Let, in the projective plane, $A,B,C,D,E$ and $F$ be the intersection points of the sides of a quadrilateral (4 lines, no 3 of which are concurrent with the same point) so that the sides are $ABF$, $ADE$, $DCF$ and $BCE$ and similarly $A',B',C',D',E'$ and $F'$ be the intersection points of the sides of a second quadrilateral so that the sides are $A'B'F'$, $A'D'E'$, $D'C'F'$ and $B'C'E'$. Let $\pi_1=ADE\barwedge A'D'E'$ be the unique projectivity mapping $A$, $D$ and $E$ onto $A'$, $D'$ and $E'$ respectively and similarly $\pi_2=DCF\barwedge D'C'F'$, $\pi_3=ABF\barwedge A'B'F'$ and $\pi_4=BCE \barwedge B'C'E'$. Suppose $l$ is a line not through $D$ and $B$, then $l$ meets $ADE$ in, say, $X_l$ and meets $DCF$ in, say, $Y_l$, they are different points and so are $\pi_1(X_l)$ and $\pi_2(Y_l)$. The same can be said for the intersection points of $l$ with $ABF$, say, $W_l$ and with $BCE$, say, $Z_l$ and their images $\pi_3(W_l)$ and $\pi_4(Z_l)$. Now, the fact that there exists a (necessarily unique) projective collineation mapping the first quadrilateral onto the second (with A to A' etc..) proves that the lines $\pi_1(X_l)\pi_2(Y_l)$ and $\pi_3(W_l)\pi_4(Z_l)$ are the same (in fact the image of $l$ under the projective collineation). I'm pretty sure that there exists a very simple way of proving this, but I'm not seeing it. Anyone with an idea? Thanks in advance.
-
This description is a lot to wade through. Maybe you can provide a diagram? – brainjam Feb 21 '24 at 21:50
-
Also: https://math.stackexchange.com/a/3426395/1257. Is this useful? – brainjam Feb 21 '24 at 21:51
-
Thank you, I worked it out myself, but haven't had the time to post my own answer, will do tonight. It was indeed this same Coxeter proof (I really hate that book) that made me wonder about all of that – Romanda de Gore Feb 23 '24 at 14:51
-
And I should add that this description is not at all 'a lot to wade through' – Romanda de Gore Feb 23 '24 at 15:06
1 Answers
Let $W$ be a point on $ABF$ and $\sigma_{W}$ be the perspectivity with centre $W$ that maps the points of $ADE$ onto the points of $DCF$. $\pi_2 \circ \sigma_{W} \circ \pi^{-1}_1$ is a projectivity that maps the points of $A'D'E'$ onto the points of $D'C'F'$ with invariant $D'$ hence it's a perspectivity $\sigma_{W'}$ with centre $W'$ (not on $A'D'E'$ and not on $D'C'F'$). $\sigma_{W'}(A')=F'$ hence the centre $W'$ lies on $A'B'F'$ and is not equal to $A',F'$. Note that, whenever $X$ is a point on $ADE$ and $Y$ a point on $CDF$ so that $X,Y$ and $W$ are collinear, then we have $\pi_2 \circ \sigma_{W} \circ \pi^{-1}_1(\pi_1(X))=\pi_2(Y)$ hence $W'$ lies on $\pi_1(X)\pi_2(Y)$, so $W'$ is the intersection of $\pi_1(X)\pi_2(Y)$ and $A'B'F'$.
Set $\eta(W)=W'$. The above definition does not work for $A$ and $F$, we set $\eta(A)=A'$ and $\eta(F)=F'$. Let $X$ be a random point on $ADE$, $X \neq A,D$ and let $\sigma_X$ be the perspectivity with centre $X$ mapping points of $ABF$ to the points on $DCF$ and $\sigma_{\pi_1(X)}$ be the perspectivity with centre $\pi_1(X)$ mapping the points on $D'C'F'$ to the points on $A'B'F'$. $\sigma_{\pi_1(X)}\circ \pi_2 \circ \sigma_{X}$ maps a point $W\neq A,F$ on $ABF$ onto the intersection of $\pi_1(X)\pi_2(\sigma_{X}(W))$ with $A'B'F'$. Since $W, X$ and $\sigma_{X}(W)$ are collinear we have $\eta(W)=\sigma_{\pi_1(X)}\circ \pi_2 \circ \sigma_{X}(W)$. Also $\eta(A)=\sigma_{\pi_1(X)}\circ \pi_2 \circ \sigma_{X}(A)=A'$ and $\eta(F)=\sigma_{\pi_1(X)}\circ \pi_2 \circ \sigma_{X}(F)=F'$. Hence $\eta$ is the projectivity $\sigma_{\pi_1(X)}\circ \pi_2 \circ \sigma_{X}$ with $X$ a random point on $ADE$ not equal to $A,D$. Now take $E$ for $X$. We have $\sigma_{\pi_1(E)}\circ \pi_2 \circ \sigma_{E}(A)=A'$, $\sigma_{\pi_1(E)}\circ \pi_2 \circ \sigma_{E}(B)=B'$ and $\sigma_{\pi_1(E)}\circ \pi_2 \circ \sigma_{E}(F)=F'$. Hence $\sigma_{\pi_1(E)}\circ \pi_2 \circ \sigma_{E}=\pi_3$ so $\eta = \pi_3$.
We now have everything we need: $X_l$ on $ADE$, $Y_l$ on $DCF$ and $W_l$ on $ABF$ collinear, so $\eta(W_l)=\pi_3(W_l)$ on $\pi_1(X_l)\pi_2(Y_l)$. We can prove that $\pi_4(Z_l)$, $\pi_1(X_l)$ and $\pi_2(Y_l)$ are collinear in the same way.