Bounds of sub-triangles
Amalia Adlerteg, Linus Carlsson · 2025
Read Bounds of sub-triangles by Amalia Adlerteg, Linus Carlsson (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-03749.txt.
Introduction
Cut a triangle by its longest side, again and again, and the angles of the shards could in drift run wild. Amalia Adlerteg and Linus Carlsson bound that drift in 2025 for a three-way cut. Their method splits the longest side in thirds and joins the cuts to the far corner, then repeats on each shard. The work is drawn from Adlerteg's thesis work.
For a general start, the top angle of any later shard stays under a fixed multiple of the start's top angle, part proved and part shown by number runs. Two clean starts, the equal-sided and the right isoceles, earn full two-sided bounds. Mesh builders need such bounds: shard angles gone wild spoil the sums that ride on the mesh.
Catalog record
| CREATOR | Amalia Adlerteg, Linus Carlsson |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 14 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-03749
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Bounds of sub-triangles Amalia Adlerteg, Linus Carlsson Abstract We show upper and lower bounds for angles in iterations of trisections of certain triangulations. 1 Introduction Mesh refinement has been of interest for several decades, see e.g. Bänsch [1] , Perdomo and Plaza [4] , Korotov et al. [3] . To estimate the numerical solutions in, for example, the finite element method, there is a need to understand how the angles in each mesh-element can vary. Lower and upper bounds have been established for the longest edge bisection method, see e.g. Gutierrez et al. [2] . We will study the behaviour of the angles when using the longest edge trisection (LE3), which is an iterative method explained in the next section. We assumethat a triangle has the angles \alpha , \beta , and \gamma , from smallest to largest. When performing the LE3, we create a family of new triangles which is denoted \Gamma . The results in this paper is an extract of some of the results in Amalia Adlerteg’s master thesis, to be presented later this year. We present the following findings: • In the general case, we numerically and partially prove that \gamma^{\prime}\leq\gamma\,\frac{\arccos\left(\frac{-5}{2\sqrt{7}}\right)}{\pi/3}, where \gamma^{\prime} is the largest angle in any triangle T\in\Gamma , see Theorem 4 and Theorem 5 . • Starting with \alpha=\beta=\gamma=\pi/3 it follows that any angle v\in T\in\Gamma satisfies 1 1 1 The lower estimate is given in [ 4 ] . \arctan{\left(\frac{\sqrt{3}}{11}\right)}\leq v\leq\arccos{\left(\frac{-5}{2\sqrt{7}}\right)}, see Theorem 4 and Theorem 5 . • Starting with \alpha=\beta=\pi/4 and \gamma=\pi/2 it follows that any angle v\in T\in\Gamma satisfies 2 2 2 The lower estimate is given in [ 4 ] . \frac{3}{4}\,\arctan{\left(\frac{\sqrt{3}}{11}\right)}\leq v\leq\frac{3}{2}\,\arccos{\left(\frac{-5}{2\sqrt{7}}\right)}, see Theorem 4 and Theorem 5 . 2 Description and preliminary results of the LE3-method 2.1 Longest edge trisection On the longest edge of any triangle there exists two points that divide the edge into three equal parts. If we draw two lines from these points to the opposing vertex we get the longest edge trisection of this triangle. This can be repeated on our new triangles until we are satisfied with the mesh refinement. 2.1.1 Mathematical understanding We can describe any triangle as a complex number, z , on the positive real and imaginary axes by letting the longest edge lay from the origin to the point 1+0i and let the shortest edge be the vector given by z . In this paper we will refer to this description of any triangle as the normalized form of it. This gives us a closed space where these triangles are defined, namely the space of triangles given by \Sigma=\{z\ |\ \Imaginary(z)>0,\ \Real(z)\leq 1/2,\ |z-1|\leq 1\} . We define W_{\textrm{L}} as the conformal mapping that transforms the left triangle obtained by LE3 into its normalized form. The conformal mappings W_{\textrm{M}} and W_{\textrm{R}} are defined similarly for the middle and right triangle respectively. From a starting triangle defined by complex z , the orbit of that starting point, \Gamma_{z} , contains the normalized triangles obtained from LE3 of all subsequent triangles. There are three points in \Sigma whose orbits consists of finite number of distinct points. They are \omega_{1}=\frac{1}{3}+\frac{\sqrt{2}}{3}i , \omega_{2}=\frac{1}{3}+\frac{\sqrt{2}}{6}i and \omega_{3}=\frac{4}{9}+\frac{\sqrt{2}}{9}i which all have the orbit \Gamma_{\omega}=\{\omega_{1},\omega_{2},\omega_{3}\} . 2.1.2 Conformal mappings To normalize a triangle we can use Möbius functions since they preserve angles. The Möbius transformation that normalizes a triangle, defined by complex z , is dependent on where z is located on \Sigma . For example, if |z-2/3|\leq 1/3 then the longest edge of the right triangle is located differently than if z was outside of that area. Thus, W_{\textrm{R}} is a piecewise function defined on two areas and, with similar reasoning, both W_{\textrm{L}} and W_{\textrm{M}} are also piecewise functions. If f(z)=\frac{az+b}{cz+d} , then f can be derived from the composition of these simple transformations: • translation : f(z)=z+c , c is a complex constant. • rotation : f(z)=\lambda z , \lambda is a complex constant such that |\lambda|=1 . • magnification : f(z)=\rho z , \rho is a real strictly positive number. • inversion : f(z)=1/z . Figure 1: A right triangle after trisection of the original triangle defined by z , such that |z-2/3|\leq 1/3 . Say that |z-2/3|\leq 1/3 and we want to normalize the right triangle illustrated in Figure 1 . We can start with defining a translation that moves the point z to the origin, that is f_{1}(\zeta)=\zeta-z. (1) Figure 2: The triangle after the translation given by Equation ( 1 ). Then we define a rotation that moves the point v , illustrated in Figure 2 , to the real axis. The resulting triangle is displayed in Figure 3 . The rotation is given by f_{2}(\zeta)=\zeta\cdot\frac{\overline{v}}{|v|}=\zeta\cdot\frac{\overline{f_{1}(1+0i)}}{|f_{1}(1+0i)|}=\zeta\cdot\frac{1-\overline{z}}{|1-z|}. (2) Figure 3: The triangle after the rotation given by Equation ( 2 ). Now we define an inversion that mirrors all points on the real axis, the result of which is illustrated in Figure 4 . The inversion is given by taking the conjugate, that is f_{3}(\zeta)=\overline{\zeta}. (3) Figure 4: The triangle after the inversion given by Equation ( 3 ). Finally we define a magnification that enlarges the triangle so that w , displayed in Figure 4 , is at 1+0i , that is f_{4}(\zeta)=\frac{\zeta}{|w|}=\frac{\zeta}{|f_{3}\circ f_{2}\circ f_{1}(1+0i)|}=\frac{\zeta}{|1-z|}. (4) Figure 5: The triangle after the magnification given by Equation ( 4 ). The final triangle after our simple transformations is illustrated in Figure 5 . Now the complex number that describes the normalized triangle is originally at 2/3+0i , so W_{\textrm{R}}(z)=f_{4}\circ f_{3}\circ f_{2}\circ f_{1}\left(\frac{2}{3}+0i\right)=\frac{3\overline{z}-2}{3\overline{z}-3}. For the other transformations see Perdomo and Plaza [4] . 2.2 Poincaré half plane metric The metric of the Poincaré half-plane model on the half-plane \{x+iy\ |\ y>0\} is given by (ds)^{2}=\frac{(dx)^{2}+(dy)^{2}}{y^{2}}, where s measures the local length along a line. The length of a curve is therefore given by l(\gamma)=\int_{\gamma}ds. With the parametrisation \gamma:[0,1]\to\mathbb{C} such that \gamma(t)=x(t)+iy(t) , we get that \frac{ds}{dt}=\frac{1}{dt}\,\sqrt{\frac{(dx)^{2}+(dy)^{2}}{y^{2}}}=\frac{1}{y}\,\sqrt{\frac{(dx)^{2}+(dy)^{2}}{(dt)^{2}}}=\frac{\sqrt{(\frac{dx}{dt})^{2}+(\frac{dy}{dt})^{2}}}{y}. With this parametrisation the length of the curve is l(\gamma)=\int_{\gamma}ds=\int_{0}^{1}\frac{ds}{dt}\,dt=\int_{0}^{1}\frac{\sqrt{(\frac{dx}{dt})^{2}+(\frac{dy}{dt})^{2}}}{y}\,dt. The hyperbolic distance, d , between z_{1} and z_{2} in the Poincaré half-plane is given by \cosh(d)=1+\frac{|z_{1}-z_{2}|^{2}}{2\cdot\Imaginary(z_{1})\cdot\Imaginary(z_{2})}. (5) In this space a geodesic is either a vertical line if the two points have the same real value, otherwise it is a circle arc that is centred on the real axis. Our derivations and simulations in this paper will rely heavily on the following Lemma, proved and presented as Lemma 3 in Perdomo and Plaza [4] . Lemma 1 (Non-increasing property) Let d(\cdot,\cdot) denote the hyperbolic distance in the Poincaré half-plane. For every z_{1} and z_{2} in the space of triangles, \Sigma , we have d(W_{j}(z_{1}),W_{j}(z_{2}))\leq d(z_{1},z_{2}),\hskip 5.69054ptj\in\{\textrm{L, M, R}\}. We end this section by defining three hyperbolic circles, that is C_{1} , C_{2} and C_{3} , with radius \ln(\sqrt{2}) and with hyperbolic centres \omega_{1} , \omega_{2} and \omega_{3} , respectively. 3 Simulations of orbits In this section we construct a fast method for orbit approximation, where one orbit is of special interest. Lemma 2 The orbit of z_{\textrm{eq}}=1/2+i\sqrt{3}/2 is \displaystyle\Gamma_{\textrm{eq}}=\mathcal{C}\bigcup\left\{\begin{array}[]{l}\frac{1}{2}+\frac{\sqrt{3}}{2}i,\frac{1}{14}+\frac{3\sqrt{3}}{14}i,\frac{1}{6}+\frac{\sqrt{3}}{6}i,\frac{5}{26}+\frac{3\sqrt{3}}{26}i,\frac{13}{42}+\frac{3\sqrt{3}}{14}i,\\ \frac{5}{14}+\frac{3\sqrt{3}}{14}i,\frac{7}{18}+\frac{3\sqrt{3}}{18}i,\frac{29}{62}+\frac{3\sqrt{3}}{62}i,\frac{1}{2}+\frac{\sqrt{3}}{18}i,\frac{1}{2}+\frac{\sqrt{3}}{6}i\end{array}\right\}, where \mathcal{C} lies inside the union of circles C_{1},\,C_{2} and C_{3} . The result in Lemma 2 can be extracted from [ 4 ] . 3.1 The method of random orbit simulation When calculating the orbit of a starting triangle, defined by