It appears that it does not matter how you move $A$, $B$, $C$, $D$, so long as $A$ and $D$ remain larger than $B$ and $C$, then you obtain the intersection between the three lines as shown ($y = x$, $[(A,B),(C,D)]$, $[(A,C),(B,D)]$).
2 Answers
Your construction can be rephrased in a more geometrical way as follows: given a triangle $ABC$, we construct a triangle $DBE$ having angle $\angle B$ in common with $ABC$ (in your case $\angle B$ is a right angle, but that is not necessary) and $AD=BD-BA=BC-BE=EC$ (see diagram below). Lines through $A$ and $D$, parallel to $BC$, meet lines though $E$ and $C$, parallel to $AB$, at $G$ and $F$ respectively. We want to prove that point $H$, where $AC$ and $DE$ meet, lies on line $FG$.
From point $H$ draw lines $HI$ and $HJ$, parallel to $BC$ and $AB$. By the intercept theorem, if we prove that $AI=EJ$ then $H$ lies on line $FG$.
Observe now that triangles $AIH$ and $HJC$ are similar, as well as triangles $DIH$ and $HJE$. We have then: $$ AI:HI=HJ:JC \quad\hbox{and}\quad DI:HI=HJ:JE, $$ whence: $$ HI\cdot HJ=AI\cdot JC=DI\cdot JE. $$ From that we get: $$ AI\cdot (JE+EC)=(AI+AD)\cdot JE, \quad\hbox{that is:}\quad AI\cdot EC=AD\cdot JE. $$ From the last equality, recalling that $EC=AD$, we obtain then $AI=JE$, which is what we wanted to prove.
- 50,470
- 4
- 42
- 77
The proof is easy with standard vector algebra: The line from (B,D) to (A,C) is parametrized by point coordinates (x,y) = (B,D)+t(A-B,C-D), 0<=t<=1, and the line from (C,D) to (A,B) is parametrized by point coordinates (x,y) = (C,D)+t'(A-C,B-D), 0<= t'<=1. The intersection between the two lines is found by equating the x and y coordinates of the two lines, solving B+t(A-B)=C+t'(A-C) together with D+t(C-D)=D+T'(B-D), which is a simple 2x2 linear system of equations. The solution is t=(D-B)/(A-C+D-B), t'=(D-C)/(A-C+D-B). Insertion of t and t' back into the line equations gives (B,D)+t(A-B,C-D) = [(AD-CB)/(A-C+D-B), (AD-CB)/(A-C+D-B)]. So indeed the intersection stays always on the diagonal because the x and y coordinates are always the same.
-
Welcome to math.stackexchange and thank you for participating. We encourage you to use the help feature if you have any question about how the site operates. We do encourage new users to learn at least the minimum mathematical formatting features of MathJax. A brief tutorial of the most common features can be found at https://math.meta.stackexchange.com/questions/5020/mathjax-basic-tutorial-and-quick-reference – John Wayland Bales Sep 07 '17 at 16:05
-
I know very well what the proof is. That's how I came across it anyway. The question is whether there is some geometric principle behind this that makes it obvious, i.e. a purely Eucliean-geometry style proof, not the algebraic version. – Project Book Sep 07 '17 at 21:25

