’s law), that a star-shaped body can be shrunk to a point, that a solid torus shrinks to the circle at the centre of its tube and no further, and that continuous shrinking does not change the Euler characteristic, are all standard. The relation between the Euler characteristics of a three-dimensional body and of its skin, that a skin of any shape has a density making the force inside vanish if the density of a thin skin may be chosen freely (the distribution of charge on an isolated conductor), and the way curvature sits on the rim of a convex body squeezed flat also appear to be classical (the literature has not been checked here). The gravitational core is used in the sense of Paper 1116 (the core reached by squeezing the confocal family as far as possible), and whether another placement could make it smaller is not asked. The gravitational core of a solid torus is not given. All bodies in this paper are taken to have no cavity inside. What is added is lining up the core that keeps the gravity outside and the core that keeps the connectivity on the same bodies; checking numerically and symbolically that the circle at the centre of a solid torus cannot give its gravity outside; counting how much Dive’s condition fails for star-shaped bodies that are not ellipsoids (a convex rounded cube, a shape with dimpled poles and a peanut shape); and counting that, when the confocal family is squeezed toward its core, the places of the two 4π on the skin move in opposite directions. Paper 1116 showed numerically the core reached by squeezing the confocal family as far as possible, and named Dive in the form that the only uniform bodies whose potential inside is a quadratic are ellipsoids. This paper sits beside it, lines that core up with the connectivity core, and tests Dive’s condition numerically on shapes that are not ellipsoids. Paper 07 wrote, about the three-dimensional Poincaré conjecture, that what drives the convergence is the Ricci flow and simple connectivity, and that the uniformity of n = 2 expresses a property of the three-sphere reached at the end. Paper 67 read 4π as 2πχ, and Paper 92 lined up the 4π of the flux of a point source and the 2πχ of Gauss–Bonnet as two examples of the generalized Stokes theorem. Paper 128 showed that the 4π of the full solid angle and 2πχ agree only in three dimensions and part on a torus. Paper 224 showed numerically that the total angle defect of a polyhedron is 4π and becomes 0 for a torus. Paper 1114 showed numerically that, in the spherical case, a point, a skin and a 1/r² ball give the same field outside. Paper 01 showed that the only density exponent for which the mass of each shell does not depend on the radius is n = 2. Section 4 of this paper adds not the parting of the two 4π itself, but that their places move in opposite directions on ellipsoids where the totals agree. The connections with the Ricci flow, with neck pinching and surgery, and with collapse in Perelman’s proof are not treated. Nor is the reading in which a three-dimensional space is placed as the skin of a body in flat four-dimensional space and a question of Dive’s type is asked one dimension up: that would assume from the start that the space is such a skin.
1. Two ways of shrinking
Paper 1116 shrank a uniform ellipsoid within its confocal family, obtained by subtracting the same number from the squared semi-axes. At equal mass the field outside does not change, and the end point is the point at the centre for a sphere, the segment between the foci for a prolate spheroid and the flat focal disc for an ellipsoid. In this paper the end point of the confocal family is called the gravitational core (as in Paper 1116, whether another placement could make it smaller is not asked).
There is another way to shrink a body: shrink its inside continuously without changing whether a loop drawn inside it can be shrunk to a point inside it. A solid ellipsoid becomes a point when its inside is moved straight toward the centre. A solid torus becomes the circle at the centre of its tube when each cross-section of the tube is shrunk to its centre, but it shrinks no further, because a loop going once around the hole cannot shrink inside the torus. This end point is called the connectivity core. Where the connectivity core is placed is not fixed, but whether it is a point or a circle is. All bodies in this paper are taken to have no cavity inside. A body with a spherical cavity inside has every loop shrinkable, but it does not shrink to a point; it stops at the sphere surrounding the cavity.
The Poincaré conjecture states that the only closed three-dimensional space in which every loop shrinks is the three-sphere, and the three-dimensional case is known to have been settled by the work of Perelman (2002, 2003). It is about the connectivity of a space itself, and sits in a different place from the bodies of this paper. This paper lines up the two cores on the same bodies in flat three-dimensional space. The Poincaré conjecture is used only as the starting point of the question.
The gravitational constant is set to $G=1$ and the masses to 1. The fields outside and the forces inside shells were found by dividing the bodies or skins into small points and adding up $(\mathbf y-\mathbf x)/|\mathbf y-\mathbf x|^3$ felt at the point $\mathbf x$ (Gaussian quadrature in double precision; equal steps along periodic directions). The curvature of a skin was found from the formula for the Gaussian curvature of an ellipsoid and from the formula for a surface of revolution; for ellipsoids it was also counted through the Gauss map (which sends each point of the surface to the direction of its normal), and the two ways were compared. For the gravitational flux through the skin in Section 4, the formula for the force inside and on a uniform ellipsoid, $\mathbf g=-4\pi\rho (N_x x, N_y y, N_z z)$ ($N_i$ are the depolarization factors of Paper 1116), was used. This formula agreed at two inner points with the sum over all directions of the distance from the point to the boundary, by relative amounts of $1.59\times10^{-14}$ and $3.99\times10^{-14}$ (with the grid of that sum made 1.5 times finer; with the first grid the agreement was of order $10^{-10}$).
2. The two cores side by side
The two cores are lined up for each shape. The relative difference in the column of the gravitational core is the difference between the fields that the original uniform shape and its core give at three outer points, divided by the magnitude of the force; it repeats the computation of Paper 1116 with the same tools (all are the size of the numerical error and move with the grid and the tool).
| shape (semi-axes) | gravitational core (dimension) | connectivity core (dimension) | relative difference at 3 outer points |
|---|---|---|---|
| ellipsoid $(3,2,1)$ | disc with semi-axes $\sqrt8$, $\sqrt3$ (2) | point (0) | $1.24\times10^{-14}$ to $3.82\times10^{-14}$ |
| prolate spheroid $(1,1,2)$ | segment of length $2\sqrt3$ (1) | point (0) | $1.58\times10^{-14}$ to $2.51\times10^{-14}$ |
| sphere $(1,1,1)$ | point at the centre (0) | point (0) | $6.90\times10^{-15}$ to $2.79\times10^{-14}$ |
| solid torus | not given here | circle at the centre of the tube (1) | gathered onto the circle: $1.51\times10^{-2}$ to $1.09\times10^{-1}$ |
The three outer points are $(4.0, 0.5, 0.3)$, $(0.5, 3.0, 1.0)$ and $(1.0, 1.0, 2.5)$ for the ellipsoid, $(2.0, 0.3, 0.5)$, $(0.4, 0.2, 3.0)$ and $(1.2, -1.0, 1.5)$ for the prolate spheroid, and $(1.5, 0.2, 0.1)$, $(0.3, -1.1, 0.8)$ and $(2, 2, 2)$ for the sphere. The dimension of the gravitational core is fixed by the shape (Paper 1116), and that of the connectivity core by whether there is a hole. They are 2 and 0 for an ellipsoid and 1 and 0 for a prolate spheroid, and do not agree. With $c$ the shortest semi-axis, the semi-axes $\sqrt{a^2-c^2}$ and $\sqrt{b^2-c^2}$ of the gravitational core both vanish only when all semi-axes are equal, and only then can the two cores be placed on the one point at the centre. This follows directly from the dimension of the core in Paper 1116.
For the torus, whether the circle of the connectivity core carries the gravity outside was counted. A solid uniform torus (mass 1) with core circle of radius $R=2$ and tube radius $r=1$ was compared at three outer points with the circle at the centre of the tube (radius 2, uniform, mass 1) and with a circle whose radius was changed to $\sqrt{R^2+r^2/4}=2.0615528$. The point $(0.3, -0.2, 0.4)$ lies in the hole, outside the body. The force of the torus moved by at most a relative $1.15\times10^{-13}$ when the quadrature grid was made 1.5 times finer.
| outer point | force of the solid torus | relative difference, circle of radius 2 | relative difference, circle of radius 2.0615528 |
|---|---|---|---|
| $(4.0, 0.5, 0.3)$ | 0.0765689339 | $1.51\times10^{-2}$ | $4.84\times10^{-4}$ |
| $(0.5, 1.0, 2.0)$ | 0.0890189524 | $4.53\times10^{-2}$ | $2.42\times10^{-3}$ |
| $(0.3, -0.2, 0.4)$ (in the hole) | 0.0486668377 | $1.09\times10^{-1}$ | $1.33\times10^{-2}$ |
Gathering the mass onto the circle changes the force outside by a relative $1.51\times10^{-2}$ to $1.09\times10^{-1}$, most at the point in the hole. Changing the radius makes the difference smaller at all three points, but not zero. The numbers readable from outside show why. The torus is symmetric about the $z$ axis and under the up-down mirror, so of the numbers per angular pattern readable from outside (Section 4 of Paper 1116) only those of even $\ell$ that do not change under rotation about the axis remain. The averages of the harmonic polynomials over a uniform torus (mass 1) are, symbolically,
$$\ell=2:\ \langle 2z^2-x^2-y^2\rangle=-R^2-\frac{r^2}{4},\qquad \ell=4:\ \langle 35z^4-30z^2|\mathbf x|^2+3|\mathbf x|^4\rangle=3R^4+\frac32R^2r^2-\frac{r^4}{8},$$and for a circle of radius $R'$ they are $-R'^2$ and $3R'^4$. With $R'^2=R^2+r^2/4$, which matches $\ell=2$, the difference for $\ell=4$ is $-5r^4/16$; for $R=2$, $r=1$ the torus has 53.875 and the circle 54.1875. Since the numbers for $\ell=4$ differ, the two fields are not the same outside the sphere of radius $R+r=3$ enclosing the body. The outside of the body is connected all the way into the hole, and the difference of the two fields is harmonic there, so if it vanished on some region it would vanish everywhere; hence the fields are not the same around the hole either. The sizes of the differences at the three points were found numerically. Unless the tube thickness $r$ is 0, a single uniform circle cannot give the field outside. How far the gravitational core of a torus can be shrunk while keeping the field outside is not given here.
The circle of the connectivity core carries the fact that a loop going once around the hole cannot shrink, but it does not carry the gravity outside. For an ellipsoid it is the other way round: the gravitational core spreads into a disc while the connectivity core shrinks to a point. The two cores are fixed by different things.
3. Similar shells and Dive’s condition
Paper 1116 showed numerically that the force inside vanishes for a thin skin on the surface of an ellipsoid whose thickness is proportional to the distance $p$ from the centre to the tangent plane (the homoeoid). This skin is a shell filled uniformly between the original surface and the surface enlarged slightly and similarly about the centre. The same construction works for any body that is star-shaped as seen from the centre. Writing the surface as $r(\theta,\varphi) \mathbf e_r$, the mass of the shell inside the cone of solid angle $d\Omega$ seen from the centre is proportional to $r^3 d\Omega$, and $p dS=r^3\sin\theta d\theta d\varphi$ (three times the volume of the cone). The direction of the surface need not be found by differentiation. That the force vanishes in the cavity enclosed by a uniform similar shell only for ellipsoidal shells is often cited under the names of Dive 1931 and Nikliborc (the original has not been read here). Paper 1116 cited the same names in the form that the only uniform bodies whose potential inside is a quadratic are ellipsoids. The two are restatements of each other. Enlarging a body of the same density by a factor $k$ turns the potential inside into $k^2\Phi(\mathbf x/k)$, so the potential of a thin similar shell is proportional to $2\Phi-\mathbf x\cdot\nabla\Phi$. Its being constant inside is the same as $\Phi$ minus a constant being homogeneous of degree 2 in the coordinates, that is, a quadratic inside (a few lines of computation). This paper calls this Dive’s condition and counts how much it fails for star-shaped bodies that are not ellipsoids.
With the mass of the shell set to 1, the magnitudes of the force at the three inner points $(0.2, 0.1, 0.05)$, $(0.5, -0.2, 0.1)$ and $(-0.3, 0.4, -0.08)$ (inside every shape) were compared. Here $r$ is the distance from the centre and $\theta$ the angle from the $z$ axis.
| shape | surface | force at the 3 inner points |
|---|---|---|
| ellipsoid $(3,2,1)$ | $x^2/9+y^2/4+z^2=1$ | $1.75\times10^{-16}$ to $4.64\times10^{-16}$ |
| ellipsoid $(1.5,1.5,0.7)$ | $(x^2+y^2)/2.25+z^2/0.49=1$ | $1.40\times10^{-15}$ to $2.48\times10^{-15}$ |
| rounded cube (convex) | $x^4+y^4+z^4=1$ | $5.63\times10^{-4}$ to $8.70\times10^{-3}$ |
| flattened shape with dimpled poles | $r=1.5-1.05\cos^2\theta+0.25\cos^4\theta$ | $2.85\times10^{-4}$ to $3.05\times10^{-3}$ |
| peanut shape | $r=1+0.35\cos2\theta$ | $1.38\times10^{-3}$ to $2.93\times10^{-2}$ |
For the two ellipsoids the force inside is the size of the numerical error (it moves with the grid and the tool), and for the three other shapes it does not vanish. The rounded cube is convex, so being convex is not enough for the condition. The dimpled shape has radius 0.7 at the poles and 1.5 at the equator, the same as the ellipsoid $(1.5,1.5,0.7)$. Around each pole it is hollowed like a bowl: at a distance 0.143459 from the axis the surface is at height 0.707704, above the 0.7 of the pole. It is close to a rice-flour dumpling pressed flat with a dimple in the middle.
What this paper counted is the thin shell. If the force vanishes for thin shells, it also vanishes for the thick similar shell obtained by stacking them. In fact, for the thick similar shell filled uniformly between the surface and the surface enlarged 1.3 times, the forces at the three inner points were $1.69\times10^{-15}$ to $4.66\times10^{-15}$ (the size of the error) for the ellipsoid $(1.5,1.5,0.7)$, $1.47\times10^{-4}$ to $1.64\times10^{-3}$ for the dimpled shape and $6.92\times10^{-4}$ to $1.32\times10^{-2}$ for the peanut shape, so they did not vanish for the two shapes that are not ellipsoids.
The force at the point $(0.5, -0.2, 0.1)$ was counted for shapes whose radius mixes the radius of the ellipsoid $(1.5,1.5,0.7)$ and that of the dimpled shape in the ratio $1-t$ to $t$. At $t=0$ it is $1.40\times10^{-15}$, and the force divided by $t$ was 0.004551 at $t=0.05$, 0.004455 at $t=0.1$ and 0.003045 at $t=1$. Among the values of $t$ tried (0, 0.05, 0.1, 0.25, 0.5 and 1), the force was of the size of the numerical error only at $t=0$. While the deviation from the ellipsoid is small, the force appears nearly in proportion to it.
For a torus a similar shell cannot even be formed. At the point $((R+r\cos u)\cos v, (R+r\cos u)\sin v, r\sin u)$ of the surface the outward normal is $(\cos u\cos v, \cos u\sin v, \sin u)$, and the distance from the centre to the tangent plane is
$$p=R\cos u+r.$$For $R=2$, $r=1$, $p$ runs from $-1$ to $3$ and is negative on the inner band where $\cos u<-r/R$. The share of the area where it is negative is $1/3-\sqrt3/(4\pi)=0.195501$. With thickness proportional to $p$, the thickness would be negative there. With the centre moved to $(0, 0, 1)$ on the axis, $(0.5, 0.3, 0.2)$ in the hole, and $(2, 0, 0)$ and $(-1.5, 0.8, 0.4)$ in the tube, the minimum of $p$ was $-3$ to $1-\sqrt5=-1.236068$, and negative parts remained.
A star-shaped body becomes a point when its inside is moved straight toward the centre, so its connectivity core is a point. A torus is not star-shaped from any centre, because a loop going once around the hole cannot shrink. A body satisfying Dive’s condition is star-shaped and its connectivity core is a point. The converse fails: the rounded cube, the dimpled shape and the peanut shape drop out of Dive’s condition while their connectivity cores remain points. Dive’s condition is stronger than the condition on connectivity. Note that if the density of a thin skin may be chosen freely, a skin of any shape, the torus included, has a density that makes the force inside vanish. It is the distribution of charge on an isolated conductor, and since it is nowhere negative, it can be laid on as mass. What Dive’s condition constrains is the distribution with density proportional to $p$, and for an ellipsoid the two coincide. That a torus cannot form a similar shell is because the distribution proportional to $p$ becomes negative, not because no skin with zero force inside exists.
4. The two 4π on a skin
On a skin there are two quantities counted with $4\pi$. One is the total curvature $\int K dA$, which by the Gauss–Bonnet theorem is $2\pi$ times the Euler characteristic $\chi$ of the skin: $4\pi$ for the skin of a relative of the sphere (a body without holes) and 0 for the skin of a torus. The Euler characteristic of the skin of a three-dimensional body is twice that of the body (the literature has not been checked here), and since continuous shrinking does not change the Euler characteristic, that of the body equals that of its connectivity core (1 for a point, 0 for a circle). Hence, for bodies without a cavity, the total curvature is $4\pi$ times the Euler characteristic of the connectivity core. The other is the gravitational flux inward through the skin, which by Gauss’s law is $4\pi$ times the mass inside (standard; for the sphere, Paper 1114 counted it in the field outside). It does not depend on the connectivity of the skin. Paper 92 lined up these two as two examples of the generalized Stokes theorem.
On the skin of a torus $K dA=\cos u du dv$, so the outer half ($K>0$) gives exactly $+4\pi$, the inner half ($K<0$) gives $-4\pi$, and the sum is 0, whatever the radii of the core circle and the tube. On the other hand, when a solid torus (mass 1) is enclosed by the skin of a fatter torus of tube radius 1.5 and the flux through that skin is found from the force obtained by adding up the torus directly, it is $-12.566370614359$ counted positive outward, differing from $-4\pi$ by $5.15\times10^{-14}$. On the same skin the flux is $4\pi$ and the total curvature is 0. That the two $4\pi$ part on a torus was written by Paper 128 as the distinction between the $4\pi$ of the full solid angle and $2\pi\chi$, and Paper 224 showed numerically that the total angle defect of a polyhedral torus is 0.
On the skin of the dimpled shape the sign of $K$ changes at $\theta=0.338312$ and $0.646182$ from the pole. The bowl around the pole has $K>0$, the ring-shaped band outside it is saddle-shaped with $K<0$, and the side toward the equator has $K>0$. The positive part is 12.660171 and the negative part $-0.093801$, and the sum is $4\pi$. Dimpling does not move the total curvature as long as the connectivity stays that of a ball.
Even on ellipsoids, where the two totals agree, the two $4\pi$ sit in different places. The confocal family of Paper 1116 (squared semi-axes $9-\lambda$, $4-\lambda$, $1-\lambda$; semi-axes written $a'$, $b'$, $c'$) was squeezed toward its core disc, and the shares in the rim band were counted. Writing the surface as $(a'\sin\theta\cos\varphi, b'\sin\theta\sin\varphi, c'\cos\theta)$, the rim band is where $|\cos\theta|<s$ ($|z|<sc'$), and $s$ was chosen for each $\lambda$ so that the band occupies exactly 1/10 of the area of the skin ($s=0.1$ for a sphere). If the band is fixed as $|z|<c'/10$, the share of area of the band itself also shrinks from 0.06542408 to 0.01009489, and it cannot be compared directly with the 0.1 of a sphere. The share of the total curvature was counted through the Gauss map and on a grid on the surface, and the two agreed to within $10^{-9}$. From the force formula and $p dS=a'b'c'\sin\theta d\theta d\varphi$ the share of the flux is
$$\frac34\Bigl[(1-N_z)\Bigl(2s-\frac{2s^3}{3}\Bigr)+\frac43N_zs^3\Bigr],$$which agreed to within $10^{-12}$ with the sum on a grid on the surface. The whole flux, counted positive outward, was $-4\pi$ for every ellipsoid.
| $\lambda$ | short semi-axis $c'$ | $s$ of the band | $N_z$ | share of total curvature in the band | share of flux in the band |
|---|---|---|---|---|---|
| 0 | 1 | 0.151349 | 0.5765452609 | 0.35942337 | 0.09739928 |
| 0.75 | 0.5 | 0.216245 | 0.7243051466 | 0.71688402 | 0.09535694 |
| 0.99 | 0.1 | 0.306641 | 0.9302428829 | 0.99034756 | 0.05790161 |
| 0.9999 | 0.01 | 0.316057 | 0.9926283350 | 0.99990805 | 0.03471734 |
| sphere (for comparison) | 1 | 0.1 | $1/3$ | 0.1 | 0.1 |
The more the family is squeezed, the more the total curvature gathers in the band and the more the flux leaves it, falling below the share 0.1 of the area. For a sphere both are 0.1, and for a sphere of mass 1 and radius 1 the density of the flux and the curvature on the skin are both 1 everywhere. In the limit of the disc, the share of area of the band becomes $s^2$, so $s$ tends to $\sqrt{0.1}=0.316228$, and the band corresponds to the ring of the core disc where the distance $q$, measured with the semi-axes of the disc, exceeds $\sqrt{0.9}=0.948683$. The share of the flux tends to the share of mass there, $(1-q^2)^{3/2}=0.1^{3/2}=0.031623$. The $4\pi$ of curvature gathers on the rim of the core disc (where the density $\sqrt{1-q^2}$ of the disc is 0), and the $4\pi$ of flux spreads over the faces, following the mass of the core disc. With the band fixed as $|z|<c'/10$, the directions were the same (total curvature 0.245575 to 0.998991, flux 0.063883 to 0.002095).
The family of prolate spheroids (squared semi-axes $1-\lambda$, $1-\lambda$, $4-\lambda$; thickness $a'$) was squeezed toward the segment between the foci, and the shares at the two ends ($|\sin\theta|<s$, where the distance from the axis is less than $s$ times the thickness) were counted. Here $s$ was chosen so that the two ends occupy exactly 1/10 of the area of the skin.
| thickness $a'$ | $s$ of the ends | share of total curvature at the ends | share of flux at the ends |
|---|---|---|---|
| 1 | 0.517518 | 0.36283411 | 0.08940606 |
| 0.3 | 0.579733 | 0.76675026 | 0.06663533 |
| 0.1 | 0.591060 | 0.92158143 | 0.05606933 |
| 0.01 | 0.592736 | 0.99215546 | 0.05318926 |
The thinner the spheroid, the more the total curvature gathers at the two ends and the more the flux leaves them, falling below the share 0.1 of the area. In the limit of the segment the two ends correspond to where $|z|/c'$ exceeds 0.805384, and the share of the flux tends to the share of the mass of the segment core (density proportional to $1-z^2/c'^2$) beyond that point, 0.053128. For the segment core, the curvature gathers at the two ends and the flux leaves from the sides.
How the curvature is shared along the rim can also be read through the Gauss map. A point of the surface is fixed by its normal $\mathbf n$ alone as $(a^2n_x, b^2n_y, c^2n_z)/\sqrt{a^2n_x^2+b^2n_y^2+c^2n_z^2}$, and the direction angle $\varphi$ of the point and the direction angle $\alpha$ of the normal are related by $\tan\varphi=(b^2/a^2)\tan\alpha$. Hence the share of the total curvature in the sector $|\varphi|<45^\circ$ (on the side $x>0$) is $\arctan(a^2/b^2)/\pi$ whatever $c$ is, and 0.366875 for semi-axes 3 and 2 (checked on a grid on the surface for $c=1$, 0.3 and 0.1 to within $10^{-9}$). An equal share in every direction would be 0.25; more sits at the ends of the long axis, where the rim bends strongly. For the rim of the core disc (semi-axes $\sqrt8$ and $\sqrt3$) it is $\arctan(8/3)/\pi=0.385800$. The total curvature where the direction angle of the normal lies between $\alpha$ and $\alpha+d\alpha$ is the solid angle of the thin lune that the normal sweeps from north to south, $\int_0^\pi\sin\beta d\beta d\alpha=2 d\alpha$, whatever $c$ is. In the flat limit all of it sits on the rim, so each point of the rim carries twice the angle through which the normal of the rim ellipse turns.
5. Three deformations and three uniqueness statements
The three deformations of the question are lined up with the numbers so far.
| deformation | shape | force inside a similar shell | connectivity core | total curvature of the skin |
|---|---|---|---|---|
| stretch a sphere | ellipsoid | of order $10^{-16}$ to $10^{-15}$ | point | $4\pi$ |
| press a dumpling with a dimple | flattened shape with dimpled poles | $2.85\times10^{-4}$ to $3.05\times10^{-3}$ | point | $4\pi$ |
| revolve a circle | solid torus | cannot be formed ($p<0$ on 0.195501 of the skin) | circle | 0 |
Stretching a sphere keeps both Dive’s condition and the connectivity. Pressing a dumpling with a dimple keeps the connectivity but drops Dive’s condition. Revolving a circle into a torus drops both. The dimpled shape separates the two conditions.
This paper lines up three statements of the form "only this one". Paper 01 showed that the only exponent of the density $1/r^n$ for which the mass of each shell does not depend on the radius is $n=2$. Dive’s theorem states that the force vanishes in the cavity enclosed by a uniform similar shell only for ellipsoidal shells. The Poincaré conjecture states that the only closed three-dimensional space in which every loop shrinks is the three-sphere. The three have the same form but choose different things. Paper 01 chooses an exponent of the density and rests on the area $4\pi r^2$ of a shell. Dive’s theorem chooses the shape of a body in flat space and rests on the Laplace equation of a field of zero mass (Paper 1116 showed numerically that for the Yukawa field a force appears even inside an ellipsoidal shell). The Poincaré conjecture chooses the space itself; it can be stated with connectivity alone and uses no field. Paper 07 wrote, about the three-dimensional Poincaré conjecture, that what drives the convergence is the Ricci flow and simple connectivity, and that the uniformity of $n=2$ expresses a property of the three-sphere reached at the end. The reading of this paper, which lines up the three as different things, does not conflict with it.
From the numbers and the arrangement of this paper, no straight road is found from Dive’s condition to the Poincaré conjecture. Dive’s condition selects bodies in flat space that have an outside, by the field the body makes. The space of the Poincaré conjecture has no outside, and its statement uses no field. The numbers of the dimpled shape show that Dive’s condition cannot be read as a condition on connectivity.
6. Same root or distinct roots
The quantities of this paper are sorted by the criterion of Paper 300.
(1) The gravitational core and the connectivity core: they keep different things (the field outside, and the loops that cannot shrink). Their dimensions do not agree either, 2 and 0 for an ellipsoid and 1 and 0 for a prolate spheroid, and the circle of the connectivity core of a torus cannot give the field outside. Distinct roots. Among ellipsoids, only for the sphere can the two be placed on one and the same point.
(2) Dive’s condition and the condition on connectivity: a body satisfying Dive’s condition has a point as its connectivity core. The converse fails for the rounded cube, the dimpled shape and the peanut shape. Dive’s condition rests on a field equation, while the condition on connectivity uses no field. Distinct roots.
(3) The two $4\pi$ on a skin: the total curvature is fixed by the connectivity and the flux by the mass inside. Distinct roots (Paper 128), and even on ellipsoids where the totals agree, squeezing moves their places in opposite directions. On a sphere the two distributions are both uniform and coincide, and for mass 1 and radius 1 the values coincide too, being 1 everywhere on the skin.
(4) The three uniqueness statements: the form "only this one" is the same, but what they choose and what they rest on differ. Distinct roots.
7. The separator
The separator is whether shrinking keeps the gravity outside or keeps the loops that cannot shrink. Keeping the gravity outside, an ellipsoid stops at a disc and a prolate spheroid at a segment, and the dimension of the end point is fixed by the shape. Keeping the loops that cannot shrink, every star-shaped body shrinks to a point and a torus stops at a circle, and the end point is fixed by the holes. Among ellipsoids, only for the sphere can the two be placed on the one point at the centre. On the skin, the $4\pi$ of flux belongs to the former and the $4\pi$ of total curvature to the latter, and squeezed toward the core, the flux goes to the faces following the mass of the core while the total curvature goes to the rim of the core.
Closing
The core reached by shrinking while keeping the gravity outside and the core reached by shrinking while keeping the connectivity keep different things, and their roots are distinct. The gravitational core of a uniform ellipsoid is a flat disc, and its connectivity core is a point. The connectivity core of a solid torus is a circle, and gathering the mass onto that circle changes the gravity outside. Among the shapes tried, the force inside a similar thin shell vanished only for ellipsoids; the dimpled shape drops out of this condition while keeping the connectivity of a ball, and a torus cannot form a similar shell. The 4π of the total curvature of the skin and the 4π of the gravitational flux through it also go to opposite places when squeezed toward the core. The sphere is the shape on which the two cores can be placed on the one point at the centre and the distributions of the two 4π coincide on the skin (for mass 1 and radius 1 the values coincide too, at 1). This paper shows nothing about the Poincaré conjecture.
References
- Dive, P. (1931). Attraction des ellipsoïdes homogènes et réciproques d’un théorème de Newton. Bulletin de la Société Mathématique de France.
- Newton, I. (1687). Philosophiæ Naturalis Principia Mathematica.
- Perelman, G. (2002). The entropy formula for the Ricci flow and its geometric applications. arXiv:math/0211159.
- Perelman, G. (2003). Ricci flow with surgery on three-manifolds. arXiv:math/0303109.
- Yamagishi, Y. (2026). The Uniqueness Theorem of $n = 2$ for Spherically Symmetric Density Distributions [Paper 01]. YOI: https://www.yamagishiyuuki.com/yoi/1; DOI: 10.5281/zenodo.21403186
- Yamagishi, Y. (2026). A Structural Interpretation of the 3D Poincaré Conjecture via the n=2 Uniqueness Theorem and Perelman’s Ricci Flow [Paper 07]. YOI: https://www.yamagishiyuuki.com/yoi/7; DOI: 10.5281/zenodo.21404387
- Yamagishi, Y. (2026). The Euler-Characteristic Origin of the $4\pi$ Patch [Paper 67]. YOI: https://www.yamagishiyuuki.com/yoi/67; DOI: 10.5281/zenodo.21642170
- Yamagishi, Y. (2026). The Shared Alphabet [Paper 92]. YOI: https://www.yamagishiyuuki.com/yoi/92; DOI: 10.5281/zenodo.21758486
- Yamagishi, Y. (2026). $4\pi=2\pi\chi$ Holds in Three Dimensions and Nowhere Else [Paper 128]. YOI: https://www.yamagishiyuuki.com/yoi/128; DOI: 10.5281/zenodo.22079576
- Yamagishi, Y. (2026). Flat Wherever You Touch It, and Four Pi in Total [Paper 224]. YOI: https://www.yamagishiyuuki.com/yoi/224; DOI: 10.5281/zenodo.22120836
- Yamagishi, Y. (2026). The Three Hundredth Census: Whether Two Things Share a Root Is Decidable [Paper 300]. YOI: https://www.yamagishiyuuki.com/yoi/300; DOI: 10.5281/zenodo.22171914
- Yamagishi, Y. (2026). m = 4π Is the Mass of a Skin of Radius 1 at Density 1, and the 1/r² Ball Is That Skin Puffed Out Evenly [Paper 1114]. YOI: https://www.yamagishiyuuki.com/yoi/1114
- Yamagishi, Y. (2026). A Uniform Ellipsoid Pulls from Outside Like a Flat Focal Disc, and a Sphere Pulling Like a Point Is the Case in Which That Disc Shrinks to One Point [Paper 1116]. YOI: https://www.yamagishiyuuki.com/yoi/1116
Reference
Yamagishi, Y. (2026). The Gravitational Core and the Connectivity Core Have Distinct Roots, and Among Ellipsoids They Meet in One Point Only for the Sphere [Paper 1118]. YOI: https://www.yamagishiyuuki.com/yoi/1118
ORCID: 0009-0006-8587-3450
PDF (English) · PDF (Japanese) · Date proof of the English PDF (OpenTimestamps .ots) · SHA256SUMS
The record page for this paper (in Japanese, with the PDFs in each language and the date proof): YOI 1118. This paper is a preprint and has not been peer reviewed.