# Elliptic curve Montgomery in projective form questions

Can someone please explain to me why don't we need the y-coordonnate in projective form, and what does this $Z_{m−n}$ in the following formulas mean (from Wikipedia) ? How can we get this value ?

$X_{m+n} = Z_{m−n}((X_m - Z_m)(X_n+Z_n)+(X_m + Z_m)(X_n-Z_n))^2$

Moreover, i would like to know why does the wikipedia article say that "there is no distinction between the affine points $( x , y )$ and $( x , − y )$ because they are both given by the point $( X : Z )$", whereas the projective coordonnates represent the lines of the plane passing through $(0,0)$ , and the lines passing through $((0,0) , (x,y))$ and $((0,0) , (x,-y))$ are not the same for me.

Thank you.

• It looks like the points in Montgomery form are specified by only the $x$ coordinate $(X:Z)$, so although $(x,y)$ and $(x,-y)$ are different points they become indistinguishable using the $(X:Z)$ representation. You can solve for the $y$-coordinate(s) from the equation for the curve given the $x$ coordinate. May 31, 2017 at 12:17
• For a given point $P$ the notation $X_n$ represents the first coordinate of the point $[n]P$ in projective coordinates while $Z_n$ represents the last coordinate. $Z_{m-n}$ is the third coordinate of the point $[m-n]P$. May 31, 2017 at 12:24

The projective coordinate pair $(X_n:Z_n)$ identifies the set of the two points $P_{\pm n}=\pm nP_1=nP_{\pm1}$. The set of the omitted projective coordinates $\{\pm Y_n\}=\{Y_{\pm n}\}$ is implicitly determined by the curve equation. The important point is that if $nP_1=O$, then $-nP_1=O$ as well, and $O$ is unambiguously identified with $(X_O:Z_O)=(0:0)$.
Adding $P_{\pm m}$ and $P_{\pm n}$ determines the set $\{P_{\pm(m+n)},P_{\pm(m-n)}\}$, but you cannot tell which two-point subset is which unless you already know one two-point subset and therefore can deduce the other. This is what happens here: By keeping track of $(X_{m-n}:Z_{m-n})$ besides $(X_n:Z_n)$ and $(X_m:Z_m)$, Montgomery uniquely deduces $(X_{m+n}:Z_{m+n})$.
In the binary Montgomery ladder algorithm, $|m-n|=1$, so $(X_1:Z_1)$ needs to be kept around. Each ladder step then computes the pair $\bigl((X_{2n}:Z_{2n}),(X_{2n+1}:Z_{2n+1})\bigr)$ or $\bigl((X_{2n+1}:Z_{2n+1}),(X_{2n+2}:Z_{2n+2})\bigr)$ from the pair $\bigl((X_n:Z_n),(X_{n+1}:Z_{n+1})\bigr)$: One $(X:Z)$ entry of the latter pair gets replaced by that for the point sum $(X_{2n+1}:Z_{2n+1})$, thereby involving $(X_1:Z_1)$; the other entry is subjected to point doubling, thereby involving the curve parameter $A$.