8

For Triangle Centers, as seen at the Encyclopedia of Triangle Centers, the various centers each have a triangle center function $f(a,b,c)$ that is homogeneous, bisymmetric, and cyclic within barycentric or trilinear coordinates. Here are functions for centers known to Euclid.

      name               trilinear  barycentric
X_1   incenter       I   1          a                 angle bisectors  
X_2   centroid       G   1/a        1                 medians
X_3   circumcenter   O   cos(A)     a^2(b^2+c^2-a)    perpendicular bisectors  
X_4   orthocenter    H   sec(A)     tan(A)            altitudes   

Are there tetrahedron center functions similar to the triangle center functions in barycentric or trilinear coordinates? The barycentric centroid and trilinear incenter are known.

enter image description here

I've made an elaborate Tetrahedron Centers demonstration, and posted code showing $4 \pi =2 \sum dihedral - \sum solid $ . At fermat point I give exact coordinates of centers for a specific tetrahedron. At Dihedral Constant Center some exact coordinates are calculated for a new tetrahedron center.

Some available items for tetrahedron $ABCD$ are :
volume, total surface area, total perimeter, dihedral constant.
solid angle $A$, face area $A$ ($\triangle BCD$), perimeter $A$ ... $B$ ... $C$ ... $D$
edge length $ab$, dihedral angle $ab$ ... $ac$ ... $ad$ ... $bc$ ... $bd$ ... $cd$

I'd like to get a list of tetrahedron center functions started.

Ed Pegg
  • 20,955

2 Answers2

8

For the comparatively-easy ones ...

As in my other answer, given tetrahedron $PABC$, I use barycentric coordinates $\rho$, $\alpha$, $\beta$, $\gamma$ to parameterize the point of interest as $$\frac{\rho\,P + \alpha\,A + \beta\, B + \gamma\,C}{\rho + \alpha + \beta + \gamma}$$

We'll take $V$ to be the volume of the tetrahedron. Other elements (side-lengths, face-areas, dihedral angles) are as labeled in the figure:

enter image description here

Of note:

  • Opposite edge-pairs are $(a,d)$, $(b,e)$, $(c,f)$.
  • Edges surrounding vertices $P$, $A$, $B$, $C$ are $(a,b,c)$, $(a,e,f)$, $(d,b,f)$, $(d,e,c)$.
  • Edges surrounding faces $W$, $X$, $Y$, $Z$ are $(d,e,f)$, $(d,b,c)$, $(a,e,c)$, $(a,b,f)$.

See the addendum to this recent answer about how to convert edge-length expressions into hedronometric (face-area-based) forms, which may be more appropriate in this context. (Finding the "best" hedronometric form is something of a challenge.)


Centroid ($G$)

$$\rho : \alpha : \beta : \gamma \;=\; 1 : 1 : 1 : 1$$


Incenter ($I$)

$$\rho : \alpha : \beta : \gamma \;=\; W : X : Y : Z$$


Circumcenter ($O$)

$$\begin{align}\rho &=\phantom{+} a^2 d^2 \left(-d^2 + e^2 + f^2 \right) \\ &\phantom{=}+ b^2 e^2 \left(\phantom{-}d^2 - e^2 + f^2 \right) \\[4pt] &\phantom{=}+ c^2 f^2 \left(\phantom{-}d^2 + e^2 - f^2 \right) \\[4pt] &\phantom{=}- 2 d^2 e^2 f^2 \\[8pt] &= 18 V^2 - a^2\,W X \cos D - b^2\,W Y \cos E - c^2\,W Z \cos F \end{align}$$


Nine/Twelve-Point Center ($T$)

(I'm taking this to mean the center of sphere through the centroids of the faces, which also contains other points of interest, as described in the Wikipedia "Tetrahedron" entry.)

$$\begin{align}\rho \;=\; &\phantom{-\;} 2 a^2 d^2 \left(-a^2 + b^2 + c^2 \right) + a^2 d^2 \left(-d^2 + e^2 + f^2 \right) \\[4pt] &+ 2 b^2 e^2 \left(\phantom{-}a^2 - b^2 + c^2 \right) + b^2 e^2 \left(\phantom{-}d^2 - e^2 + f^2 \right) \\[4pt] &+ 2 c^2 f^2 \left(\phantom{-}a^2 + b^2 - c^2 \right) + c^2 f^2 \left(\phantom{-}d^2 + e^2 - f^2 \right) \\[4pt] &- 2 b^2 c^2 d^2 - 2 a^2 c^2 e^2 - 2 a^2 b^2 f^2 \end{align}$$


Isogonal Conjugates, and the Symmedian Point ($K$)

According to the Forum Geometricorum article "Isogonal Conjugates in a Tetrahedron" (PDF) by Sadek, et al., if isogonal conjugates have barycentric coordinates $(\rho,\alpha,\beta,\gamma)$ and $(\rho^\prime, \alpha^\prime, \beta^\prime, \gamma^\prime)$, then

$$\frac{\rho\rho^\prime}{W^2} = \frac{\alpha\alpha^\prime}{X^2} = \frac{\beta\beta^\prime}{Y^2} = \frac{\gamma\gamma^\prime}{Z^2}$$

(This result reconfirms that the Incenter ($I$) is its own isogonal conjugate. But I digress ...) Thus, $$\rho^\prime : \alpha^\prime : \beta^\prime : \gamma^\prime \;=\; \frac{W^2}{\rho} : \frac{X^2}{\alpha} = \frac{Y^2}{\beta} = \frac{Z^2}{\gamma}$$

Since the Symmedian Point ($K$) is the isogonal conjugate of the centroid ($G$), its barycentric coordinates satisfy

$$\rho : \alpha : \beta : \gamma \;=\; W^2 : X^2 : Y^2 : Z^2$$


Monge Point ($M$)

$$\begin{align} \rho &=\phantom{+}a^2d^2 \left(-a^2 + b^2 + c^2\right) \\[4pt] &\phantom{=\,}+ b^2 e^2 \left(\phantom{-}a^2 - b^2 + c^2\right) \\[4pt] &\phantom{=\,}+ c^2 f^2 \left(\phantom{-}a^2 + b^2 - c^2\right) \\[4pt] &\phantom{=\,}+ d^2 e^2 f^2 - b^2 c^2 d^2 - a^2 c^2 e^2 - a^2 b^2 f^2 \end{align}$$


Spieker Point ($S$)

The Spieker Point is the incenter of the medial tetrahedron (ie, the tetrahedron whose vertices are the centroids of the faces) of $PABC$.

By the Incenter formula above, $S$ is given by

$$\frac{W^\prime\,P^\prime + X^\prime\,A^\prime + Y^\prime\,B^\prime + Z^\prime\,C^\prime}{W^\prime + X^\prime + Y^\prime + Z^\prime}$$

for the medial tetrahedron with vertices $P^\prime = \frac13(A+B+C)$, etc, and face-areas $W^\prime = \frac19 W$, etc. This can be re-written as

$$\frac{\frac13(X+Y+Z)\,P + \frac13(W+Y+Z)\,A + \frac13(W+X+Z)\,B + \frac13(W+X+Y)\,C}{W + X + Y + Z}$$ from which we observe $$\rho : \alpha : \beta : \gamma \;=\; X + Y + Z : W + Y + Z : W + X + Z : W + Z + Y $$


Orthocenter ($H$)

In general, a tetrahedron does not have an orthocenter. Equivalent conditions that force a tetrahedron to be orthocentric are $$a^2 + d^2 = b^2 + e^2 = c^2 + f^2 $$ $$\cos A \cos D = \cos B \cos E = \cos C \cos F$$ $$\overrightarrow{PA}\perp\overrightarrow{BC} \qquad \overrightarrow{PB}\perp\overrightarrow{CA} \qquad \overrightarrow{PC}\perp\overrightarrow{AB}$$

Note: If any two orthogonality conditions hold, then the third one rides for free.

Symmetric-looking barycentric coordinates for the orthocenter of such a tetrahedron are a bit elusive. I'll have to come back to this.



Edit. (17 October, 2023. Five years later!) I'm expanding this compilation with a couple of items from OP's recent questions.


Tangential Tetrahedron (question)

Let $P'A'B'C'$ be a tetrahedron tangent to the circumsphere of $PABC$ at its vertices, with $P'$ opposite the face tangent to $P$, etc. Barycentric coordinates of $P'$ are ...

$$\begin{align} -2d^2e^2f^2 &:\,d^2(-a^2d^2+b^2e^2+c^2f^2) \\ &:\,e^2(\phantom{-}a^2d^2-b^2e^2+c^2f^2) \\ &:f^2(\phantom{-}a^2d^2+b^2e^2-c^2f^2) \end{align}$$


Extangents Tetrahedron (mentioned in comments to this question)

Let $P'A'B'C'$ be the "outer" tetrahedron such that plane $A'B'C'$ is tangent to the exspheres of $PABC$ opposite vertices $A$, $B$, $C$; etc. Barycentric coordinates of $P'$ are ...

$$\begin{align} - W(3 + \cos A + \cos B + \cos C ) &: X(1 + \cos D + \cos B + \cos C ) \\ &: Y(1 + \cos A + \cos E + \cos C ) \\ &: Z(1 + \cos A + \cos B + \cos F ) \end{align}$$

The shaded tetrahedron is the extangents tetrahedron:

enter image description here

Blue
  • 75,673
  • These are brilliant. I usually substitute Monge for the Orthocenter. I believe when H=M when H exists. The Fermat point is on my want list. Interesting that K is the 2nd power point.
    We can calculate the Euler line. In a triangle, the incenter can be added and the Euler line and 3 induced Euler lines concur. Are there points like that in a Tetrahedron?
    – Ed Pegg Aug 15 '18 at 19:26
  • Also need to look at anticomplementary, contact, excentral, extangents, extouch, incentral, medial, orthic, and tangential tetrahedra. – Ed Pegg Aug 15 '18 at 19:31
  • So there exists a tetrahedral analogue to the “isogonal conjugate”, but what about the “isotomic conjugate” (as well as the “cyclocevian conjugate”)? – user688486 Feb 15 '24 at 04:53
  • @user688486: The Wolfram Function Repo defines IsotomicConjugate. In bary-coords, the conj of $U=(r:s:t:u)$ is $U'=(1/r:1/s:1/t:1/u)$. Geometrically, cevians through $U$ and $U'$ from a tet vertex meet the opposite face in pts that are in turn isotomic conjugates with respect to that face. (This is analogous to the triangle case, where the cevians from a vertex meet the opposite edge in "isotomic conjugates" (having reciprocal bary-coords) with respect to that edge.) ... I haven't checked about cyclocevians. – Blue Feb 15 '24 at 09:23
  • @Blue Thanks! Strangely, whereas the trilinear coordinate system is much more popular than the barycentric coordinate system in Euclidean triangle geometry, the homologous tetraplanar coordinate system has almost never (Why?) been as popular as the corresponding barycentric coordinate system in ℝ³. Can those formulae be generalized to tetraplanar coordinates as well? – user688486 Feb 15 '24 at 13:17
  • @user688486: My system of choice is barycentric (for both triangles and tetrahedra); treating a point as a weighted average of vertices is simple and intuitive. (That the ETC tends to favor trilinears makes consulting, say, the glossary more work for me. This may someday drive me to switch allegiances.) Tetraplanars may be less common for whatever-reason, but they work. So, one can indeed use them to capture notions of parallelism, concurrency, etc. – Blue Feb 15 '24 at 14:37
  • In addition, although Kimberling's points (centers and bicentric pairs) are the centre of attention for most people at present, algebraic curves (central lines, inconics, circumconics, triangle cubics, etc.) are getting more and more regard in planar triangle geometry. In accordance with the Wikipedia article, both the Euler line and the nine-point circle can be generalized. In view of these, can something like the Feuerbach hyperbola, the Neuberg cubic, and the Steiner deltoid be generalized likewise? – user688486 Feb 16 '24 at 08:29
  • @Blue Many thanks. I notice that MathWorld also refers to the tripolar coordinate system; can the cited "Euler 1786" equation be extended to quadripolar coordinates on tetrahedra? Moreover, as the Desargues's theorem (as well as the Menelaus's theorem and the Ceva's theorem) can be extended into 3D, can the trilinear pole and the trilinear polar also be extended to points and planes associated with tetrahedra? – user688486 Feb 16 '24 at 08:31
  • @user688486: Some things get just a little bit trickier with tetrahedra. Concurrencies are harder to guarantee (eg, no orthocenter). Circumquadrics require five non-vertex pts compared to triangle circumconics needing only two, and higher-degree surfaces require even more. So, how (and whether) things generalize seems to require consideration on a case-by-case basis. ... But at least trilinear poles and polars generalize directly, with comparable formulas. ... (Extended discussion warning. This may be my last reply.) – Blue Feb 16 '24 at 19:24
  • 1
    Thanks very much! Given that the orthocentre (point of concurrence of three lines) has to be generalised to the Monge point (intersection of six or four planes), I reckon that the concept of planes effectively makes more sense than that of lines in modern tetrahedron geometry. – user688486 Feb 16 '24 at 20:54
  • It is said that in a tetrahedron, there actually exists the second symmedian point (although the authors use the term “Lemoine point”). Reasons: Note that in a triangle, the centroid minimizes the sum of squared (Euclidean) distances between itself and each vertex, and the symmedian point minimizes that between itself and each side; in a tetrahedron, the centroid is still the point with a minimal sum of squared distances from the (four) vertices, yet now there are two objects, i.e., faces and edges …. How about the other Lemoine point? – user688486 Mar 11 '24 at 19:00
  • @user688486: "How about the other Lemoine point?" I don't have a ready answer for that, and finding one seems a bit computationally difficult. I'll have to give it some thought. ... I'll double-up on this comment to note that our previous exchange inspired me to do a separate investigation of the Monge pt: I've determined that any circumhyperboloid whose intersections with the faces are rectangular hyperbolas has its center at the Monge pt. This is comparable —though not directly analogous— to the fact that a circumhyperbola of a triangle is rectangular iff it contains the orthocenter. – Blue Mar 12 '24 at 17:52
3

Inspired by this previous question by OP, I investigated "hedronometric" (face-area-based) counterparts of a triangle's Congruent Isoscelizers Point and Congruent Parallelians Point. I wrote a note about them.

In the following, barycentric coordinates $\rho$, $\alpha$, $\beta$, $\gamma$ describe the point $$\frac{\rho\,P + \alpha\,A + \beta\,B + \gamma\,C}{\rho + \alpha + \beta + \gamma} \tag{$\star$}$$ The corresponding trilinear (er, um, tetraplanar) coordinates of the point derive by dividing each barycentric by the area of the appropriate opposite face (here, $W$, $X$, $Y$, $Z$):

$$\frac{\rho}{W} : \frac{\alpha}{X} : \frac{\beta}{Y} : \frac{\gamma}{Z} \tag{$\star\star$}$$

Equal-Area Parallelians Point

Definition and Theorem. A parallelian of tetrahedron $PABC$ with respect to vertex $P$ is a triangle, with vertices on the edge-lines through $P$, in a plane parallel to that of $\triangle ABC$. Any non-degenerate tetrahedron admits a unique point common to the planes of four equal-area parallelians (one for each vertex); this is the Equal-Area Parallelians Point (EAPP).

The barycentric coordinates of the EAPP are $$\rho =-\frac{2}{\sqrt{W}}+\frac{1}{\sqrt{X}}+\frac{1}{\sqrt{Y}}+\frac{1}{\sqrt{Z}}\qquad \alpha =\phantom{-}\frac{1}{\sqrt{W}}-\frac{2}{\sqrt{X}}+\frac{1}{\sqrt{Y}}+\frac{1}{\sqrt{Z}}\qquad\text{etc} \tag{1}$$


Equal-Area Trisohedralizers Point

Definition and Theorem. A trisohedralizer of a tetrahedron $PABC$ with respect to vertex $P$ is a triangle $\triangle A^\prime B^\prime C^\prime$, with vertices on the edge-lines through $P$, that serves as a base for a "trisohedral" tetrahedron. (That is, $|\triangle PB^\prime C^\prime| = |\triangle P C^\prime A^\prime| = |\triangle P A^\prime B^\prime |$.) Any non-degenerate triangle admits a unique point common to the planes of four equal-area trisohedralizers (one for each vertex); this is the Equal-Area Trisohedralizers Point (EATP).

The barycentric coordinates of the EATP are $$\rho = W \left(\;2\sqrt{W/\lambda_P} - \sqrt{X/\lambda_A} - \sqrt{Y/\lambda_B} - \sqrt{Z/\lambda_C}\;\right) \qquad \text{etc} \tag{2}$$ where $$\lambda_V^2 \;=\; 3 - 2\;\sum_{\theta} \cos\theta \tag{3}$$ with the sum taken over the three dihedral angles along edges emanating from vertex $V$.

Ed Pegg
  • 20,955
Blue
  • 75,673