2

Let's suppose we have 5 of the 6 dihedral angles of a tetrahedron, and the vertices A,B,C,D. Where $A = (0,0,0)$ and $B = (1,0,0)$. We also limit $C$ to be of the form: $(x_1,x_2,0)$. and lastly $D = (x_3,x_4,x_5)$. Given the 5 known values of 5 of the dihedral angles, we should be able to calculate the 6th dihedral angle and the values $x_i$. But the math gets fairly complicated and I have not been able to find a correct solution.

So the question is: what are these values?

Any help would be greatly appreciated.

Control
  • 162
  • 1
    See here where Cayley-Menger determinant plays a main role. – Jean Marie Jan 11 '24 at 22:44
  • That's really an interesting problem. May be you can add your attempt, or a sketch of your attempt. That would really enhance the quality of your question. – Hosam Hajeer Jan 11 '24 at 23:08
  • @JeanMarie i've read that paper, didn't quite help me not make a bunch of mistakes somewhere. – Control Jan 11 '24 at 23:17
  • @HosamHAJEER I can give you my progress rn: desmos code: yblhponpx2 (idk if i can post links in here) there are 5 equations between a bunch of equal signs, if you can solve those for x1 to x5 that would be lovely – Control Jan 11 '24 at 23:19
  • Yes. Definitely, please edit your question and add these equations. – Hosam Hajeer Jan 11 '24 at 23:48
  • 2
    Consider tet $OABC$ with face-areas $W:=|\triangle ABC|$, $X:=|\triangle OBC|$, $Y:=|\triangle OCA|$, $Z:=|\triangle OAB|$. We have $$\begin{align} W &=;X\cos\angle BC+;Y\cos\angle CA+;Z\cos\angle AB \ X &=W\cos\angle BC+;Y\cos\angle OC+;Z\cos\angle OB \ Y &=;X\cos\angle OC+W\cos\angle CA+;Z\cos\angle OA \ Z &=;X\cos\angle OB+;Y\cos\angle OA+W\cos\angle AB \end{align}$$ (These are analogous to $c=a\cos B+b\cos A$ in a triangle.) Eliminating $W$, $X$, $Y$, $Z$ from the system leaves an expression in just the cosines, so that we can solve for any one from the other five. – Blue Jan 12 '24 at 00:16
  • @Blue Okay so after a bit of time I finally got the right angles for these to work with mine.

    The formulas work, but I don't see how these help me to get x1 to x5. I don't understand what you mean by elimininating W,X,Y,Z, or how you would go about that

    – Control Jan 12 '24 at 03:22
  • 2
    @Control: Eliminate $W$, $X$, $Y$, $Z$ by solving any 3 of the eqns for $X$, $Y$, $Z$ in terms of $W$, and substituting into the 4th eqn; the $W$ cancels, leaving a relation in the dihedral cosines. ... Knowing the dihedral angles, calculate the face angles via a Spherical Law of Cosines; eg, $$\cos BOC = \frac{\cos OA+\cos OB\cos OC}{\sin OB\sin OC}$$ Then, given an edge-length, find others by using the (ordinary) Laws of Sines and Cosines on the triangular faces. Finally, calculate whatever coordinates you desire. – Blue Jan 12 '24 at 04:13
  • @Blue just making sure, isn't it $cos OA - cos OB cos OC$ ? – Control Jan 12 '24 at 04:50
  • @Control, Look here and here, very similar to the formulas you are trying to find. – kabenyuk Jan 12 '24 at 04:57
  • 2
    @Control: It's important to double-check ... I sometimes get the signs mixed-up (esp. when I mix this stuff with hyperbolic trig, which adds its own sign changes). :) ... When applying the Spherical Laws of Cosines to dihedral and face-angles about a vertex of a tetrahedron, the "angles" of the spherical triangle (upper-case in the Wikipedia entry) correspond to the dihedral angles; the "sides" (lower-case) correspond to the face-angles. (To see why: The triangle is effectively the intersection with the sphere of a tet with its vertex at the center.) ... So, "$+$" is indeed correct here. – Blue Jan 12 '24 at 04:58
  • 2
    @Control BTW: If you ever need to sanity-check which sign to use and don't have a reference handy, recall that the corner of a regular tetrahedron has face-angles of measure $60^\circ$ (ie, with cosine $1/2$) and dihedral angles with cosine $1/3$. We have $$\frac12 = \frac{\dfrac13 ;+; \dfrac13\cdot\dfrac13}{\sqrt{1-\left(\dfrac13\right)^2}\cdot\sqrt{1-\left(\dfrac13\right)^2}}=\frac{4/9}{8/9} \qquad \frac13 = \frac{\dfrac12 ;-; \dfrac12\cdot\dfrac12}{\sqrt{1-\left(\dfrac12\right)^2}\cdot\sqrt{1-\left(\dfrac12\right)^2}}=\frac{1/4}{3/4}$$ – Blue Jan 12 '24 at 05:13
  • @Blue Im still not sure if it's actually a "+" the wiki says: cos(c)=cos(a)cos(b)+sin(a)sin(b)cos(C) where C is the face angle solving for the face angle gives me cos(C) = (cos(c)-cos(a)cos(b))/(sin(a)sin(b)) that's a "-"

    and my formula's work with a "-" sign, but not with a "+".

    anyway... I have used a cool fact: $sin(ij)sin(kl)/edge(ij)edge(kl) = sin(ik)sin(jl)/edge(ik)edge(jl)$

    which lets me calculate the 6th angle fairly easily. (this is equivalent to the polar law of sines).

    (sorry for the horrible mathjax, it's almost 7am i pulled an all nighter)

    – Control Jan 12 '24 at 05:36
  • @Control: In $\cos c=\cos a \cos b+\sin a\sin b\cos C$, the "$C$" is an angle of the spherical triangle, not a face-angle of the tet corner. ... In the tet context, we have a corner with its vertex at the (unit) sphere's center, cutting through the sphere to make the spherical triangle you see. The sides of that triangle match the corner's face-angles, because they're unit-radius circular arcs subtending those face-angles; the angles of that triangle match the corner's dihedral angles. ... FYI: The "cool fact" exploits the Spherical Law of Sines $\sin A/\sin a=\sin B/\sin b=\sin C/\sin c$. – Blue Jan 12 '24 at 05:58

1 Answers1

2

As you did, take $A = (0, 0, 0), B = (1, 0, 0), C = (x_1, x_2 , 0) $, and finally $D = (x_3, x_4, x_5) $.

There are four faces: $ABC$, $ABD$, $BCD$, $CAD$, with unit outward normal vectors $n_1, n_2, n_3, n_4 $. Let assume that $x_5 \gt 0 $, so we can take

$ n_1 = (0, 0, -1) $

The minus sign comes from the fact that we're taking the unit normal to point outward of the tetrahedron.

Now suppose $\theta_1$ is the first given dihedral angle between face $ABC$ and $ ABD $. Let $n_2$ be the outward unit normal vector of $ABD$, then $n_2$ is a rotation of $n_1$ about the $x$ axis by an angle of $\pi - \theta_1$, therefore,

$n_2 = (0, \sin( \pi - \theta_1) , - \cos( \pi - \theta_1) ) = (0, - \sin(\theta_1), \cos( \theta_1) ) $

Let $n_3$ be the outward unit normal vector of $BCD$, then $n_3$, and let $\theta_2$ be the dihedral angle between $ABC$ and $BCD$, then $n_2$ can be viewed as a rotation about the $x$ axis by $\pi - \theta_2$ followed by a rotation by $\phi_1 \in (\dfrac{\pi}{2}, \pi ) $ about the $z$ axis applied to the vector $n_1$. This leads to the expression for $n_2$:

$n_3 = R_z( \phi_1 ) R_x (\pi - \theta_2 ) [0, 0, -1]^T $

Hence,

$ n_3 = Rz ( \phi_1) [ 0, - \sin(\theta_2) , \cos ( \theta_2 ) ]^T \\ = [ \sin(\phi_1) \sin(\theta_2) , - \cos(\phi_1) \sin(\theta_2) , \cos(\theta_2) ]^T $

In exactly the same way, we deduce that $n_4$ the outward normal to face $CAD$ is

$n_4 = [ \sin(\phi_2) \sin(\theta_3) , - \cos(\phi_2) \sin(\theta_3) , \cos(\theta_3) ]^T $

where $ \phi_2 \in ( - \pi , - \dfrac{\pi}{2} ) $

These are the expressions for all the normals, and they are all unit vectors pointing out of the tetrahedron surface.

Next, we want to find $\phi_1 $ and $\phi_2$ using the given angles $\theta_4$ (the dihedral angle between $ABD$ and $BCD$ ) and $\theta_5$ (the dihedral angle between $ABD$ and $CAD$ ).

For that we have the following:

$ \cos(\pi - \theta_4) = - \cos(\theta_4) = n_2 \cdot n_3 $

And similarly,

$ - \cos(\theta_5) = n_2 \cdot n_4 $

It is straightforward to solve these two equations for $\phi_1$ and $\phi_2$ given that $\phi_1 \in (\dfrac{\pi}{2} , \pi )$ and $ \phi_2 \in (- \pi , - \dfrac{ \pi}{2} ) $

Now the angle $\theta_6$ between $BCD$ and $CAD$ follows, it is equal to

$ \theta_6 = \pi - \cos^{-1}( n_3 \cdot n_4 ) $

In addition,

$ \angle CBA = \pi - \phi_1 $

$ \angle CAB = \pi + \phi_2 $

$ \angle ACB = - \pi + \phi_1 - \phi_2 $

It follows that

$ AC = AB \left(\dfrac{ \sin \angle CBA }{\sin \angle ACB } \right) $

Therefore,

$ x_1 = AC \cos \angle CAB $

$ x_2 = AC \sin \angle CAB $

To find $(x_3, x_4, x_5)$ we just intersect the three planes $ABD$, $BCD$ and $CAD$.

Hosam Hajeer
  • 21,978
  • wonderful answer, I hadn't even thought about this! My question would be why the angles $$\angle CBA , \angle CAB , \angle ACB$$ can be calculated like that?

    and if that is the case, what would the remaining face angles be?

    or more specifically my main question is;

    is there a nice way to calculate the edge lengths?

    – Control Jan 18 '24 at 16:57
  • These formulas for $\angle CBA , \angle CAB, \angle ACB $ are consistent with the expressions for $n_3$ and $n_4$. – Hosam Hajeer Jan 18 '24 at 17:16