Theorem.

Any line l divides the plane into two parts such that a line L connecting a point P on one
side of l to Q on the other side of l must intersect l.

l(t)
= i#‰


·

+ t · #‰  #‰
PP 
#‰
PP 
PP d
PP
N
·
PP 
‰ Q
PP
q
P
# Q
PP
PP 
sP 
PP
#‰ +
PP
P PP
P PP P
PP = 
) ·
(sPPPP
L
P

#‰  L(k) PP
P

· PP
PP
P




#‰ #‰ #‰ #‰ #‰ #‰ #‰
Proof. Take some line l(t) = i + t · d . Let N be some vector such that N ∦ d ∗ . Then, N and d
form a basis.
#‰ #‰ #‰ #‰ #‰ #‰ #‰ #‰ #‰ #‰ #‰
We then define P and Q in terms of d and N ; take some P = i + Pd d + PN N and Q = i + Qd d +
#‰
QN N . We’d like for one to be “on one side of l” and the other to be “on the other side”, so we assert
that PN < 0 and QN > 0.
Note, then, that:
# ‰
P Q = hQ0 − P0 , Q1 − P1 i
#‰ #‰ #‰ #‰ #‰ #‰
= h(i0 + Qd d0 + QN N0 ) − (i0 + Pd d0 + PN N ),
#‰ #‰ #‰ #‰ #‰ #‰
(i1 + Qd d1 + QN N1 ) − (i1 + Pd d1 + PN N1 )i
#‰ #‰ #‰ #‰
= hi0 − i0 , i1 − i1 i
#‰ #‰ #‰ #‰
+ hQd d0 − Pd d0 , Qd d1 − Pd d1 i
#‰ #‰ #‰ #‰
+ hQN N0 − PN N0 , QN N1 − PN N1 i
#‰ #‰
= (QN − PN )N + (Qd − Pd ) d

−−→ #‰ # ‰ #‰
Let L be the line P Q; define L(s) = P + sP Q. Also let N (x) denote the N component of x when
#‰ #‰
expressed in terms of N and d . Then for some s,
#‰ # ‰
N (L(s)) = N (P + sP Q) (1)
#‰ # ‰
= N (P ) + sN (P Q) (2)
#‰ #‰ #‰ # ‰
= N ( i + Pd d + PN lN ) + sN (P Q) (3)
#‰ #‰ #‰ #‰ #‰
= N ( i ) + N (Pd d + PN lN ) + sN ((QN − PN )N + (Qd − Pd ) d ) (4)
= 0 + PN + s(QN − PN ) (5)
= PN + s(QN − PN ) (6)
#‰ #‰
Note that we don’t justify that N ( #‰
a + b ) = N ( #‰
a )+N ( b ) for lines (2) and (4) nor that N (s #‰
a ) = sN ( #‰
a)
for line (2). These will not be justified because the proof may be easily transformed to be more tedious
#‰ #‰
but not use these properties. Note that, for line (5), N ( i ) = 0 since i is on l.
Note that N (P ) = PN < 0 and N (Q) = QN > 0. Note as well that L(0) = P and L(1) = Q. Thus,
(N ◦ L)(0) < 0 and (N ◦ L)(1) > 0. Since N ◦ L is continuous, then by the Intermediate Value Theorem,
∃k | (N ◦ L)(k) = 0.
#‰ #‰ #‰ #‰ #‰
Since N and d form a basis, let us write L(k) as i + a d + bN for some a and b. We know that
#‰ #‰
N (L(k)) = 0, so b = 0 and L(k) = i + a d . Thus, L(k) = l(a) for some a, and is therefore on l. 
#‰
∗N #‰
certainly exists; let N = h−d0 , −d1 i.